OpenCL is a hardware abstraction layer for processors. It's raison d'être is to shift some types of processing tasks away from the CPU and onto the GPU.
But it doesn't have to ve used that way.
Quote:
Intel has the most powerful CPUs on the market, but lags in the GPU space. Logically, they would be against any technology that allows the GPU take over the tasks of the CPU.
To the point of loosing customers? If Intel doesn't have a fast OpenCL implementation that is what will happen.
Quote:
In fact, there is no indication that Intel will ever support OpenCL properly. They are listed as a member of the Chronos Consortium, which administers OpenCL, but have only produced an initial implementation of OpenCL which runs on the *CPU*.
Which just recently went Alpha. As you note this is a CPU based OpenCL implementation and does not indicate future GPU capability. So we still don't know if Intels Sandy Brige has a gpu that actually helps people
Quote:
AMD is much more enthusiastic about OpenCL. Too bad Bobcat is a thoroughly low end product.
From what I'm hearing Bobcat would have worked just fine in the AIRs, at least from the GPU standpoint. The problem is information is very tight with respect to actual CPU performance. Supposedly the embargo on benchmarks ends sometime this week, so hopefully we will have a more complete picture soon.
Llano (sp) is still a ways off so who knows. You are right about AMD being more serious, after a slow start they seem to be after OpenCL with gusto. Very aggressive is the word. However the simple fact that they can in fact run OpenCL code well gives them an advantage over Intel.
Quote:
Perhaps their future chips (Lanos?) will get serious consideration from Apple.
To the point of loosing customers? If Intel doesn't have a fast OpenCL implementation that is what will happen.
I suspect that this is a bit of a showdown between Apple and Intel. Apple must be hopping mad if Intel is really refusing to support OpenCL. On the other hand, Intel's CPUs are so far ahead of the competition (right now, and for the foreseeable future) that Apple can't jump ship.
Quote:
Originally Posted by wizard69
From what I'm hearing Bobcat would have worked just fine in the AIRs, at least from the GPU standpoint. The problem is information is very tight with respect to actual CPU performance. Supposedly the embargo on benchmarks ends sometime this week, so hopefully we will have a more complete picture soon.
The embargo is over. Here's a review that looks at the top of the line "Zacate" version of Brazos APU (Bobcat + GPU) - whew, so many codenames! It compares Zacate to an Intel Celeron SU2300 CULV processor in combination with an NVIDIA ION graphics / chipset. The Celeron/ION combo is lower end than the Core2Duo/320M in the MacBook Air. (Celeron is similar to the Core2Duo but slower clockspeed and smaller L2 cache, and ION includes essentially the same graphics as the 9400M in the previous generation Air).
Zacate is competitive with Celeron/ION (a bit slower on CPU, slightly better graphics), though it does seem to be a good performer considering its low price and power consumption. It would definitely be a step down from the current MacBook Air internals.
Of course this depends upon what you are expecting from the processor but considering is power profile it is pretty impressive. It basically destroys ATOM.
Quote:
Originally Posted by hypercommunist
I suspect that this is a bit of a showdown between Apple and Intel. Apple must be hopping mad if Intel is really refusing to support OpenCL.
Livid! Lets face it, it will be a very very long time before an Intel CPU can do what a GPU can do well. More so Apple must be looking at AMDs and Intels roadmaps and be wondering why they are doing business with Intel. It really looks like AMD has a vision of the future more in line with Apples.
Quote:
On the other hand, Intel's CPUs are so far ahead of the competition (right now, and for the foreseeable future) that Apple can't jump ship.
Well this i don't buy. For one thing they don't need to jump ship, rather they just need to pick an choose the best implementations for each product.
Beyound that many of Apples products don't stress ultimate performance but rather other features. For example the Mini is a low power compact device. Similarly the Mac Book and even the AIRs are not CPU power houses, AMD could slot in processors today with similar performance profiles.
As to AMD the little bit of reviewing I've seen so far is very impressive. Remember this is an extremely low power solution with a die not much different in size than ATOM. I do not see Intel offering anything similar.
[quote]
The embargo is over. Here's a review that looks at the top of the line "Zacate" version of Brazos APU (Bobcat + GPU) - whew, so many codenames! It compares Zacate to an Intel Celeron SU2300 CULV processor in combination with an NVIDIA ION graphics / chipset. The Celeron/ION combo is lower end than the Core2Duo/320M in the MacBook Air. (Celeron is similar to the Core2Duo but slower clockspeed and smaller L2 cache, and ION includes essentially the same graphics as the 9400M in the previous generation Air).
[quote]
If I find the time this weekend I will review some of these benchmarks in greater depth. I must admit though these numbers look really good for a processor that sits on a die approximately the size of an ATOM.
Zacate is competitive with Celeron/ION (a bit slower on CPU, slightly better graphics), though it does seem to be a good performer considering its low price and power consumption.
Exactly; excellent performance given the power consumption.
Quote:
It would definitely be a step down from the current MacBook Air internals.
Well at 1.6 GHz maybe but what will happen at say 2.4 GHz assuming AMD can produce Zacates at that clock rate and maintain thermals at tolerable levels. If the whole chip can handle thermals of around 25 watts would Intel be able to compete?
Even more interesting is that AMD has said that the GPUs of the Fusion products may rev quicker than the rest of the chip. So maybe by the time AIRs are ready for their next rev we will see a suitable Bobcat based fusion product. Or maybe Apple will do something for the low end. Right now Mac Book plastic is poorly placed with the AIRs and 13" MBP all clumped together. Maybe Apple will turn Mac Book into a very low cost laptop. In a nut shell AMD is giving Apple options it has never had before.
Maybe Zacate is less than ideal for AIR today, but it is a damn good start. More so it should allow for surprisingly low cost, low power but well performing devices.
So, the big unanswered questions about Sandy Bridge processors and Apple's next MacBook Pro refresh are:
- Will the 13" MBP get a discrete graphics processor or will it rely on Intel's inbuilt one?
This is really tough to respond to as we simply don't know the full capabilities of Intels Sandy Bridge GPU. My perspective is that it will be to close in oerformance to the AIRs and Mac Books thus Apple will go with a suplemental GPU. The wild card is that they wait for AMDs high performamce Fusion products.
Quote:
- What discrete GPUs will the 15" and 17" MBPs get?
That is a good question. I would love to see an upgrade of course.
Quote:
- Is it totally beyond the bounds of possibility that Intel produces an integral GPU that's good enough to replace the discrete GPU on all MBPs?
For pro level use I think it will be a couple of years before that will happen. Apple could try it with Sandy Bridge but I think they would have a user revolt. Beyound that AMD is likely to have better GPUs in their Fusion products for the foreseeable future.
Quote:
- Will all MBPs get USB 3.0?
Possibly. The rumor is AMD will be supporting USB 3 before Intel.
Quote:
- Will all MBPs get standard 256 Gb SSDs?
That should be the minimal. What ever they get it will be a SSD.
Quote:
- Will the DVD disappear?
I really hope so in the 13". But not in the larger machines as again the pros would revolt.
Quote:
- Will we get a hardware redesign too, i.e. a thinner form factor and lighter weight?
I looked that over and it only supports OCL on the cpu. Perhaps I missed something but I don't see intel IGP being able to pitch in and help on OCL applications. And it only works on Windows.
Of course this depends upon what you are expecting from the processor but considering is power profile it is pretty impressive. It basically destroys ATOM. Lets face it, it will be a very very long time before an Intel CPU can do what a GPU can do well. More so Apple must be looking at AMDs and Intels roadmaps and be wondering why they are doing business with Intel. It really looks like AMD has a vision of the future more in line with Apples.
Next year, NVIDIA will air its Tegra 2 system-on-a-chip which comes complete with a dual-core ARM9 and super-powered mobile GPU. *Not to be outdone, Qualcomm just announced its own plans for mobile hardware domination and they're shaping up to be equally impressive.
Qualcomm's Chip Plans for 2011
First up will be the pair of previously announced dual CPU core 45 nm system-on-a-chips. *These chips will launch in phones early next year. *Dubbed the MSM8260 and 8660, these system-on-a-chips are powered by the Adreno 205 GPU and two Scorpion cores clocked at 1.2 GHz. *The key difference between the two chip models is in mobile broadcast standards support. *The 8260 only supports HSPA+, while the 8660 supports HSPA+, CDMA2000 and 1xEV-DO Rev. B. *These chips have already been completed and sampled to hardware partners, so phones should be soon coming to market.
This week Qualcomm also disclosed that it would be moving to the 28 nm process node. The first SoC to be produced at the new node will be the MSM8960. *The new SoC won't just be a die shrink; it will also feature a new core design. *While the company refuses to hint at clock speeds, it will say the new chip will be 5 times as powerful as the original (single Scorpion core) SnapDragon, meaning that each core will be roughly 2.5 times as powerful as its predecessor.
The company also claims it can achieve all of that while operating at 75% of the current generation's power (though it was less specific about what kind of power measure and which core -- 45 nm or 65 nm -- it was comparing to). *This claim has been met with a bit of confusion and skepticism, but if it's as good as it sounds, Qualcomm should be good shape.
In the vague boasting department, Qualcomm also bragged that the 8960 would 4x as powerful graphically (as some nonspecific design). *The Adreno 205 is roughly twice as powerful as the original Adreno 200, so 4x the 200 would be twice the current 205's power.
Graphics as Powerful as a PS3 -- in the Palm of Your Hand
Some time in the 2011-2013 window, Qualcomm plans to air the Adreno 3xx which could be its crowning achievement -- if it pulls it off. *The GPU will be made for use with SoCs on the 28 nm node. *Qualcomm claims it will be graphically as powerful as an Xbox 360 or PlayStation 3. *Of course, those console GPUs are paired with lots of graphics RAM so it seems unlikely that the true performance would match these next-gen consoles.
(Side note: an Adreno GPU will power the upcoming PlayStation Phone.)
Nonetheless, if it can even match the processing power of these consoles, that could make for some impressive smartphone graphics. *The new GPU will support the upcoming OpenGL ES "Haiti", the successor to OpenGL ES 2.0. *It will also jump on the GPU computing bandwagon, adding support for OpenCL 1.1. *What good GPU computing on a smartphone would be seems questionable, but then again, several years back few could predict the growing uses of GPU computing in the PC/server sector today.
Qualcomm is also promising that it will outcompete competitors like NVIDIA in chip cost and size. *
The company is also developing dual-mode chips, which will support both 3G technologies and 4G (LTE).
Conclusions
To put this week's presentation by Qualcomm in perspective, it appears that some very powerful hardware is coming down the pipe. *One thing we see as a clear problem is that the amounts of memory found in current generation smartphones won't be capable of supporting these kinds of phones. *
If hardware partners can work together, to say, triple the memory of current Android models (1 GB for CPU, 512 MB for GPU), then we could have some mighty impressive products on our hands. *Otherwise Qualcomm, NVIDIA, and others miss overshooting on the processing power mark.
So Intel's bundlegate continues unabated. Shovelling you useless GPUs...
Given Anand's benchmarks shows it to be about the same performance as the Radeon HD 5450 it's not that bad.
WRT To OpenCL...if OpenCL on the CPU works well enough it might not matter enough Apple if the IGP isn't supporting it if the total package is better then alternatives.
WRT To OpenCL...if OpenCL on the CPU works well enough it might not matter enough Apple if the IGP isn't supporting it if the total package is better then alternatives.
OCL on the CPU is not very important on the Mac platform. GCD is available to see that all cpu cores are utilized when its useful to do so.
OCL is meaningful to bring the gpu into action when its feasible to do so.
While I don't agree with every bit of your spin below the info correctly highlights where the industry is going.
Quote:
Originally Posted by hypercommunist
Next year, NVIDIA will air its Tegra 2 system-on-a-chip which comes complete with a dual-core ARM9 and super-powered mobile GPU. *Not to be outdone, Qualcomm just announced its own plans for mobile hardware domination and they're shaping up to be equally impressive.
This is one of the reasons I'm expecting vastly improved hardware on the iPad 2. These new generations of SoC ARM based hardware sould be very impressive.
Quote:
Qualcomm's Chip Plans for 2011
First up will be the pair of previously announced dual CPU core 45 nm system-on-a-chips. *These chips will launch in phones early next year. *Dubbed the MSM8260 and 8660, these system-on-a-chips are powered by the Adreno 205 GPU and two Scorpion cores clocked at 1.2 GHz. *The key difference between the two chip models is in mobile broadcast standards support. *The 8260 only supports HSPA+, while the 8660 supports HSPA+, CDMA2000 and 1xEV-DO Rev. B. *These chips have already been completed and sampled to hardware partners, so phones should be soon coming to market.
I doubt very much that Apple will look outside of Apple for the next SoC. However these releases highlight where the industry will be very early next year. They are packing a lot of stuff on to these chips and vastly increasing performance. In many cases they are doing so at a lower power profile.
I see these as hints as to what Apples A4 replacement will look like.
Quote:
This week Qualcomm also disclosed that it would be moving to the 28 nm process node. The first SoC to be produced at the new node will be the MSM8960. *The new SoC won't just be a die shrink; it will also feature a new core design. *While the company refuses to hint at clock speeds, it will say the new chip will be 5 times as powerful as the original (single Scorpion core) SnapDragon, meaning that each core will be roughly 2.5 times as powerful as its predecessor.
When they hit 28 nm they will actually be ahead of Intel.
With respect to Apple though they have many patents related to CPU optimization. I could see Apple deliver a hybrid ARM design with CPU optimizations for Objective C code. In any event it should be remembered that anything the other players can do on the SoC apple can do too. Plus they can do it without the kitchen sink.
Quote:
The company also claims it can achieve all of that while operating at 75% of the current generation's power (though it was less specific about what kind of power measure and which core -- 45 nm or 65 nm -- it was comparing to). *This claim has been met with a bit of confusion and skepticism, but if it's as good as it sounds, Qualcomm should be good shape.
I've seen some very impressive power figures for processes in the 22-28nm range so the possibility is there. What is notable here is the additional core.
Quote:
In the vague boasting department, Qualcomm also bragged that the 8960 would 4x as powerful graphically (as some nonspecific design). *The Adreno 205 is roughly twice as powerful as the original Adreno 200, so 4x the 200 would be twice the current 205's power.
Samsung has also boasted about 4x improvements to 3D graphics. This leads one to believe the IP supplier has made significant strides in core performance.
Quote:
Graphics as Powerful as a PS3 -- in the Palm of Your Hand
Some time in the 2011-2013 window, Qualcomm plans to air the Adreno 3xx which could be its crowning achievement -- if it pulls it off. *The GPU will be made for use with SoCs on the 28 nm node. *Qualcomm claims it will be graphically as powerful as an Xbox 360 or PlayStation 3. *Of course, those console GPUs are paired with lots of graphics RAM so it seems unlikely that the true performance would match these next-gen consoles.
Every year brings a crowning achievement!!!!!!
Just remember we really haven't seen the fruits of Apples PA Semi purchase yet.
Quote:
(Side note: an Adreno GPU will power the upcoming PlayStation Phone.)
Nonetheless, if it can even match the processing power of these consoles, that could make for some impressive smartphone graphics. *The new GPU will support the upcoming OpenGL ES "Haiti", the successor to OpenGL ES 2.0. *It will also jump on the GPU computing bandwagon, adding support for OpenCL 1.1. *What good GPU computing on a smartphone would be seems questionable, but then again, several years back few could predict the growing uses of GPU computing in the PC/server sector today.
What good is OpenCL on a portable device?
1. It allows Apple to accelerate parts of the OS to the benefit of everyone.
2. It keeps SIMD code in maintainable and portable "C" code for developers.
3. It should lower power usage in many cases.
Quote:
Qualcomm is also promising that it will outcompete competitors like NVIDIA in chip cost and size. *
The company is also developing dual-mode chips, which will support both 3G technologies and 4G (LTE).
Conclusions
To put this week's presentation by Qualcomm in perspective, it appears that some very powerful hardware is coming down the pipe. *One thing we see as a clear problem is that the amounts of memory found in current generation smartphones won't be capable of supporting these kinds of phones. *
If hardware partners can work together, to say, triple the memory of current Android models (1 GB for CPU, 512 MB for GPU), then we could have some mighty impressive products on our hands. *Otherwise Qualcomm, NVIDIA, and others miss overshooting on the processing power mark.
This did sound like a bit of an ad for Qualcomm. .
I along with many others expressed concern about Apple going into the ARM processor business all those years ago. The thing Apple needs to watch out for is the rather stiff competition in the ARM SoC marketplace. While you focused primarily on Qualcomm there are actually many design houses making some very impressive SOC. Apple will either set the pace or teail far in the back of the pack. Apple will either be successful or it won't be with its CPU venture.
Any guesses as to when Apple will update it's iMac and MacBook Pro lines then? If they launch the Chips Q1 2011 then I'm guessing earliest.... Late January/Early February, and as late as early Summer?
Anyone?
Quad-Core i7 with 3.4 MHz looks tasty... Especially considering they'll update the Graphics Card too, ATI 5870 or something probably, and possibly even 1600 MHz Ram
Maybe coming out in early April heard from a reliable source.
Maybe coming out in early April heard from a reliable source.
Was that the IMac line or the MacBookPro line? I actually expect a portable update earlier than the iMac line. However Sandy Bridge could be very compelling in both platforms.
So, in the end, will Intel's Sandy Bridge be available for MBP 15/17in during February? If no, why not? Is it because of the Graphics support?
I highly doubt Apple will not have any updates to the Macbook pro early 2011 (not as late as april, maybe feb), im just wondering, will this include Sandy bridge?
So, in the end, will Intel's Sandy Bridge be available for MBP 15/17in during February? If no, why not? Is it because of the Graphics support?
I highly doubt Apple will not have any updates to the Macbook pro early 2011 (not as late as april, maybe feb), im just wondering, will this include Sandy bridge?
Maybe, it all depends on supply and demand. Remember that Apple needs a lot of the newer chips on the ready. This is do to their limited model lineup and use of the same chips across as many models as possible.
I think it’s a safe bet that Sandy Bridge will come to the 15” and 17” MBPs when Apple can get the components it needs. Unfortunately, unless Apple removes the ODD or gets some special deal with Intel to add OpenCL to Sandy Bridge (since OpenCL seems to be a requirement in all new Macs) I don’t think we’ll see Core-i in the ≤ 13" Mac notebooks. I’m hoping we’ll see a revision of the entire Mac notebook line that does remove the ODD and mimics the new MBA design, albeit thicker, to allow for larger batteries, Core-i chips, more performance, all whilst lighting the weight and size.
Also note that the CPU isn’t that important to most users these days. Very few need to have the fastest, most bestest CPU on the market for a given product category. There are many other aspects that can make your system feel faster, and unless you are doing some heavy crunching you really don’t need the fastest CPU. I’ve been recommending going with a newer SSD and a slightly slower CPU for people that want a new machine.
We should keep in mind that even if Apple can get the CPUs it needs there could be other components that cause Apple to delay a release. Maybe they will be spearheading LightPeak or are working on their own USB3.0 controller.
So, in the end, will Intel's Sandy Bridge be available for MBP 15/17in during February?
This may sound like a broken record but how in the hell would anybody on these forums know for sure? Some of us are thousands of miles from Apple HQ.
Quote:
If no, why not? Is it because of the Graphics support?
It could very well be an issue of graphics support. Or they could be waiting for any of a number of other things such as Lightpeak hardware. Plus others have already indicated that hardware demands could force Apple to delay intro while intel ramps up.
Quote:
I highly doubt Apple will not have any updates to the Macbook pro early 2011 (not as late as april, maybe feb), im just wondering, will this include Sandy bridge?
Apple could drag the current models out for months into 2011. It is simply a question of having everything ready to go. As to Sandy Bridge it would be suitable for some models but again you ask the question like you expect people to really know on this forum. The reality is that people who do know can't talk about it. The probability of a Sandy Btidge based portable is pretty high but it is not written in stone.
OCL on the CPU is not very important on the Mac platform. GCD is available to see that all cpu cores are utilized when its useful to do so.
OCL is meaningful to bring the gpu into action when its feasible to do so.
Someone stated that OCL is somehow a requirement for OS X and the Core APIs. This strikes me as false but even if it was true, so long as the OCL on CPU replicated all OCL functionality used by the OS then the performance hit is likely acceptable if the GPU itself can work as well as the 320M. You're simply trading the the advantages of the Sandy Bridge platform for the somewhat minor advantage of OGL acceleration on an embedded GPU. The 320M is not like a Tesla or something.
So a the MB, mini, MBA all having the Sandy Bridge GPU alone is possible. Adding a dedicated GPU to the 13" MBP would be nice.
Someone stated that OCL is somehow a requirement for OS X and the Core APIs. This strikes me as false but even if it was true, so long as the OCL on CPU replicated all OCL functionality used by the OS then the performance hit is likely acceptable if the GPU itself can work as well as the 320M.
Apple provides for a way to fall back to the CPU when the GPU is not available. However this often results in slow ups of ten time or worst. In some cases much worst.
The slow ups are due to multiple things including the factvthat the GPU can act in parallel with the CPU. Plus the GPU is much wider, in other words it can process much more data per cycle than the CPU.
Quote:
You're simply trading the the advantages of the Sandy Bridge platform for the somewhat minor advantage of OGL acceleration on an embedded GPU. The 320M is not like a Tesla or something.
What advantages? The Sandy Bridge platform isn't uniformly faster. Further the opertunities for extremely wide SIMD is not there on a CPU. That given that this is the one area where Sandy Bidge is improved the most.
Quote:
So a the MB, mini, MBA all having the Sandy Bridge GPU alone is possible. Adding a dedicated GPU to the 13" MBP would be nice.
It is certainly possible but maybe not wise. It all depends upon how much of a step backwards it is for the platform. Except for the MBA where right now it appears to be an impossibility.
It all depends upon what you are looking for in the machine you buy. For some uses a Mini with a Core 2 and integrated graphics is enough. However the majority of Mac users are or would not be happy with such a rig. I think many underestimate how important the GPU can be when talking about the overall feel of a machine. Beyound that Apple and others are working on delivering GPU accelerated web browsers and other apps that can really leverage the GPU. It just strikes me as being extremely silly to pull away from fast OpenCL hardware just when all the pieces are coming together.
In the end the last thing Apple needs to do is stagnate GPU performance. People will not want to regress.
Someone stated that OCL is somehow a requirement for OS X and the Core APIs. This strikes me as false but even if it was true, so long as the OCL on CPU replicated all OCL functionality used by the OS then the performance hit is likely acceptable if the GPU itself can work as well as the 320M. You're simply trading the the advantages of the Sandy Bridge platform for the somewhat minor advantage of OGL acceleration on an embedded GPU. The 320M is not like a Tesla or something.
So a the MB, mini, MBA all having the Sandy Bridge GPU alone is possible. Adding a dedicated GPU to the 13" MBP would be nice.
I don't know this for sure but I would bet that in applications that could leverage OCL on a 320m they would be significantly faster than that same application run on a Sandy Bridge processor (cpu only) that would go in a MB, mini or MBA. Those are likely dual core cpus, (maybe with hyperthreading) that'll clock under 3.0 ghz.
IIRC, applications that ran on OCL using a 9400m were done faster than on 2.0 ghz dual C2Ds MPs. That bakeoff was a while ago so I may be mistaken. I bet Marvin would remember and could add more to the discussion.
Its my understanding that massively parallel tasks are far better performed on a GPU. Otherwise our GPUs would look a lot like CPUs, no? Tasks like that are well suited for OCL. Its a shame we don't have more applications that leverage the GPU when its advantageous to do so.
Comments
OpenCL is a hardware abstraction layer for processors. It's raison d'être is to shift some types of processing tasks away from the CPU and onto the GPU.
But it doesn't have to ve used that way.
Intel has the most powerful CPUs on the market, but lags in the GPU space. Logically, they would be against any technology that allows the GPU take over the tasks of the CPU.
To the point of loosing customers? If Intel doesn't have a fast OpenCL implementation that is what will happen.
In fact, there is no indication that Intel will ever support OpenCL properly. They are listed as a member of the Chronos Consortium, which administers OpenCL, but have only produced an initial implementation of OpenCL which runs on the *CPU*.
Which just recently went Alpha. As you note this is a CPU based OpenCL implementation and does not indicate future GPU capability. So we still don't know if Intels Sandy Brige has a gpu that actually helps people
AMD is much more enthusiastic about OpenCL. Too bad Bobcat is a thoroughly low end product.
From what I'm hearing Bobcat would have worked just fine in the AIRs, at least from the GPU standpoint. The problem is information is very tight with respect to actual CPU performance. Supposedly the embargo on benchmarks ends sometime this week, so hopefully we will have a more complete picture soon.
Llano (sp) is still a ways off so who knows. You are right about AMD being more serious, after a slow start they seem to be after OpenCL with gusto. Very aggressive is the word. However the simple fact that they can in fact run OpenCL code well gives them an advantage over Intel.
Perhaps their future chips (Lanos?) will get serious consideration from Apple.
To the point of loosing customers? If Intel doesn't have a fast OpenCL implementation that is what will happen.
I suspect that this is a bit of a showdown between Apple and Intel. Apple must be hopping mad if Intel is really refusing to support OpenCL. On the other hand, Intel's CPUs are so far ahead of the competition (right now, and for the foreseeable future) that Apple can't jump ship.
From what I'm hearing Bobcat would have worked just fine in the AIRs, at least from the GPU standpoint. The problem is information is very tight with respect to actual CPU performance. Supposedly the embargo on benchmarks ends sometime this week, so hopefully we will have a more complete picture soon.
The embargo is over. Here's a review that looks at the top of the line "Zacate" version of Brazos APU (Bobcat + GPU) - whew, so many codenames! It compares Zacate to an Intel Celeron SU2300 CULV processor in combination with an NVIDIA ION graphics / chipset. The Celeron/ION combo is lower end than the Core2Duo/320M in the MacBook Air. (Celeron is similar to the Core2Duo but slower clockspeed and smaller L2 cache, and ION includes essentially the same graphics as the 9400M in the previous generation Air).
http://www.pcper.com/article.php?aid...e=expert&pid=3
http://www.pcper.com/article.php?aid...e=expert&pid=4
http://www.pcper.com/article.php?aid...e=expert&pid=5
http://www.pcper.com/article.php?aid...e=expert&pid=6
Zacate is competitive with Celeron/ION (a bit slower on CPU, slightly better graphics), though it does seem to be a good performer considering its low price and power consumption. It would definitely be a step down from the current MacBook Air internals.
- Will the 13" MBP get a discrete graphics processor or will it rely on Intel's inbuilt one?
- What discrete GPUs will the 15" and 17" MBPs get?
- Is it totally beyond the bounds of possibility that Intel produces an integral GPU that's good enough to replace the discrete GPU on all MBPs?
- Will all MBPs get USB 3.0?
- Will all MBPs get standard 256 Gb SSDs?
- Will the DVD disappear?
- Will we get a hardware redesign too, i.e. a thinner form factor and lighter weight?
So, the big unanswered questions about Sandy Bridge processors and Apple's next MacBook Pro refresh are:
1- Will the 13" MBP get a discrete graphics processor or will it rely on Intel's inbuilt one?
2- What discrete GPUs will the 15" and 17" MBPs get?
3- Is it totally beyond the bounds of possibility that Intel produces an integral GPU that's good enough to replace the discrete GPU on all MBPs?
4- Will all MBPs get USB 3.0?
5- Will all MBPs get standard 256 Gb SSDs?
6- Will the DVD disappear?
7- Will we get a hardware redesign too, i.e. a thinner form factor and lighter weight?
1- I hope it will get Core ix cpu + dedicated gpu even if that means removing the ODD
2- nvidia 4xxM series (10-23W) or Radeon 5650 (15-19W), the current 330M is a 23W part
3- yes
4- probably not
5- not in 2011
6- gradually on the notebooks, the next 13" MBP would be a good start
7- hopefully on the 13" MBP early 2011, later on the 15/17" models, with the inclusion of LightPeak and other things (added or removed)...
Early 2011 MacBooks:
$999 13" alu MacBook (1280*800) 2.40 C2D, 320M, 2GB RAM, ODD (gains the alu enclosure, FW, SD card slot) 10 hours
$1199 13" alu MacBook (1280*800) 2.66 C2D, 320M, 2GB RAM, ODD (much lower price) 10 hours
$1499 13" MacBook Pro (1440*900) 2.50 Core i5-2520M, dedicated gpu, 4GB RAM, no ODD, 8-9 hours
$1799 15" MacBook Pro (1680*1050) 2.50 Core i5-2520M, dedicated gpu, 4GB RAM, ODD, 8-9 hours
$1999 15" MacBook Pro (1680*1050) 2.60 Core i5-2540M, dedicated gpu, 4GB RAM, ODD, 8-9 hours
$2199 15" MacBook Pro (1680*1050) 2.70 Core i7-2620M, dedicated gpu, 4GB RAM, ODD, 8-9 hours
$2399 17" MacBook Pro (1920*1200) 2.70 Core i7-2620M, dedicated gpu, 4GB RAM, ODD, 8-9 hours
I suspect that this is a bit of a showdown between Apple and Intel. Apple must be hopping mad if Intel is really refusing to support OpenCL.
Livid! Lets face it, it will be a very very long time before an Intel CPU can do what a GPU can do well. More so Apple must be looking at AMDs and Intels roadmaps and be wondering why they are doing business with Intel. It really looks like AMD has a vision of the future more in line with Apples.
On the other hand, Intel's CPUs are so far ahead of the competition (right now, and for the foreseeable future) that Apple can't jump ship.
Well this i don't buy. For one thing they don't need to jump ship, rather they just need to pick an choose the best implementations for each product.
Beyound that many of Apples products don't stress ultimate performance but rather other features. For example the Mini is a low power compact device. Similarly the Mac Book and even the AIRs are not CPU power houses, AMD could slot in processors today with similar performance profiles.
As to AMD the little bit of reviewing I've seen so far is very impressive. Remember this is an extremely low power solution with a die not much different in size than ATOM. I do not see Intel offering anything similar.
[quote]
The embargo is over. Here's a review that looks at the top of the line "Zacate" version of Brazos APU (Bobcat + GPU) - whew, so many codenames! It compares Zacate to an Intel Celeron SU2300 CULV processor in combination with an NVIDIA ION graphics / chipset. The Celeron/ION combo is lower end than the Core2Duo/320M in the MacBook Air. (Celeron is similar to the Core2Duo but slower clockspeed and smaller L2 cache, and ION includes essentially the same graphics as the 9400M in the previous generation Air).
[quote]
If I find the time this weekend I will review some of these benchmarks in greater depth. I must admit though these numbers look really good for a processor that sits on a die approximately the size of an ATOM.
http://www.pcper.com/article.php?aid...e=expert&pid=3
http://www.pcper.com/article.php?aid...e=expert&pid=4
http://www.pcper.com/article.php?aid...e=expert&pid=5
http://www.pcper.com/article.php?aid...e=expert&pid=6
Zacate is competitive with Celeron/ION (a bit slower on CPU, slightly better graphics), though it does seem to be a good performer considering its low price and power consumption.
Exactly; excellent performance given the power consumption.
It would definitely be a step down from the current MacBook Air internals.
Well at 1.6 GHz maybe but what will happen at say 2.4 GHz assuming AMD can produce Zacates at that clock rate and maintain thermals at tolerable levels. If the whole chip can handle thermals of around 25 watts would Intel be able to compete?
Even more interesting is that AMD has said that the GPUs of the Fusion products may rev quicker than the rest of the chip. So maybe by the time AIRs are ready for their next rev we will see a suitable Bobcat based fusion product. Or maybe Apple will do something for the low end. Right now Mac Book plastic is poorly placed with the AIRs and 13" MBP all clumped together. Maybe Apple will turn Mac Book into a very low cost laptop. In a nut shell AMD is giving Apple options it has never had before.
Maybe Zacate is less than ideal for AIR today, but it is a damn good start. More so it should allow for surprisingly low cost, low power but well performing devices.
So, the big unanswered questions about Sandy Bridge processors and Apple's next MacBook Pro refresh are:
- Will the 13" MBP get a discrete graphics processor or will it rely on Intel's inbuilt one?
This is really tough to respond to as we simply don't know the full capabilities of Intels Sandy Bridge GPU. My perspective is that it will be to close in oerformance to the AIRs and Mac Books thus Apple will go with a suplemental GPU. The wild card is that they wait for AMDs high performamce Fusion products.
- What discrete GPUs will the 15" and 17" MBPs get?
That is a good question. I would love to see an upgrade of course.
- Is it totally beyond the bounds of possibility that Intel produces an integral GPU that's good enough to replace the discrete GPU on all MBPs?
For pro level use I think it will be a couple of years before that will happen. Apple could try it with Sandy Bridge but I think they would have a user revolt. Beyound that AMD is likely to have better GPUs in their Fusion products for the foreseeable future.
- Will all MBPs get USB 3.0?
Possibly. The rumor is AMD will be supporting USB 3 before Intel.
- Will all MBPs get standard 256 Gb SSDs?
That should be the minimal. What ever they get it will be a SSD.
- Will the DVD disappear?
I really hope so in the 13". But not in the larger machines as again the pros would revolt.
- Will we get a hardware redesign too, i.e. a thinner form factor and lighter weight?
Most certainly.
Looks like the rumours of Sandy Bridge supporting OpenCL may be true.
I looked that over and it only supports OCL on the cpu. Perhaps I missed something but I don't see intel IGP being able to pitch in and help on OCL applications. And it only works on Windows.
Of course this depends upon what you are expecting from the processor but considering is power profile it is pretty impressive. It basically destroys ATOM. Lets face it, it will be a very very long time before an Intel CPU can do what a GPU can do well. More so Apple must be looking at AMDs and Intels roadmaps and be wondering why they are doing business with Intel. It really looks like AMD has a vision of the future more in line with Apples.
Next year, NVIDIA will air its Tegra 2 system-on-a-chip which comes complete with a dual-core ARM9 and super-powered mobile GPU. *Not to be outdone, Qualcomm just announced its own plans for mobile hardware domination and they're shaping up to be equally impressive.
Qualcomm's Chip Plans for 2011
First up will be the pair of previously announced dual CPU core 45 nm system-on-a-chips. *These chips will launch in phones early next year. *Dubbed the MSM8260 and 8660, these system-on-a-chips are powered by the Adreno 205 GPU and two Scorpion cores clocked at 1.2 GHz. *The key difference between the two chip models is in mobile broadcast standards support. *The 8260 only supports HSPA+, while the 8660 supports HSPA+, CDMA2000 and 1xEV-DO Rev. B. *These chips have already been completed and sampled to hardware partners, so phones should be soon coming to market.
This week Qualcomm also disclosed that it would be moving to the 28 nm process node. The first SoC to be produced at the new node will be the MSM8960. *The new SoC won't just be a die shrink; it will also feature a new core design. *While the company refuses to hint at clock speeds, it will say the new chip will be 5 times as powerful as the original (single Scorpion core) SnapDragon, meaning that each core will be roughly 2.5 times as powerful as its predecessor.
The company also claims it can achieve all of that while operating at 75% of the current generation's power (though it was less specific about what kind of power measure and which core -- 45 nm or 65 nm -- it was comparing to). *This claim has been met with a bit of confusion and skepticism, but if it's as good as it sounds, Qualcomm should be good shape.
In the vague boasting department, Qualcomm also bragged that the 8960 would 4x as powerful graphically (as some nonspecific design). *The Adreno 205 is roughly twice as powerful as the original Adreno 200, so 4x the 200 would be twice the current 205's power.
Graphics as Powerful as a PS3 -- in the Palm of Your Hand
Some time in the 2011-2013 window, Qualcomm plans to air the Adreno 3xx which could be its crowning achievement -- if it pulls it off. *The GPU will be made for use with SoCs on the 28 nm node. *Qualcomm claims it will be graphically as powerful as an Xbox 360 or PlayStation 3. *Of course, those console GPUs are paired with lots of graphics RAM so it seems unlikely that the true performance would match these next-gen consoles.
(Side note: an Adreno GPU will power the upcoming PlayStation Phone.)
Nonetheless, if it can even match the processing power of these consoles, that could make for some impressive smartphone graphics. *The new GPU will support the upcoming OpenGL ES "Haiti", the successor to OpenGL ES 2.0. *It will also jump on the GPU computing bandwagon, adding support for OpenCL 1.1. *What good GPU computing on a smartphone would be seems questionable, but then again, several years back few could predict the growing uses of GPU computing in the PC/server sector today.
Qualcomm is also promising that it will outcompete competitors like NVIDIA in chip cost and size. *
The company is also developing dual-mode chips, which will support both 3G technologies and 4G (LTE).
Conclusions
To put this week's presentation by Qualcomm in perspective, it appears that some very powerful hardware is coming down the pipe. *One thing we see as a clear problem is that the amounts of memory found in current generation smartphones won't be capable of supporting these kinds of phones. *
If hardware partners can work together, to say, triple the memory of current Android models (1 GB for CPU, 512 MB for GPU), then we could have some mighty impressive products on our hands. *Otherwise Qualcomm, NVIDIA, and others miss overshooting on the processing power mark.
http://www.dailytech.com/article.aspx?newsid=20194
1st post w00t ... It's been a while.
So Intel's bundlegate continues unabated. Shovelling you useless GPUs...
Given Anand's benchmarks shows it to be about the same performance as the Radeon HD 5450 it's not that bad.
WRT To OpenCL...if OpenCL on the CPU works well enough it might not matter enough Apple if the IGP isn't supporting it if the total package is better then alternatives.
I see fusion on Apple as a longish shot.
WRT To OpenCL...if OpenCL on the CPU works well enough it might not matter enough Apple if the IGP isn't supporting it if the total package is better then alternatives.
OCL on the CPU is not very important on the Mac platform. GCD is available to see that all cpu cores are utilized when its useful to do so.
OCL is meaningful to bring the gpu into action when its feasible to do so.
Next year, NVIDIA will air its Tegra 2 system-on-a-chip which comes complete with a dual-core ARM9 and super-powered mobile GPU. *Not to be outdone, Qualcomm just announced its own plans for mobile hardware domination and they're shaping up to be equally impressive.
This is one of the reasons I'm expecting vastly improved hardware on the iPad 2. These new generations of SoC ARM based hardware sould be very impressive.
Qualcomm's Chip Plans for 2011
First up will be the pair of previously announced dual CPU core 45 nm system-on-a-chips. *These chips will launch in phones early next year. *Dubbed the MSM8260 and 8660, these system-on-a-chips are powered by the Adreno 205 GPU and two Scorpion cores clocked at 1.2 GHz. *The key difference between the two chip models is in mobile broadcast standards support. *The 8260 only supports HSPA+, while the 8660 supports HSPA+, CDMA2000 and 1xEV-DO Rev. B. *These chips have already been completed and sampled to hardware partners, so phones should be soon coming to market.
I doubt very much that Apple will look outside of Apple for the next SoC. However these releases highlight where the industry will be very early next year. They are packing a lot of stuff on to these chips and vastly increasing performance. In many cases they are doing so at a lower power profile.
I see these as hints as to what Apples A4 replacement will look like.
This week Qualcomm also disclosed that it would be moving to the 28 nm process node. The first SoC to be produced at the new node will be the MSM8960. *The new SoC won't just be a die shrink; it will also feature a new core design. *While the company refuses to hint at clock speeds, it will say the new chip will be 5 times as powerful as the original (single Scorpion core) SnapDragon, meaning that each core will be roughly 2.5 times as powerful as its predecessor.
When they hit 28 nm they will actually be ahead of Intel.
With respect to Apple though they have many patents related to CPU optimization. I could see Apple deliver a hybrid ARM design with CPU optimizations for Objective C code. In any event it should be remembered that anything the other players can do on the SoC apple can do too. Plus they can do it without the kitchen sink.
The company also claims it can achieve all of that while operating at 75% of the current generation's power (though it was less specific about what kind of power measure and which core -- 45 nm or 65 nm -- it was comparing to). *This claim has been met with a bit of confusion and skepticism, but if it's as good as it sounds, Qualcomm should be good shape.
I've seen some very impressive power figures for processes in the 22-28nm range so the possibility is there. What is notable here is the additional core.
In the vague boasting department, Qualcomm also bragged that the 8960 would 4x as powerful graphically (as some nonspecific design). *The Adreno 205 is roughly twice as powerful as the original Adreno 200, so 4x the 200 would be twice the current 205's power.
Samsung has also boasted about 4x improvements to 3D graphics. This leads one to believe the IP supplier has made significant strides in core performance.
Graphics as Powerful as a PS3 -- in the Palm of Your Hand
Some time in the 2011-2013 window, Qualcomm plans to air the Adreno 3xx which could be its crowning achievement -- if it pulls it off. *The GPU will be made for use with SoCs on the 28 nm node. *Qualcomm claims it will be graphically as powerful as an Xbox 360 or PlayStation 3. *Of course, those console GPUs are paired with lots of graphics RAM so it seems unlikely that the true performance would match these next-gen consoles.
Every year brings a crowning achievement!!!!!!
Just remember we really haven't seen the fruits of Apples PA Semi purchase yet.
(Side note: an Adreno GPU will power the upcoming PlayStation Phone.)
Nonetheless, if it can even match the processing power of these consoles, that could make for some impressive smartphone graphics. *The new GPU will support the upcoming OpenGL ES "Haiti", the successor to OpenGL ES 2.0. *It will also jump on the GPU computing bandwagon, adding support for OpenCL 1.1. *What good GPU computing on a smartphone would be seems questionable, but then again, several years back few could predict the growing uses of GPU computing in the PC/server sector today.
What good is OpenCL on a portable device?
1. It allows Apple to accelerate parts of the OS to the benefit of everyone.
2. It keeps SIMD code in maintainable and portable "C" code for developers.
3. It should lower power usage in many cases.
Qualcomm is also promising that it will outcompete competitors like NVIDIA in chip cost and size. *
The company is also developing dual-mode chips, which will support both 3G technologies and 4G (LTE).
Conclusions
To put this week's presentation by Qualcomm in perspective, it appears that some very powerful hardware is coming down the pipe. *One thing we see as a clear problem is that the amounts of memory found in current generation smartphones won't be capable of supporting these kinds of phones. *
If hardware partners can work together, to say, triple the memory of current Android models (1 GB for CPU, 512 MB for GPU), then we could have some mighty impressive products on our hands. *Otherwise Qualcomm, NVIDIA, and others miss overshooting on the processing power mark.
http://www.dailytech.com/article.aspx?newsid=20194
This did sound like a bit of an ad for Qualcomm.
I along with many others expressed concern about Apple going into the ARM processor business all those years ago. The thing Apple needs to watch out for is the rather stiff competition in the ARM SoC marketplace. While you focused primarily on Qualcomm there are actually many design houses making some very impressive SOC. Apple will either set the pace or teail far in the back of the pack. Apple will either be successful or it won't be with its CPU venture.
Any guesses as to when Apple will update it's iMac and MacBook Pro lines then? If they launch the Chips Q1 2011 then I'm guessing earliest.... Late January/Early February, and as late as early Summer?
Anyone?
Quad-Core i7 with 3.4 MHz looks tasty... Especially considering they'll update the Graphics Card too, ATI 5870 or something probably, and possibly even 1600 MHz Ram
Maybe coming out in early April heard from a reliable source.
Maybe coming out in early April heard from a reliable source.
Was that the IMac line or the MacBookPro line? I actually expect a portable update earlier than the iMac line. However Sandy Bridge could be very compelling in both platforms.
I highly doubt Apple will not have any updates to the Macbook pro early 2011 (not as late as april, maybe feb), im just wondering, will this include Sandy bridge?
So, in the end, will Intel's Sandy Bridge be available for MBP 15/17in during February? If no, why not? Is it because of the Graphics support?
I highly doubt Apple will not have any updates to the Macbook pro early 2011 (not as late as april, maybe feb), im just wondering, will this include Sandy bridge?
Maybe, it all depends on supply and demand. Remember that Apple needs a lot of the newer chips on the ready. This is do to their limited model lineup and use of the same chips across as many models as possible.
I think it’s a safe bet that Sandy Bridge will come to the 15” and 17” MBPs when Apple can get the components it needs. Unfortunately, unless Apple removes the ODD or gets some special deal with Intel to add OpenCL to Sandy Bridge (since OpenCL seems to be a requirement in all new Macs) I don’t think we’ll see Core-i in the ≤ 13" Mac notebooks. I’m hoping we’ll see a revision of the entire Mac notebook line that does remove the ODD and mimics the new MBA design, albeit thicker, to allow for larger batteries, Core-i chips, more performance, all whilst lighting the weight and size.
Also note that the CPU isn’t that important to most users these days. Very few need to have the fastest, most bestest CPU on the market for a given product category. There are many other aspects that can make your system feel faster, and unless you are doing some heavy crunching you really don’t need the fastest CPU. I’ve been recommending going with a newer SSD and a slightly slower CPU for people that want a new machine.
We should keep in mind that even if Apple can get the CPUs it needs there could be other components that cause Apple to delay a release. Maybe they will be spearheading LightPeak or are working on their own USB3.0 controller.
So, in the end, will Intel's Sandy Bridge be available for MBP 15/17in during February?
This may sound like a broken record but how in the hell would anybody on these forums know for sure? Some of us are thousands of miles from Apple HQ.
If no, why not? Is it because of the Graphics support?
It could very well be an issue of graphics support. Or they could be waiting for any of a number of other things such as Lightpeak hardware. Plus others have already indicated that hardware demands could force Apple to delay intro while intel ramps up.
I highly doubt Apple will not have any updates to the Macbook pro early 2011 (not as late as april, maybe feb), im just wondering, will this include Sandy bridge?
Apple could drag the current models out for months into 2011. It is simply a question of having everything ready to go. As to Sandy Bridge it would be suitable for some models but again you ask the question like you expect people to really know on this forum. The reality is that people who do know can't talk about it. The probability of a Sandy Btidge based portable is pretty high but it is not written in stone.
OCL on the CPU is not very important on the Mac platform. GCD is available to see that all cpu cores are utilized when its useful to do so.
OCL is meaningful to bring the gpu into action when its feasible to do so.
Someone stated that OCL is somehow a requirement for OS X and the Core APIs. This strikes me as false but even if it was true, so long as the OCL on CPU replicated all OCL functionality used by the OS then the performance hit is likely acceptable if the GPU itself can work as well as the 320M. You're simply trading the the advantages of the Sandy Bridge platform for the somewhat minor advantage of OGL acceleration on an embedded GPU. The 320M is not like a Tesla or something.
So a the MB, mini, MBA all having the Sandy Bridge GPU alone is possible. Adding a dedicated GPU to the 13" MBP would be nice.
Someone stated that OCL is somehow a requirement for OS X and the Core APIs. This strikes me as false but even if it was true, so long as the OCL on CPU replicated all OCL functionality used by the OS then the performance hit is likely acceptable if the GPU itself can work as well as the 320M.
Apple provides for a way to fall back to the CPU when the GPU is not available. However this often results in slow ups of ten time or worst. In some cases much worst.
The slow ups are due to multiple things including the factvthat the GPU can act in parallel with the CPU. Plus the GPU is much wider, in other words it can process much more data per cycle than the CPU.
You're simply trading the the advantages of the Sandy Bridge platform for the somewhat minor advantage of OGL acceleration on an embedded GPU. The 320M is not like a Tesla or something.
What advantages? The Sandy Bridge platform isn't uniformly faster. Further the opertunities for extremely wide SIMD is not there on a CPU. That given that this is the one area where Sandy Bidge is improved the most.
So a the MB, mini, MBA all having the Sandy Bridge GPU alone is possible. Adding a dedicated GPU to the 13" MBP would be nice.
It is certainly possible but maybe not wise. It all depends upon how much of a step backwards it is for the platform. Except for the MBA where right now it appears to be an impossibility.
It all depends upon what you are looking for in the machine you buy. For some uses a Mini with a Core 2 and integrated graphics is enough. However the majority of Mac users are or would not be happy with such a rig. I think many underestimate how important the GPU can be when talking about the overall feel of a machine. Beyound that Apple and others are working on delivering GPU accelerated web browsers and other apps that can really leverage the GPU. It just strikes me as being extremely silly to pull away from fast OpenCL hardware just when all the pieces are coming together.
In the end the last thing Apple needs to do is stagnate GPU performance. People will not want to regress.
Someone stated that OCL is somehow a requirement for OS X and the Core APIs. This strikes me as false but even if it was true, so long as the OCL on CPU replicated all OCL functionality used by the OS then the performance hit is likely acceptable if the GPU itself can work as well as the 320M. You're simply trading the the advantages of the Sandy Bridge platform for the somewhat minor advantage of OGL acceleration on an embedded GPU. The 320M is not like a Tesla or something.
So a the MB, mini, MBA all having the Sandy Bridge GPU alone is possible. Adding a dedicated GPU to the 13" MBP would be nice.
I don't know this for sure but I would bet that in applications that could leverage OCL on a 320m they would be significantly faster than that same application run on a Sandy Bridge processor (cpu only) that would go in a MB, mini or MBA. Those are likely dual core cpus, (maybe with hyperthreading) that'll clock under 3.0 ghz.
IIRC, applications that ran on OCL using a 9400m were done faster than on 2.0 ghz dual C2Ds MPs. That bakeoff was a while ago so I may be mistaken. I bet Marvin would remember and could add more to the discussion.
Its my understanding that massively parallel tasks are far better performed on a GPU. Otherwise our GPUs would look a lot like CPUs, no? Tasks like that are well suited for OCL. Its a shame we don't have more applications that leverage the GPU when its advantageous to do so.