Anandtech's preview of Sandy Bridge makes it sound very promising. Even the integrated graphics sounds pretty good.
"The [SB] Core i5 2400 should actually perform like a Core i7 880 despite not having Hyper Threading enabled... [With turbo mode] I'd estimate you can add another 3 - 7%... Not only will Sandy Bridge be noticeably quicker than Lynnfield, it'll draw less power."
"Sandy Bridge's integrated graphics is good. It's fast enough to put all previous attempts at integrated graphics to shame and compete with entry level discrete GPUs. The fact that you can get Radeon HD 5450 performance for free with a Core i5 2400 is just awesome. As I mentioned before, you won't want to throw away your GTX 460, but if you were planning on spending $50 on a GPU - you may not need to with Sandy Bridge."
The main problem with Sandy Bridge has been that there was no word of OpenCL support in the GPU. Now it seems that the graphics will support OpenCL:
"Intel is also working on OpenCL for the graphics part of Sandy Bridge, according to sources.
Intel declined to comment directly on Apple's plans, but regarding OpenCL it would only tell CNET: 'In terms of full product support, we continue to evaluate when and where OpenCL will intercept our various products.' "
Looks very, very good. Good enough that I'll probably finally upgrade from my trusty 2006 Macbook 2,1
Rumors of CpenCL support so far are just that - rumors. Intel has stated that it doesn't support OpenCL. Whether it will in the future is an open question. So, unless Apple is working to do that, perhaps with their help, it won't happen. And as we all know, having Apple say something before the new models come out is just wishful thinking.
This is where I can see Apple throwing its weight around. I can see Apple strong arming Intel into supporting OpenCL, even if it?s special category just for Mac notebook that aren?t found on Intel?s price list. It?s not like there isn?t precedence to support this possibility.
We have to be careful with this, as it's more than software or firmware that makes OpenCL possible on any given gpu. The gpu has to be programmable, and it must support features that will allow OpenCL firmware and software to do what is needed. If that's not possible, then nothing can be done to make it so.
That would mean that Intel would have to make special designs just for Apple, and I don't think Apple wants that. Remember the problems Apple had with latter G4 and G5 chips? That comes from having limited edition chip lines. What if Intel agrees, but can't get them ready in time? Apple is screwed. And as we all know, that happens all the time in chip making. Look at all the problems AMD has had over the years, which is why I don't trust them.
By using what everyone else has, Apple is at least assured that if there are delays, it's for the entire industry, not just them. Otherwise, we could have the problem of waiting months for a machine to come out that should have been out way before. Not good!
The only thing you can do is wait and see. As you may know none (zero) of Intels previous chips supported OpenGL well. So you will need to check out the benchmarks very closely.
Quote:
Originally Posted by Kali
I already tested the current MBP 13" with the apps I'm using, and it's working well. The 9400M that I have in my late 2009 mini is barely working well for my needs (I'm experiencing some textures weirdness with it).
Even worst is that OpenGL demands often go up as programs are updated. So even if the current machine passes with flying colors you may have issues in the future if you don't have enough margin with respect to performance.
Beyound that I do hope you are running the latest drivers in Snow Leopard.
Quote:
I'm using Celestia (full 3D OpenGL astronomy app) with large hi-res textures and some heavy 3D models, to teach astronomy in a classroom with a wall projector. I'm also using several full HD videos in QT.
cool!.
I've always found this science fascinating especially the vastness of space and our current in ability to navigate it.
Quote:
The 15" MBP is too large, for a portable. The 13" is just perfect, sizewise and is not too heavy to bring under the arm, during a full day.
I understand what you are saying but it doesn't look real gold right now if the rumor is true. The lack of decent GPU performance in the 13" is why I wish they would drop the CD drive and put a real GPU in the machine with a larger battery. The MBP could really use such a configuration. On the otherhand simply waiting for Llano from AMD would likely give you the results you are looking for.
In the end you won't know what is up until the hardware is released and you can test your software on it. I wouldn't be to trustful of bench marks either, this is a case where you need to test with your software to see what is happening visually. Drivers play a key role so what doesn't work at release could six months later.
Rumors of CpenCL support so far are just that - rumors. Intel has stated that it doesn't support OpenCL. Whether it will in the future is an open question. So, unless Apple is working to do that, perhaps with their help, it won't happen. And as we all know, having Apple say something before the new models come out is just wishful thinking.
So are you saying Intel's Director of Graphics Architecture is lying?
Quote:
Thomas Piazza, an Intel fellow and director of graphics architecture for the Intel Architecture Group, said that Sandy Bridge-based chips in their current implementation will not support DirectX 11, a Microsoft technology for accelerating multimedia and games. Currently, Sandy Bridge supports DirectX 10.1 and OpenCL 1.1--the latter used on Apple's Mac operating systems, according to Piazza.
Intel's Director of Graphics Architecture has confirmed that Sandy Bridge supports OpenCL 1.1. Apple currently supports OpenCL 1.0 so Apple can at least tout that as an upgrade, although current AMD and nVidia solutions support OpenCL 1.1 as well, if Apple were to release drivers for it. Sandy Bridge's IGP won't likely beat the 320M in raw performance unless Apple really invests in the drivers, but it's certainly 320M class so it'll be a sidegrade. Architecturally, Sandy Bridge's IGP should be superior for OpenCL to both nVidia's off-chip IGP and AMD's Fusion whose IGP connects to the CPU via a crossbar bus, because Sandy Bridge's IGP is on die and shares an L3 cache with the CPU allow efficient low-latency, high-bandwidth data sharing between the IGP and CPU for OpenCL. No longer will the CPU have to set up the OpenCL data and commands, transfer it to the GPU, wait for the GPU to process it, and then wait for the results to be copied back. Everything will be in a common data location that both the IGP and CPU can read and write to. This efficiency should help make up some of the raw performance deficit of Sandy Bridge's IGP for OpenCL.
Intel is actively hiring OpenCL driver developers specifically for their IGP, so OpenCL support is definitely coming, it's just a matter of when and how good it'll be.
As well, Sandy Bridge's CPU already includes enhancements that target many of the same benefits of OpenCL. For example, Sandy Bridge includes a dedicated hardware video encode accelerator which should be faster and more power efficient than running an implementation on OpenCL on a GPU. AVX also greatly improves the CPU's vector and multimedia performance and should be easier for software developers to modify their existing SSE code to adopt rather than writing a new OpenCL path. AMD's Fusion processors support neither in the 2010-2011 timeframe.
Intel's Director of Graphics Architecture has confirmed that Sandy Bridge supports OpenCL 1.1.
As you probably know, the Intel has only officially announced *CPU*-based implementations of OpenCL. If Intel is starting to implement OpenCL support on the GPU (as the recent hiring ad you posted indicates) this is excellent news! But don't claim that they've been supporting it all along.
Quote:
Originally Posted by ltcommander.data
Sandy Bridge's IGP won't likely beat the 320M in raw performance unless Apple really invests in the drivers, but it's certainly 320M class so it'll be a sidegrade. Architecturally, Sandy Bridge's IGP should be superior for OpenCL to both nVidia's off-chip IGP and AMD's Fusion whose IGP connects to the CPU via a crossbar bus, because Sandy Bridge's IGP is on die and shares an L3 cache.
Sounds nice, we'll know for sure when the new Intel and AMD hardware is out.
Quote:
Originally Posted by ltcommander.data
Sandy Bridge includes a dedicated hardware video encode accelerator which should be faster and more power efficient than running an implementation on OpenCL on a GPU.
Oh yeah, I just hope Handbrake developers have got their hands on this!
The main problem with Sandy Bridge has been that there was no word of OpenCL support in the GPU. Now it seems that the graphics will support OpenCL:
"Intel is also working on OpenCL for the graphics part of Sandy Bridge, according to sources.
Intel declined to comment directly on Apple's plans, but regarding OpenCL it would only tell CNET: 'In terms of full product support, we continue to evaluate when and where OpenCL will intercept our various products.' "
Are we talking hardware support for a GPU-based Open GL? Or are we talking about functional drivers for a GPU-based Open GL?
Quote:
Originally Posted by ltcommander.data
So are you saying Intel's Director of Graphics Architecture is lying?
I'm saying I would be embarassed to be out in public with the title "Director of Graphics Architecture for Intel". Everything they've done for years has been sh!t. Why should we believe him now? I say the guy should put up or shut up. Show us real world benchmarks of a shipping product, only then would I be willing to re-evaluate my assumptions of how bad any future Intel IGP will be vs. the competition.
From the cnet article, this is what you call being damned with faint praise: "It will integrate Intel's best graphics chip technology to date directly onto the central processing unit."
As you probably know, the Intel has only officially announced *CPU*-based implementations of OpenCL. If Intel is starting to implement OpenCL support on the GPU (as the recent hiring ad you posted indicates) this is excellent news! But don't claim that they've been supporting it all along.
It is kind of open to interpretation whether Piazza's OpenCL statement was referring to the IGP or just the CPU implementation, although seeing the article's focus is on the IGP, Piazza is the director of the graphics group, and OpenCL was mentioned in context with DirectX, I'd hope he was talking about OpenCL for IGP.
Quote:
Originally Posted by hypercommunist
Sounds nice, we'll know for sure when the new Intel and AMD hardware is out.
Just to avoid getting people's hopes up and getting it quoted back to me later, when I say Intel's IGP/CPU implementation is superior for OpenCL I really do mean architecturally only. I have no idea what the final raw performance numbers are going to be like, but it's a good design idea.
Quote:
Originally Posted by John.B
I'm saying I would be embarassed to be out in public with the title "Director of Graphics Architecture for Intel". Everything they've done for years has been sh!t. Why should we believe him now? I say the guy should put up or shut up. Show us real world benchmarks of a shipping product, only then would I be willing to re-evaluate my assumptions of how bad any future Intel IGP will be vs. the competition.
In fairness, I think the problem with Intel's IGP group is more what they want to do rather than can they do it. From a technical perspective, the GMA X4500MHD had 10 execution units while Arrandale's IGP has 12 execution units with performance doubling. Sandy Bridge looks to double Arrandale's performance with the same number of execution units. I guess that can be interpreted as very inefficient designing on the older IGPs, but it can also point to Intel's IGP really knowing their stuff in order to extract significantly more performance without drastically increasing the resources used to do it. Certainly, AMD and nVidia's method of increasing performance is generally to increase execution units rather than make the existing ones more efficient. In fact, performance analysis of AMD's HD5000 series seems to show that the HD5000's shader units are actually less efficient than the HD4000 units, just that there are more of them.
It's just that up to now, Intel IGP's have been too focused on the corporate market. It's only recently in Windows that everyday computer tasks require decent IGP acceleration, like the Aero GUI, video decode, Flash acceleration, etc. For the rest of the decade, Intel IGPs were sufficient for Windows XP and certainly it didn't seem like corporations were complaining. In the consumer spectrum, computer average selling prices have been falling with consumers expecting to pay less and less for computers. That means that OEMs are generally not able to put in discrete GPUs which increases the presence of Intel IGPs in consumer computers and consumers want to play games on them which Intel IGPs clearly weren't able to do. It's only now that Intel is belatedly addressing this, that computer average selling prices will remain low, most consumer computers will use Intel IGPs, and most consumers will expect decent gaming performance from them, so Intel has to develop a decent IGP capable of gaming. Hopefully we'll know next month whether Intel will be successful with Sandy Bridge.
In fairness, I think the problem with Intel's IGP group is more what they want to do rather than can they do it. From a technical perspective, the GMA X4500MHD had 10 execution units while Arrandale's IGP has 12 execution units with performance doubling. Sandy Bridge looks to double Arrandale's performance with the same number of execution units. I guess that can be interpreted as very inefficient designing on the older IGPs, but it can also point to Intel's IGP really knowing their stuff in order to extract significantly more performance without drastically increasing the resources used to do it. Certainly, AMD and nVidia's method of increasing performance is generally to increase execution units rather than make the existing ones more efficient. In fact, performance analysis of AMD's HD5000 series seems to show that the HD5000's shader units are actually less efficient than the HD4000 units, just that there are more of them.
It's just that up to now, Intel IGP's have been too focused on the corporate market. It's only recently in Windows that everyday computer tasks require decent IGP acceleration, like the Aero GUI, video decode, Flash acceleration, etc. For the rest of the decade, Intel IGPs were sufficient for Windows XP and certainly it didn't seem like corporations were complaining. In the consumer spectrum, computer average selling prices have been falling with consumers expecting to pay less and less for computers. That means that OEMs are generally not able to put in discrete GPUs which increases the presence of Intel IGPs in consumer computers and consumers want to play games on them which Intel IGPs clearly weren't able to do. It's only now that Intel is belatedly addressing this, that computer average selling prices will remain low, most consumer computers will use Intel IGPs, and most consumers will expect decent gaming performance from them, so Intel has to develop a decent IGP capable of gaming. Hopefully we'll know next month whether Intel will be successful with Sandy Bridge.
Great post! Thank you.
Since Apple seems to be headed toward Sandy Bridge, I'm really hoping you're right.
No Apple, DON'T DO IT! I know that for most of your MacBook base of users the crappy graphics from Intel would do the job just fine but don't punish the rest of us, especially after steam etc has finally come to the Mac! \
It is kind of open to interpretation whether Piazza's OpenCL statement was referring to the IGP or just the CPU implementation, although seeing the article's focus is on the IGP, Piazza is the director of the graphics group, and OpenCL was mentioned in context with DirectX, I'd hope he was talking about OpenCL for IGP.
Just to avoid getting people's hopes up and getting it quoted back to me later, when I say Intel's IGP/CPU implementation is superior for OpenCL I really do mean architecturally only. I have no idea what the final raw performance numbers are going to be like, but it's a good design idea.
In fairness, I think the problem with Intel's IGP group is more what they want to do rather than can they do it. From a technical perspective, the GMA X4500MHD had 10 execution units while Arrandale's IGP has 12 execution units with performance doubling. Sandy Bridge looks to double Arrandale's performance with the same number of execution units. I guess that can be interpreted as very inefficient designing on the older IGPs, but it can also point to Intel's IGP really knowing their stuff in order to extract significantly more performance without drastically increasing the resources used to do it. Certainly, AMD and nVidia's method of increasing performance is generally to increase execution units rather than make the existing ones more efficient. In fact, performance analysis of AMD's HD5000 series seems to show that the HD5000's shader units are actually less efficient than the HD4000 units, just that there are more of them.
It's just that up to now, Intel IGP's have been too focused on the corporate market. It's only recently in Windows that everyday computer tasks require decent IGP acceleration, like the Aero GUI, video decode, Flash acceleration, etc. For the rest of the decade, Intel IGPs were sufficient for Windows XP and certainly it didn't seem like corporations were complaining. In the consumer spectrum, computer average selling prices have been falling with consumers expecting to pay less and less for computers. That means that OEMs are generally not able to put in discrete GPUs which increases the presence of Intel IGPs in consumer computers and consumers want to play games on them which Intel IGPs clearly weren't able to do. It's only now that Intel is belatedly addressing this, that computer average selling prices will remain low, most consumer computers will use Intel IGPs, and most consumers will expect decent gaming performance from them, so Intel has to develop a decent IGP capable of gaming. Hopefully we'll know next month whether Intel will be successful with Sandy Bridge.
i have a core i5 laptop at work and the intel HD is more than enough for normal web browsing including youtube. with flash 10.2 beta it's insanely fast at video playback. and the laptop stays cool even when using flash.
not like the old HP POS i had that you could use as a food warmer
It's just that up to now, Intel IGP's have been too focused on the corporate market. It's only recently in Windows that everyday computer tasks require decent IGP acceleration, like the Aero GUI, video decode, Flash acceleration, etc. For the rest of the decade, Intel IGPs were sufficient for Windows XP and certainly it didn't seem like corporations were complaining. In the consumer spectrum, computer average selling prices have been falling with consumers expecting to pay less and less for computers. That means that OEMs are generally not able to put in discrete GPUs which increases the presence of Intel IGPs in consumer computers and consumers want to play games on them which Intel IGPs clearly weren't able to do. It's only now that Intel is belatedly addressing this, that computer average selling prices will remain low, most consumer computers will use Intel IGPs, and most consumers will expect decent gaming performance from them, so Intel has to develop a decent IGP capable of gaming. Hopefully we'll know next month whether Intel will be successful with Sandy Bridge.
Geez - you guys. I'm still gaming on my IBM XT and Apple II, and the text game "Softporn" is just fine with me - who needs fancy graphics? Sheesh.
i have a core i5 laptop at work and the intel HD is more than enough for normal web browsing including youtube. with flash 10.2 beta it's insanely fast at video playback. and the laptop stays cool even when using flash.
not like the old HP POS i had that you could use as a food warmer
What is the battery life difference between using HTML5 video vs. Flash video onYouTube for the same video across different resolutions?
Anandtech's preview of Sandy Bridge makes it sound very promising. Even the integrated graphics sounds pretty good.
Looks very, very good. Good enough that I'll probably finally upgrade from my trusty 2006 Macbook 2,1
The problem there is you are looking at the desktop version of Sandy Bridge so you won't see it in the Macbook. It compares favourably to the desktop Radeon 5450, which is impressive but still only matches the lowest end GPU in the current iMac.
Apple can use the saving from the dedicated GPU to put an SSD in place and the overall iMac power consumption would lower but it will be a side-step in terms of graphics performance at best.
The media conversion hardware looks good but not as impressive as the 400FPS claim:
The demo there was only twice as fast as the software encoding. It does look like it's programmable though so it might be possible to extend beyond just AVC encoding.
The mobile hardware is still uncertain. The good thing is they have stated the mobile GPUs will all have 12EUs and 2 graphics cores where the desktop versions might only have 6 in some models. Anand thought the above desktop version was a 6EU model but it still has no bearing on the mobile part.
1 EU I'd assume is equivalent in description to 1 NVidia core of which the 320M has 48. The number isn't comparable across manufacturers though.
The tested desktop version of Sandy Bridge is faster than the NVidia 320M by about 15-20%. If it was a 6EU model, then it suggests the 12-core mobile GPU might be worthwhile.
Notebookcheck have estimated the position lower than the 320M based on the 5450 mobile/desktop comparisons:
My main concern is when it comes to feature support. Sometimes the raw performance of the GPU is fast enough but when you use hardware-accelerated apps, none of the options are supported whereas they are always supported with NVidia and sometimes supported with AMD.
OpenCL doesn't look likely but we can't tell this until Intel say one way or the other. Given their stance on GPGPU up until now, I'd say it's not going to happen. They have pages about Ct detailing why it's better than GPUs - task parallels, standard code etc.
Apple cannot use AMD right now. LLano isn't coming out until Summer so Fusion is out for now.
I hope Apple will go Sandy Bridge + dedicated but for cost, it's not likely. I guess they will make a choice based on the following:
- is the power consumption better => yes
- is the performance faster than C2D => yes
- will it be fast enough for casual games => probably
- it likely has no GPU OpenCL support but it supports hardware media encode/decode and OpenCL code can be tested on the CPU
There's nothing in there that immediately puts them off using it alone. It's good enough but not so good you will want it over a more expensive model.
What is the battery life difference between using HTML5 video vs. Flash video onYouTube for the same video across different resolutions?
haven't measured it, but on my old POS HP centrino laptop it would get hot just by seeing non-video flash content and the battery would die in 90 minutes instead of the 3 hour rated time. and that was with an ATI HD2600 discrete graphics chip in there. the CPU would spike up to close to 100% and videos would skip frames. full screen was unwatchable
on my Core i5 the battery seems to last pretty close to the rated time when watching flash videos, but i usually have it charging when i'm using it. with flash 10.1 and 10.2 beta the CPU goes to high teens when watching the video in 720p and full screen
haven't measured it, but on my old POS HP centrino laptop it would get hot just by seeing non-video flash content and the battery would die in 90 minutes instead of the 3 hour rated time. and that was with an ATI HD2600 discrete graphics chip in there. the CPU would spike up to close to 100% and videos would skip frames. full screen was unwatchable
on my Core i5 the battery seems to last pretty close to the rated time when watching flash videos, but i usually have it charging when i'm using it. with flash 10.1 and 10.2 beta the CPU goes to high teens when watching the video in 720p and full screen
on my 2010 13” MBP I get about 30% longer battery life with HTML5 video from the same YouTube file on 720p vids with Flash 10.2, which seems 5 to 10% better than Flash 10.1, which was a huge improvement over Flash 10.0. But that’s a very weak way of measuring the difference.
Due to Flash crashes I uninstalled it and went back to Flash 10.1. Not that it matters much since I rarely need to use the Flash plug-in with HTML5 video on pretty much every site I frequent. It is mice that that Adobe adds the uninstaller to the page for the Flash beta installer, instead of making for you hunt it down.
Intel's Director of Graphics Architecture has confirmed that Sandy Bridge supports OpenCL 1.1. Apple currently supports OpenCL 1.0 so Apple can at least tout that as an upgrade, although current AMD and nVidia solutions support OpenCL 1.1 as well, if Apple were to release drivers for it. Sandy Bridge's IGP won't likely beat the 320M in raw performance unless Apple really invests in the drivers, but it's certainly 320M class so it'll be a sidegrade. Architecturally, Sandy Bridge's IGP should be superior for OpenCL to both nVidia's off-chip IGP and AMD's Fusion whose IGP connects to the CPU via a crossbar bus, because Sandy Bridge's IGP is on die and shares an L3 cache with the CPU allow efficient low-latency, high-bandwidth data sharing between the IGP and CPU for OpenCL. No longer will the CPU have to set up the OpenCL data and commands, transfer it to the GPU, wait for the GPU to process it, and then wait for the results to be copied back. Everything will be in a common data location that both the IGP and CPU can read and write to. This efficiency should help make up some of the raw performance deficit of Sandy Bridge's IGP for OpenCL.
Intel is actively hiring OpenCL driver developers specifically for their IGP, so OpenCL support is definitely coming, it's just a matter of when and how good it'll be.
As well, Sandy Bridge's CPU already includes enhancements that target many of the same benefits of OpenCL. For example, Sandy Bridge includes a dedicated hardware video encode accelerator which should be faster and more power efficient than running an implementation on OpenCL on a GPU. AVX also greatly improves the CPU's vector and multimedia performance and should be easier for software developers to modify their existing SSE code to adopt rather than writing a new OpenCL path. AMD's Fusion processors support neither in the 2010-2011 timeframe.
This is from an article very recently in Ars Technica, a site that keeps up with this, and has some very qualified people in the field writing for them. It also follows what Intel has been sayig publically about this issue. so while I'm certainly not saying that this guy is lying, it may not be what you think, or it may be a new development.
Quote:
The IGP in Intel's upcoming Sandy Bridge mobile processors is expected to have vastly improved performance over its Arrandale predecessor, but is still based on relatively ancient architecture and, from what we have been able to gather, is not OpenCL compatible.
Intel's next architecture update, Ivy Bridge, is expected to have a design that is OpenCL compatible, but Intel won't commit to providing OpenCL drivers. "Intel is evaluating when and where OpenCL support will intercept our products, however no announcement has been made," spokesperson Nick Knupffer told Ars. Apple isn't in a position to wait any longer for Intel to step up its IGP game, and NVIDIA likely has designs waiting in the wing for Apple to use once the legal issues are cleared. A settlement now could ultimately be a win-win-win for everyone involved.
And as OpenCl doesn't only work on gpu's, that may also what we're seeing from Intel right now.
on my 2010 13? MBP I get about 30% longer battery life with HTML5 video from the same YouTube file on 720p vids with Flash 10.2, which seems 5 to 10% better than Flash 10.1, which was a huge improvement over Flash 10.0. But that?s a very weak way of measuring the difference.
Due to Flash crashes I uninstalled it and went back to Flash 10.1. Not that it matters much since I rarely need to use the Flash plug-in with HTML5 video on pretty much every site I frequent. It is mice that that Adobe adds the uninstaller to the page for the Flash beta installer, instead of making for you hunt it down.
depends on the browser as well
in the last year or two someone did a test and believe it or not IE came out as the most energy efficient browser. firefox was the worst. forgot where safari, opera and chrome ended up
Not reliable at all. Apple is depending on OpenCL for performance. Otherwise they wouldn't be willing to take the flack for still using Core 2 chips in its lower machines so they can use OpenCL GPU's. I don't see them backing away from that, unless, somehow, they've figured out a way around it. Being that's it's Apple, that's not impossible, but not likely.
Not entirely true. There's little advantage in the i3 over the C2D and the IGP isn't as good as in sandy bridge. In the case of sandy bridge I believe the mobile parts are all i5 for now. The difference in CPU performance with be huge.
As argued back and forth in the other thread OpenCL support IMHO isn't a necessarily a hard requirement for the GPU.
The quick recap of my position is:
1) Intel is supporting OpenCL on the CPU. Therefore OpenCL calls could be implemented by on die hardware in Sandy Bridge (i.e. via 256 bit wide SIMD or the hardware encoder/decoders).
2) Intel is providing on die hardware support/enhancement for the 2 most common GP GPU uses: cryptography and transcoding. How can we guess these are the two most common use cases? Because these are the two GPGPU benchmarks added the SiSoft Sandra benchmark suite. [1][2]
3) The GPU performance as a GPU is slightly better than the 320M as the 5450 is slightly better than the 320M.
Given these three, going with just a Sandy Bridge i5 in the mini, MB and MBP 13" is plausible. The MBA I believe uses lower TDP parts and will simply have to wait.
---
[1] "The demo used Cyberlink?s Media Espresso to convert a ~1 minute long 30Mbps 1080p HD video clip to an iPhone compatible format. On Sandy Bridge the conversion finished in a matter of a few seconds (< 10 seconds by my watch)."
[2] "The interesting bit is that SiSoft expects how CPUs should be able to challenge GPGPUs in the near future, given that "Using 256-bit register width (instead of 128-bit of SSE/2/3/4) yields further performance gains through greater parallelism in most algorithms. Combined with the increase in processor cores and threads we will soon have CPUs rivaling GPGPUs in performance."
Comments
Anandtech's preview of Sandy Bridge makes it sound very promising. Even the integrated graphics sounds pretty good.
"The [SB] Core i5 2400 should actually perform like a Core i7 880 despite not having Hyper Threading enabled... [With turbo mode] I'd estimate you can add another 3 - 7%... Not only will Sandy Bridge be noticeably quicker than Lynnfield, it'll draw less power."
"Sandy Bridge's integrated graphics is good. It's fast enough to put all previous attempts at integrated graphics to shame and compete with entry level discrete GPUs. The fact that you can get Radeon HD 5450 performance for free with a Core i5 2400 is just awesome. As I mentioned before, you won't want to throw away your GTX 460, but if you were planning on spending $50 on a GPU - you may not need to with Sandy Bridge."
http://www.anandtech.com/show/3871/t...ns-in-a-row/13
The main problem with Sandy Bridge has been that there was no word of OpenCL support in the GPU. Now it seems that the graphics will support OpenCL:
"Intel is also working on OpenCL for the graphics part of Sandy Bridge, according to sources.
Intel declined to comment directly on Apple's plans, but regarding OpenCL it would only tell CNET: 'In terms of full product support, we continue to evaluate when and where OpenCL will intercept our various products.' "
http://news.cnet.com/8301-13924_3-20023505-64.html
Looks very, very good. Good enough that I'll probably finally upgrade from my trusty 2006 Macbook 2,1
Rumors of CpenCL support so far are just that - rumors. Intel has stated that it doesn't support OpenCL. Whether it will in the future is an open question. So, unless Apple is working to do that, perhaps with their help, it won't happen. And as we all know, having Apple say something before the new models come out is just wishful thinking.
This is where I can see Apple throwing its weight around. I can see Apple strong arming Intel into supporting OpenCL, even if it?s special category just for Mac notebook that aren?t found on Intel?s price list. It?s not like there isn?t precedence to support this possibility.
We have to be careful with this, as it's more than software or firmware that makes OpenCL possible on any given gpu. The gpu has to be programmable, and it must support features that will allow OpenCL firmware and software to do what is needed. If that's not possible, then nothing can be done to make it so.
That would mean that Intel would have to make special designs just for Apple, and I don't think Apple wants that. Remember the problems Apple had with latter G4 and G5 chips? That comes from having limited edition chip lines. What if Intel agrees, but can't get them ready in time? Apple is screwed. And as we all know, that happens all the time in chip making. Look at all the problems AMD has had over the years, which is why I don't trust them.
By using what everyone else has, Apple is at least assured that if there are delays, it's for the entire industry, not just them. Otherwise, we could have the problem of waiting months for a machine to come out that should have been out way before. Not good!
I already tested the current MBP 13" with the apps I'm using, and it's working well. The 9400M that I have in my late 2009 mini is barely working well for my needs (I'm experiencing some textures weirdness with it).
Even worst is that OpenGL demands often go up as programs are updated. So even if the current machine passes with flying colors you may have issues in the future if you don't have enough margin with respect to performance.
Beyound that I do hope you are running the latest drivers in Snow Leopard.
I'm using Celestia (full 3D OpenGL astronomy app) with large hi-res textures and some heavy 3D models, to teach astronomy in a classroom with a wall projector. I'm also using several full HD videos in QT.
cool!.
I've always found this science fascinating especially the vastness of space and our current in ability to navigate it.
The 15" MBP is too large, for a portable. The 13" is just perfect, sizewise and is not too heavy to bring under the arm, during a full day.
I understand what you are saying but it doesn't look real gold right now if the rumor is true. The lack of decent GPU performance in the 13" is why I wish they would drop the CD drive and put a real GPU in the machine with a larger battery. The MBP could really use such a configuration. On the otherhand simply waiting for Llano from AMD would likely give you the results you are looking for.
In the end you won't know what is up until the hardware is released and you can test your software on it. I wouldn't be to trustful of bench marks either, this is a case where you need to test with your software to see what is happening visually. Drivers play a key role so what doesn't work at release could six months later.
Rumors of CpenCL support so far are just that - rumors. Intel has stated that it doesn't support OpenCL. Whether it will in the future is an open question. So, unless Apple is working to do that, perhaps with their help, it won't happen. And as we all know, having Apple say something before the new models come out is just wishful thinking.
So are you saying Intel's Director of Graphics Architecture is lying?
Thomas Piazza, an Intel fellow and director of graphics architecture for the Intel Architecture Group, said that Sandy Bridge-based chips in their current implementation will not support DirectX 11, a Microsoft technology for accelerating multimedia and games. Currently, Sandy Bridge supports DirectX 10.1 and OpenCL 1.1--the latter used on Apple's Mac operating systems, according to Piazza.
Read more: http://news.cnet.com/8301-13924_3-20...#ixzz17dOiAnNQ
Intel's Director of Graphics Architecture has confirmed that Sandy Bridge supports OpenCL 1.1. Apple currently supports OpenCL 1.0 so Apple can at least tout that as an upgrade, although current AMD and nVidia solutions support OpenCL 1.1 as well, if Apple were to release drivers for it. Sandy Bridge's IGP won't likely beat the 320M in raw performance unless Apple really invests in the drivers, but it's certainly 320M class so it'll be a sidegrade. Architecturally, Sandy Bridge's IGP should be superior for OpenCL to both nVidia's off-chip IGP and AMD's Fusion whose IGP connects to the CPU via a crossbar bus, because Sandy Bridge's IGP is on die and shares an L3 cache with the CPU allow efficient low-latency, high-bandwidth data sharing between the IGP and CPU for OpenCL. No longer will the CPU have to set up the OpenCL data and commands, transfer it to the GPU, wait for the GPU to process it, and then wait for the results to be copied back. Everything will be in a common data location that both the IGP and CPU can read and write to. This efficiency should help make up some of the raw performance deficit of Sandy Bridge's IGP for OpenCL.
http://lists.cs.uiuc.edu/pipermail/l...er/036284.html
Intel is actively hiring OpenCL driver developers specifically for their IGP, so OpenCL support is definitely coming, it's just a matter of when and how good it'll be.
As well, Sandy Bridge's CPU already includes enhancements that target many of the same benefits of OpenCL. For example, Sandy Bridge includes a dedicated hardware video encode accelerator which should be faster and more power efficient than running an implementation on OpenCL on a GPU. AVX also greatly improves the CPU's vector and multimedia performance and should be easier for software developers to modify their existing SSE code to adopt rather than writing a new OpenCL path. AMD's Fusion processors support neither in the 2010-2011 timeframe.
So are you saying Intel's Director of Graphics Architecture is lying?
Read more: http://news.cnet.com/8301-13924_3-20...#ixzz17dOiAnNQ
Intel's Director of Graphics Architecture has confirmed that Sandy Bridge supports OpenCL 1.1.
As you probably know, the Intel has only officially announced *CPU*-based implementations of OpenCL. If Intel is starting to implement OpenCL support on the GPU (as the recent hiring ad you posted indicates) this is excellent news! But don't claim that they've been supporting it all along.
Sandy Bridge's IGP won't likely beat the 320M in raw performance unless Apple really invests in the drivers, but it's certainly 320M class so it'll be a sidegrade. Architecturally, Sandy Bridge's IGP should be superior for OpenCL to both nVidia's off-chip IGP and AMD's Fusion whose IGP connects to the CPU via a crossbar bus, because Sandy Bridge's IGP is on die and shares an L3 cache.
Sounds nice, we'll know for sure when the new Intel and AMD hardware is out.
Sandy Bridge includes a dedicated hardware video encode accelerator which should be faster and more power efficient than running an implementation on OpenCL on a GPU.
Oh yeah, I just hope Handbrake developers have got their hands on this!
The main problem with Sandy Bridge has been that there was no word of OpenCL support in the GPU. Now it seems that the graphics will support OpenCL:
"Intel is also working on OpenCL for the graphics part of Sandy Bridge, according to sources.
Intel declined to comment directly on Apple's plans, but regarding OpenCL it would only tell CNET: 'In terms of full product support, we continue to evaluate when and where OpenCL will intercept our various products.' "
Are we talking hardware support for a GPU-based Open GL? Or are we talking about functional drivers for a GPU-based Open GL?
So are you saying Intel's Director of Graphics Architecture is lying?
I'm saying I would be embarassed to be out in public with the title "Director of Graphics Architecture for Intel". Everything they've done for years has been sh!t. Why should we believe him now? I say the guy should put up or shut up. Show us real world benchmarks of a shipping product, only then would I be willing to re-evaluate my assumptions of how bad any future Intel IGP will be vs. the competition.
From the cnet article, this is what you call being damned with faint praise: "It will integrate Intel's best graphics chip technology to date directly onto the central processing unit."
As you probably know, the Intel has only officially announced *CPU*-based implementations of OpenCL. If Intel is starting to implement OpenCL support on the GPU (as the recent hiring ad you posted indicates) this is excellent news! But don't claim that they've been supporting it all along.
It is kind of open to interpretation whether Piazza's OpenCL statement was referring to the IGP or just the CPU implementation, although seeing the article's focus is on the IGP, Piazza is the director of the graphics group, and OpenCL was mentioned in context with DirectX, I'd hope he was talking about OpenCL for IGP.
Sounds nice, we'll know for sure when the new Intel and AMD hardware is out.
Just to avoid getting people's hopes up and getting it quoted back to me later, when I say Intel's IGP/CPU implementation is superior for OpenCL I really do mean architecturally only. I have no idea what the final raw performance numbers are going to be like, but it's a good design idea.
I'm saying I would be embarassed to be out in public with the title "Director of Graphics Architecture for Intel". Everything they've done for years has been sh!t. Why should we believe him now? I say the guy should put up or shut up. Show us real world benchmarks of a shipping product, only then would I be willing to re-evaluate my assumptions of how bad any future Intel IGP will be vs. the competition.
In fairness, I think the problem with Intel's IGP group is more what they want to do rather than can they do it. From a technical perspective, the GMA X4500MHD had 10 execution units while Arrandale's IGP has 12 execution units with performance doubling. Sandy Bridge looks to double Arrandale's performance with the same number of execution units. I guess that can be interpreted as very inefficient designing on the older IGPs, but it can also point to Intel's IGP really knowing their stuff in order to extract significantly more performance without drastically increasing the resources used to do it. Certainly, AMD and nVidia's method of increasing performance is generally to increase execution units rather than make the existing ones more efficient. In fact, performance analysis of AMD's HD5000 series seems to show that the HD5000's shader units are actually less efficient than the HD4000 units, just that there are more of them.
It's just that up to now, Intel IGP's have been too focused on the corporate market. It's only recently in Windows that everyday computer tasks require decent IGP acceleration, like the Aero GUI, video decode, Flash acceleration, etc. For the rest of the decade, Intel IGPs were sufficient for Windows XP and certainly it didn't seem like corporations were complaining. In the consumer spectrum, computer average selling prices have been falling with consumers expecting to pay less and less for computers. That means that OEMs are generally not able to put in discrete GPUs which increases the presence of Intel IGPs in consumer computers and consumers want to play games on them which Intel IGPs clearly weren't able to do. It's only now that Intel is belatedly addressing this, that computer average selling prices will remain low, most consumer computers will use Intel IGPs, and most consumers will expect decent gaming performance from them, so Intel has to develop a decent IGP capable of gaming. Hopefully we'll know next month whether Intel will be successful with Sandy Bridge.
In fairness, I think the problem with Intel's IGP group is more what they want to do rather than can they do it. From a technical perspective, the GMA X4500MHD had 10 execution units while Arrandale's IGP has 12 execution units with performance doubling. Sandy Bridge looks to double Arrandale's performance with the same number of execution units. I guess that can be interpreted as very inefficient designing on the older IGPs, but it can also point to Intel's IGP really knowing their stuff in order to extract significantly more performance without drastically increasing the resources used to do it. Certainly, AMD and nVidia's method of increasing performance is generally to increase execution units rather than make the existing ones more efficient. In fact, performance analysis of AMD's HD5000 series seems to show that the HD5000's shader units are actually less efficient than the HD4000 units, just that there are more of them.
It's just that up to now, Intel IGP's have been too focused on the corporate market. It's only recently in Windows that everyday computer tasks require decent IGP acceleration, like the Aero GUI, video decode, Flash acceleration, etc. For the rest of the decade, Intel IGPs were sufficient for Windows XP and certainly it didn't seem like corporations were complaining. In the consumer spectrum, computer average selling prices have been falling with consumers expecting to pay less and less for computers. That means that OEMs are generally not able to put in discrete GPUs which increases the presence of Intel IGPs in consumer computers and consumers want to play games on them which Intel IGPs clearly weren't able to do. It's only now that Intel is belatedly addressing this, that computer average selling prices will remain low, most consumer computers will use Intel IGPs, and most consumers will expect decent gaming performance from them, so Intel has to develop a decent IGP capable of gaming. Hopefully we'll know next month whether Intel will be successful with Sandy Bridge.
Great post! Thank you.
Since Apple seems to be headed toward Sandy Bridge, I'm really hoping you're right.
It is kind of open to interpretation whether Piazza's OpenCL statement was referring to the IGP or just the CPU implementation, although seeing the article's focus is on the IGP, Piazza is the director of the graphics group, and OpenCL was mentioned in context with DirectX, I'd hope he was talking about OpenCL for IGP.
Just to avoid getting people's hopes up and getting it quoted back to me later, when I say Intel's IGP/CPU implementation is superior for OpenCL I really do mean architecturally only. I have no idea what the final raw performance numbers are going to be like, but it's a good design idea.
In fairness, I think the problem with Intel's IGP group is more what they want to do rather than can they do it. From a technical perspective, the GMA X4500MHD had 10 execution units while Arrandale's IGP has 12 execution units with performance doubling. Sandy Bridge looks to double Arrandale's performance with the same number of execution units. I guess that can be interpreted as very inefficient designing on the older IGPs, but it can also point to Intel's IGP really knowing their stuff in order to extract significantly more performance without drastically increasing the resources used to do it. Certainly, AMD and nVidia's method of increasing performance is generally to increase execution units rather than make the existing ones more efficient. In fact, performance analysis of AMD's HD5000 series seems to show that the HD5000's shader units are actually less efficient than the HD4000 units, just that there are more of them.
It's just that up to now, Intel IGP's have been too focused on the corporate market. It's only recently in Windows that everyday computer tasks require decent IGP acceleration, like the Aero GUI, video decode, Flash acceleration, etc. For the rest of the decade, Intel IGPs were sufficient for Windows XP and certainly it didn't seem like corporations were complaining. In the consumer spectrum, computer average selling prices have been falling with consumers expecting to pay less and less for computers. That means that OEMs are generally not able to put in discrete GPUs which increases the presence of Intel IGPs in consumer computers and consumers want to play games on them which Intel IGPs clearly weren't able to do. It's only now that Intel is belatedly addressing this, that computer average selling prices will remain low, most consumer computers will use Intel IGPs, and most consumers will expect decent gaming performance from them, so Intel has to develop a decent IGP capable of gaming. Hopefully we'll know next month whether Intel will be successful with Sandy Bridge.
i have a core i5 laptop at work and the intel HD is more than enough for normal web browsing including youtube. with flash 10.2 beta it's insanely fast at video playback. and the laptop stays cool even when using flash.
not like the old HP POS i had that you could use as a food warmer
It's just that up to now, Intel IGP's have been too focused on the corporate market. It's only recently in Windows that everyday computer tasks require decent IGP acceleration, like the Aero GUI, video decode, Flash acceleration, etc. For the rest of the decade, Intel IGPs were sufficient for Windows XP and certainly it didn't seem like corporations were complaining. In the consumer spectrum, computer average selling prices have been falling with consumers expecting to pay less and less for computers. That means that OEMs are generally not able to put in discrete GPUs which increases the presence of Intel IGPs in consumer computers and consumers want to play games on them which Intel IGPs clearly weren't able to do. It's only now that Intel is belatedly addressing this, that computer average selling prices will remain low, most consumer computers will use Intel IGPs, and most consumers will expect decent gaming performance from them, so Intel has to develop a decent IGP capable of gaming. Hopefully we'll know next month whether Intel will be successful with Sandy Bridge.
Geez - you guys. I'm still gaming on my IBM XT and Apple II, and the text game "Softporn" is just fine with me - who needs fancy graphics? Sheesh.
i have a core i5 laptop at work and the intel HD is more than enough for normal web browsing including youtube. with flash 10.2 beta it's insanely fast at video playback. and the laptop stays cool even when using flash.
not like the old HP POS i had that you could use as a food warmer
What is the battery life difference between using HTML5 video vs. Flash video onYouTube for the same video across different resolutions?
Anandtech's preview of Sandy Bridge makes it sound very promising. Even the integrated graphics sounds pretty good.
Looks very, very good. Good enough that I'll probably finally upgrade from my trusty 2006 Macbook 2,1
The problem there is you are looking at the desktop version of Sandy Bridge so you won't see it in the Macbook. It compares favourably to the desktop Radeon 5450, which is impressive but still only matches the lowest end GPU in the current iMac.
Apple can use the saving from the dedicated GPU to put an SSD in place and the overall iMac power consumption would lower but it will be a side-step in terms of graphics performance at best.
The media conversion hardware looks good but not as impressive as the 400FPS claim:
http://www.youtube.com/watch?v=DopdfPnWlFA
The demo there was only twice as fast as the software encoding. It does look like it's programmable though so it might be possible to extend beyond just AVC encoding.
The mobile hardware is still uncertain. The good thing is they have stated the mobile GPUs will all have 12EUs and 2 graphics cores where the desktop versions might only have 6 in some models. Anand thought the above desktop version was a 6EU model but it still has no bearing on the mobile part.
1 EU I'd assume is equivalent in description to 1 NVidia core of which the 320M has 48. The number isn't comparable across manufacturers though.
The tested desktop version of Sandy Bridge is faster than the NVidia 320M by about 15-20%. If it was a 6EU model, then it suggests the 12-core mobile GPU might be worthwhile.
Notebookcheck have estimated the position lower than the 320M based on the 5450 mobile/desktop comparisons:
http://www.notebookcheck.net/Intel-H...0.37948.0.html
My main concern is when it comes to feature support. Sometimes the raw performance of the GPU is fast enough but when you use hardware-accelerated apps, none of the options are supported whereas they are always supported with NVidia and sometimes supported with AMD.
OpenCL doesn't look likely but we can't tell this until Intel say one way or the other. Given their stance on GPGPU up until now, I'd say it's not going to happen. They have pages about Ct detailing why it's better than GPUs - task parallels, standard code etc.
Apple cannot use AMD right now. LLano isn't coming out until Summer so Fusion is out for now.
I hope Apple will go Sandy Bridge + dedicated but for cost, it's not likely. I guess they will make a choice based on the following:
- is the power consumption better => yes
- is the performance faster than C2D => yes
- will it be fast enough for casual games => probably
- it likely has no GPU OpenCL support but it supports hardware media encode/decode and OpenCL code can be tested on the CPU
There's nothing in there that immediately puts them off using it alone. It's good enough but not so good you will want it over a more expensive model.
What is the battery life difference between using HTML5 video vs. Flash video onYouTube for the same video across different resolutions?
haven't measured it, but on my old POS HP centrino laptop it would get hot just by seeing non-video flash content and the battery would die in 90 minutes instead of the 3 hour rated time. and that was with an ATI HD2600 discrete graphics chip in there. the CPU would spike up to close to 100% and videos would skip frames. full screen was unwatchable
on my Core i5 the battery seems to last pretty close to the rated time when watching flash videos, but i usually have it charging when i'm using it. with flash 10.1 and 10.2 beta the CPU goes to high teens when watching the video in 720p and full screen
haven't measured it, but on my old POS HP centrino laptop it would get hot just by seeing non-video flash content and the battery would die in 90 minutes instead of the 3 hour rated time. and that was with an ATI HD2600 discrete graphics chip in there. the CPU would spike up to close to 100% and videos would skip frames. full screen was unwatchable
on my Core i5 the battery seems to last pretty close to the rated time when watching flash videos, but i usually have it charging when i'm using it. with flash 10.1 and 10.2 beta the CPU goes to high teens when watching the video in 720p and full screen
on my 2010 13” MBP I get about 30% longer battery life with HTML5 video from the same YouTube file on 720p vids with Flash 10.2, which seems 5 to 10% better than Flash 10.1, which was a huge improvement over Flash 10.0. But that’s a very weak way of measuring the difference.
Due to Flash crashes I uninstalled it and went back to Flash 10.1. Not that it matters much since I rarely need to use the Flash plug-in with HTML5 video on pretty much every site I frequent. It is mice that that Adobe adds the uninstaller to the page for the Flash beta installer, instead of making for you hunt it down.
So are you saying Intel's Director of Graphics Architecture is lying?
Read more: http://news.cnet.com/8301-13924_3-20...#ixzz17dOiAnNQ
Intel's Director of Graphics Architecture has confirmed that Sandy Bridge supports OpenCL 1.1. Apple currently supports OpenCL 1.0 so Apple can at least tout that as an upgrade, although current AMD and nVidia solutions support OpenCL 1.1 as well, if Apple were to release drivers for it. Sandy Bridge's IGP won't likely beat the 320M in raw performance unless Apple really invests in the drivers, but it's certainly 320M class so it'll be a sidegrade. Architecturally, Sandy Bridge's IGP should be superior for OpenCL to both nVidia's off-chip IGP and AMD's Fusion whose IGP connects to the CPU via a crossbar bus, because Sandy Bridge's IGP is on die and shares an L3 cache with the CPU allow efficient low-latency, high-bandwidth data sharing between the IGP and CPU for OpenCL. No longer will the CPU have to set up the OpenCL data and commands, transfer it to the GPU, wait for the GPU to process it, and then wait for the results to be copied back. Everything will be in a common data location that both the IGP and CPU can read and write to. This efficiency should help make up some of the raw performance deficit of Sandy Bridge's IGP for OpenCL.
http://lists.cs.uiuc.edu/pipermail/l...er/036284.html
Intel is actively hiring OpenCL driver developers specifically for their IGP, so OpenCL support is definitely coming, it's just a matter of when and how good it'll be.
As well, Sandy Bridge's CPU already includes enhancements that target many of the same benefits of OpenCL. For example, Sandy Bridge includes a dedicated hardware video encode accelerator which should be faster and more power efficient than running an implementation on OpenCL on a GPU. AVX also greatly improves the CPU's vector and multimedia performance and should be easier for software developers to modify their existing SSE code to adopt rather than writing a new OpenCL path. AMD's Fusion processors support neither in the 2010-2011 timeframe.
This is from an article very recently in Ars Technica, a site that keeps up with this, and has some very qualified people in the field writing for them. It also follows what Intel has been sayig publically about this issue. so while I'm certainly not saying that this guy is lying, it may not be what you think, or it may be a new development.
The IGP in Intel's upcoming Sandy Bridge mobile processors is expected to have vastly improved performance over its Arrandale predecessor, but is still based on relatively ancient architecture and, from what we have been able to gather, is not OpenCL compatible.
Intel's next architecture update, Ivy Bridge, is expected to have a design that is OpenCL compatible, but Intel won't commit to providing OpenCL drivers. "Intel is evaluating when and where OpenCL support will intercept our products, however no announcement has been made," spokesperson Nick Knupffer told Ars. Apple isn't in a position to wait any longer for Intel to step up its IGP game, and NVIDIA likely has designs waiting in the wing for Apple to use once the legal issues are cleared. A settlement now could ultimately be a win-win-win for everyone involved.
And as OpenCl doesn't only work on gpu's, that may also what we're seeing from Intel right now.
on my 2010 13? MBP I get about 30% longer battery life with HTML5 video from the same YouTube file on 720p vids with Flash 10.2, which seems 5 to 10% better than Flash 10.1, which was a huge improvement over Flash 10.0. But that?s a very weak way of measuring the difference.
Due to Flash crashes I uninstalled it and went back to Flash 10.1. Not that it matters much since I rarely need to use the Flash plug-in with HTML5 video on pretty much every site I frequent. It is mice that that Adobe adds the uninstaller to the page for the Flash beta installer, instead of making for you hunt it down.
depends on the browser as well
in the last year or two someone did a test and believe it or not IE came out as the most energy efficient browser. firefox was the worst. forgot where safari, opera and chrome ended up
Not reliable at all. Apple is depending on OpenCL for performance. Otherwise they wouldn't be willing to take the flack for still using Core 2 chips in its lower machines so they can use OpenCL GPU's. I don't see them backing away from that, unless, somehow, they've figured out a way around it. Being that's it's Apple, that's not impossible, but not likely.
Not entirely true. There's little advantage in the i3 over the C2D and the IGP isn't as good as in sandy bridge. In the case of sandy bridge I believe the mobile parts are all i5 for now. The difference in CPU performance with be huge.
As argued back and forth in the other thread OpenCL support IMHO isn't a necessarily a hard requirement for the GPU.
The quick recap of my position is:
1) Intel is supporting OpenCL on the CPU. Therefore OpenCL calls could be implemented by on die hardware in Sandy Bridge (i.e. via 256 bit wide SIMD or the hardware encoder/decoders).
2) Intel is providing on die hardware support/enhancement for the 2 most common GP GPU uses: cryptography and transcoding. How can we guess these are the two most common use cases? Because these are the two GPGPU benchmarks added the SiSoft Sandra benchmark suite. [1][2]
3) The GPU performance as a GPU is slightly better than the 320M as the 5450 is slightly better than the 320M.
Given these three, going with just a Sandy Bridge i5 in the mini, MB and MBP 13" is plausible. The MBA I believe uses lower TDP parts and will simply have to wait.
---
[1] "The demo used Cyberlink?s Media Espresso to convert a ~1 minute long 30Mbps 1080p HD video clip to an iPhone compatible format. On Sandy Bridge the conversion finished in a matter of a few seconds (< 10 seconds by my watch)."
http://www.anandtech.com/show/3916/i...anscode-engine
[2] "The interesting bit is that SiSoft expects how CPUs should be able to challenge GPGPUs in the near future, given that "Using 256-bit register width (instead of 128-bit of SSE/2/3/4) yields further performance gains through greater parallelism in most algorithms. Combined with the increase in processor cores and threads we will soon have CPUs rivaling GPGPUs in performance."
http://www.brightsideofnews.com/news...gpu-tests.aspx