I like seeing the anger, disappointment etc in this thread. Feel my pain.
The reason I am not posting is because John.B, nvidia2008 and nht are covering all the bullet points so well.
Intel's IGP are crap, plain and simple. If they double performance every year for the next ten years, they will still be crap relative to all other options available. "Sucking less than previous Intel IGPs" does not qualify as an 'accomplishment.' It qualifies as Moore's Law plus clockspeed increases.
Benchmarks that make Intel IGP look good are somehow borked.
Anand's benchmarks were for a desktop (higher clocked), 12 EU version (mobile will be only 6).
OpenCL has left a yawning chasm that AMD alone can fill. Let's hope Apple carpe the diem.
Intel's IGP are crap, plain and simple. If they double performance every year for the next ten years, they will still be crap relative to all other options available.
Where's the point where they become good enough that it doesn't matter though?
Right now, a 320M is about half of what an XBox 360 is capable of and it can play any game that's out today - that's about 3,500 graphically intense games. The highest end games push it to the limit but even on low quality, a game like Metro 2033 looks good:
Years ago, it used to matter a lot to get the highest GPU and CPU as their capability was so low. There also weren't that many decent games available. We're not at that stage any more.
If Intel manage to get in the region of the 320M with this revision, it's a side-step but Ivy Bridge could double it again within 12 months. Then you have XBox 360 performance in a laptop. Nobody complains about that quality.
You might say that in 2012, new consoles could arrive with new standards to meet but will the games really look much better than they do now and will PC versions be prevented from ramping down to suit lower hardware? I think not. From seeing the graphics quality that is in games right now compared to post-production graphics, I don't think we are much more than a couple of generations away from having enough visual quality that it doesn't matter if you only have a low-end GPU.
If you compare the entry iMac to the highest-end desktop card you can buy that is a single card and not two jammed into a single unit, the difference is just 2x. With the dual card models, it's still under 3x.
This means that you can either have a huge desktop tower consuming 200-300W running games at 50-75FPS with the fans blasting away or you have a Macbook Air running them at 25FPS consuming under 35W. The 320M stays extremely cool while playing.
It's slightly hard to compare as the Mini one has an external camera but the experience is comparable.
Quote:
Originally Posted by 1337_5L4Xx0R
Anand's benchmarks were for a desktop (higher clocked), 12 EU version (mobile will be only 6).
The desktop ones have the same clock as the comparative mobile ones and the mobile ones all have 12 EUs. It's the desktop models that come with either 6 or 12 EUs.
Fair points Marvin. But... For me, console quality graphics is a very low standard nowadays. I'm trying NFS Hot Pursuit on my PC and not only is it a shitty port with stupid unnecessary lags but the overall 3D environment and quality really feels "cheap". Even CoDMW2 had some scenes where the background sky was clearly one pixellated single image plane. Consoles have really mangled PC gaming from it's wonderful heights. Look at HAWX2... To think that a flight sim in 2010 looks and plays like that is a little disturbing.
But your point about console gaming being the standard that everyone expects and enjoys is also true. There is a certain aesthetic and gameplay, once reached, that makes it acceptable and successful, despite technical deficiencies compared to full-fledged PC titles like Starcraft2.
One thing that kills me about consoles is the lack of proper antialiasing, which results in most cases causing the PC port to also lack antialiasing, even though mid range PC GPUs can easily chew on 2xAA or 4xAA at 1920x1080.
On another note 320M is impressive but I think 4x that is just about a nice setting to be "good enough for most people not to have to care anymore". I'm obviously biased but I think the ATI 5850 for gaming performance and Nvidia 460 for GPGPU etc is the "benchmark" for 2011-2015. Any next gen console would be killer using at least that level of graphics quality. Also, there is no getting by the fact that dedicated 1GB VRAM does absolute wonders for all kinds of stuff at 1080p.
I was initially skeptical about DX10 but implemented wisely and on decent hardware the shaders, aesthetic, realism, etc. is quite nice. DX11, while mainly focused on tessellation, is also quite impressive. Run UniEngine and some of the latest DX11 titles on a very high-end PC and it is like going from standard def to HD, you really might not want to go back to anything else.
Gameplay is of course an entirely different issue. Cut the Rope on my iPad has been really quite good. Not so much Angry Birds - that's just pure digital cocaine - you get the rush but the edginess too when you can't complete a level!
I like seeing the anger, disappointment etc in this thread. Feel my pain.
Which should be saved until something ships! Otherwise we have people speculating on what they think a processor is capable of. If Apple ships a bunch of Mac Laptops which show a step backwards in performance then I might very well join you in releasing anger. However I'm not going to jump the gun on this issue and will instead simply wait for Apple to debut something.
[quote]
The reason I am not posting is because John.B, nvidia2008 and nht are covering all the bullet points so well.
Intel's IGP are crap, plain and simple. If they double performance every year for the next ten years, they will still be crap relative to all other options available. "Sucking less than previous Intel IGPs" does not qualify as an 'accomplishment.' It qualifies as Moore's Law plus clockspeed increases.
[/quote[
The current intel IGPU's do suck, there is no way around that, All I'm suggesting is avoiding passing judgment upon the The coming SB chips. In the end you need an open mind until intel convinces you to close it.
Quote:
Benchmarks that make Intel IGP look good are somehow borked.
Well I don't have much respect for the people doing the prerelease testing as I really think they are part of Intels marketing campaign.
However what are you going to say if some respectable people get the hardware, test it and post results showing better than 320M performance? It is pretty silly to think that it is impossible for Intel to do a good GPU.
Quote:
Anand's benchmarks were for a desktop (higher clocked), 12 EU version (mobile will be only 6).
This is one of the guys I have trouble with.
Quote:
OpenCL has left a yawning chasm that AMD alone can fill. Let's hope Apple carpe the diem.
This is again a questionable point of view. AMD definitely has a plan that may or may not pan out. Intel on the other hand has decided to keep us in the dak
There is a certain aesthetic and gameplay, once reached, that makes it acceptable and successful, despite technical deficiencies compared to full-fledged PC titles like Starcraft2.
One thing that kills me about consoles is the lack of proper antialiasing, which results in most cases causing the PC port to also lack antialiasing, even though mid range PC GPUs can easily chew on 2xAA or 4xAA at 1920x1080.
I agree entirely. I think consoles have badly affected PC gaming simply because that's where the focus is now and PCs gaming is optional. 007 Blood Stone is one of the worst game ports I've seen recently. Clearly it came to the PC as an afterthought.
But it's also to do with volume. The costs to make the game are so high ($100m) that to make a return on a $50 game, you have to sell over 2 million copies. Only big titles like CoD are doing this successfully. A lot of other studios are just dropping out. This is why the IGP market is so important because it makes up over 50% of all shipped computers (about 150 million target) vs a fraction of 30% for high-end gaming (probably 5% = 15 million people).
Quote:
Originally Posted by nvidia2008
On another note 320M is impressive but I think 4x that is just about a nice setting to be "good enough for most people not to have to care anymore". I'm obviously biased but I think the ATI 5850 for gaming performance and Nvidia 460 for GPGPU etc is the "benchmark" for 2011-2015.
I would agree that 4x current mobile hardware would suffice for pretty much anything you need visually. This could be a reality with with Intel IGP in 3 years. We'll probably have 6 or even 8-core CPUs in the entry level by then.
Jan 2011: Sandy Bridge, dual-core, IGP = 320M, 32nm architecture
Jan 2014: Brickland die-shrink, 6-core mobile (48EUs) = 5850/5870 desktop, 16nm die-shrink
The dedicated GPU will be dead in 2014 and I'd say that large towers will stop being manufactured for consumers. Obviously dedicated GPUs move on too but the visual bar that nobody needs the 300W tower for will ensure that desktop sales are eroded to under 10% and we'll all have MBAs with 1TB SSD a 6-core CPU and an Intel IGP.
Quote:
Originally Posted by nvidia2008
Gameplay is of course an entirely different issue. Cut the Rope on my iPad has been really quite good. Not so much Angry Birds - that's just pure digital cocaine - you get the rush but the edginess too when you can't complete a level!
The good thing about the GPU plateau is that it allows more of a creative focus on content without having to worry that the game won't run on however many systems.
My Gawd. Case in point. Just installed COD Black Ops tonight. What an absolute bloody nightmare. First of all you have to go through Steam. Which is normally fine but Internet is rubbish in my country.
More concerning though is the well known lag issues due to something or other even in single player. To this date Treyarch has released a few updates and mentions that they are still "working on issues". Update and update later there is still some weird lagging as though the CPU is pegged and stuttering.
The state of PC gaming is abysmal. The hardware power is there. But the whole Windows PC model combined with game development economics and/or rampant greed and/or challenges with piracy makes PC gaming one big pile of rubbish to wade through.
I am still desperate enough though that I might struggle through COD Black Ops because there is a gaming experience there not available on Mac or iPad or iPhone. Same for NFS Hot Pursuit ... 30 minutes a day of brainless racing is tolerable.
I'm ready to throw the PC out the window and get an Xbox360. Worse thing is though, for who knows what reason it is not officially available in Malaysia so the grey market... Well it seems pretty scummy to me. Plus I have to get a HDMI converter or 1080p HDTV and so on.
AAAAARARARARARARRRGGGHHHHHHHHHH
I really swear, Apple is so close to moving into console gaming, they should just make the jump and do it. Put PC gaming well and truly out of its misery oh I just don't know anymore
PC gaming may well pass, but to think that Apple will do it, to me, is utterly laughable. I'm bailing on Mac hardware this xmas, when I buy/build a PC that can actually competently run 3D apps. PC gaming is SOTA. Macs are, uh, behind the curve.
I agree entirely. I think consoles have badly affected PC gaming simply because that's where the focus is now and PCs gaming is optional. 007 Blood Stone is one of the worst game ports I've seen recently. Clearly it came to the PC as an afterthought.
I'm not a gamer so I don't care from that perspective. At least I never was but Angery Birds and a few others on my iPhone are a big distraction. However I think this highlights at least on potential reason for Apples disinterest in gaming on the Mac, that is the consoles have won.
Quote:
But it's also to do with volume. The costs to make the game are so high ($100m) that to make a return on a $50 game, you have to sell over 2 million copies. Only big titles like CoD are doing this successfully. A lot of other studios are just dropping out. This is why the IGP market is so important because it makes up over 50% of all shipped computers (about 150 million target) vs a fraction of 30% for high-end gaming (probably 5% = 15 million people).
the cost of production is likely another reason that gaming consoles are favored. Like it or not developers need to make a buck or two and on the PC there is no way to really do that.
Quote:
I would agree that 4x current mobile hardware would suffice for pretty much anything you need visually. This could be a reality with with Intel IGP in 3 years. We'll probably have 6 or even 8-core CPUs in the entry level by then.
Display tech has a very long ways to go. Imagine a Retina like resolution on 35" monitor. Even at half of that you are still talking about a lot of pixels to push. I don't see the demand for faster GPUs stopping anytime soon.
Quote:
Jan 2011: Sandy Bridge, dual-core, IGP = 320M, 32nm architecture
Jan 2014: Brickland die-shrink, 6-core mobile (48EUs) = 5850/5870 desktop, 16nm die-shrink
AMD has indicated that they intend to rev the GPU sections of their Fusion processors faster than the CPU portion. I actually think we will see a race here to put as much GPU into these processors as is possible. The demand is certainly there.
Quote:
The dedicated GPU will be dead in 2014 and I'd say that large towers will stop being manufactured for consumers. Obviously dedicated GPUs move on too but the visual bar that nobody needs the 300W tower for will ensure that desktop sales are eroded to under 10% and we'll all have MBAs with 1TB SSD a 6-core CPU and an Intel IGP.
I'm not sure I agree with the points above. For one the process shrinks that alllow for on die GPUs also allow for far more capable discrete GPUs. I just don't see a need for an end to discrete GPUs. They may become more specialized and less mainstream but I'd be shocked if they where gone by 2014.
As for towers there are lots of reasons to buy such. So I don't see them leaving either. On the other hand somthing like Apples Mini will be an extremely powerful platform when the 22nm processes hit. More so we are entering the age where a ten watt computer might actually be very useful. This is the flip side of the process improvements we have seen, having a device optimized for low power doesn't mean that it has to suck.
This brings us back to Apples laptops and the rumors of Sandy Bridge only GPUs in the MacBook. I know this might not be the case for many in this thread but a lower power chip, that is one that runs cooler and has a long battery life may be a very hot seller. The reality is many Mac Book owners don't care about 3D.
Quote:
The good thing about the GPU plateau is that it allows more of a creative focus on content without having to worry that the game won't run on however many systems.
This is likely why Apple has had so much success with iOS devices. The capabilities from one model to the next are clearly defined. There is little mystery about how an iPhone 4 performs compared to an iPad. However I don't see a plateau as much as I see ckear and well defined rungs on a ladder. For example I'm expecting a clear jump in performance when iPad 2 comes out. This ends up being another incremental rung on the ladder to higher performing systems.
On the Macs, especially the Mac Books, I don't see a plateau either. The transition to Sandy Bridge IGP (if it actually happens) is just that a transition. It is somthing that has to happen and as you note there will be rapid advancements in performance afterward. Again this can be seen as a rung on a ladder but it is a new ladder going in a slightly different direction.
All of this hand wringing about the move to Intel integrated graphics will likely be over in a few weeks. We will either have or not have laptop systems based on Sandy Bridge, if we do we can get a clearer picture of just what the goods and bass are. Due to the new architecture I'm expecting both outstanding good points as well as a few bad points. Will the new processor make for a good gaming platform? I kinda doubt it but is that is not what most people are looking for.
Display tech has a very long ways to go. Imagine a Retina like resolution on 35" monitor. Even at half of that you are still talking about a lot of pixels to push. I don't see the demand for faster GPUs stopping anytime soon.
The current iMac display could probably be considered a retina display already because you sit 30" away from it not 12" like a phone. So if you have 960 x 640 on a 3.5" screen at 12" away, you only need the iPad's 132ppi density for a computer. So a 21.5" would need 2400 x 1600, the 27" would need 3000 x 2000 and a 35" would need 3800 x 2500 or thereabouts. That's only about 15% more than we have now.
You don't need to push that many pixels in a game though because the textures would have to be that high too to be worthwhile and the game itself would be huge. Given that consumer movies won't go beyond 1080p, that can be the benchmark for games too and the 5850 handles that easily with Bad Company 2 at maximum and 4xAA. There's not many places to go when a card can play a game like the following:
I just don't see a need for an end to discrete GPUs. They may become more specialized and less mainstream but I'd be shocked if they where gone by 2014.
Right now 70% of all machines shipped are laptops, over 50% of all machines only have IGPs. That trend towards mobile is increasing. Some research suggests desktops won't drop to 10% in 2014 like I suggested but 19%:
A portion of those machines will still push the higher end GPUs for the remaining enthusiasts but I don't think that market will be big enough to sustain the companies developing the dedicated GPUs. As the market lowers, the prices have to go up, which continues the market decline until they give up. AMD's tagline is "the future is fusion" - the only way to survive is if you have both CPU and GPU together. Sadly for NVidia, they only have one of the parts they need, Intel and AMD have both.
Quote:
Originally Posted by wizard69
On the Macs, especially the Mac Books, I don't see a plateau either. The transition to Sandy Bridge IGP (if it actually happens) is just that a transition. It is somthing that has to happen and as you note there will be rapid advancements in performance afterward. Again this can be seen as a rung on a ladder but it is a new ladder going in a slightly different direction.
I think there is a trend towards a plateau. A certain level of performance beyond which, there's no demand for it. Inevitably come the quotes from people who have said 'no one will ever need more than x' and have been proved wrong but we can see it today with the rise in popularity of ultra-mobile devices. You see it in TVs, which are at a point where you walk into a store and you have screens bigger than you are and you'd likely opt for one you can actually get into your car.
When it comes to machines, it's a case of 'what's the cheapest machine I can get that does what I need'. If you're a casual gamer, you can get away with the lowest end Macbook Air:
If its too good to be true, it probably isn't true.
What I've been reading on this is that it should equal low end discreet gpu's. That would be very good, but it doesn't mean it will run OpenCL, which is even more important to Apple.
Intel won, not by being better but by being bigger and pushing them out of the way. It had to happen but NVidia don't deserve to go out this way.
NVidia has made plenty of mistakes. They didn't exactly endear themselves to manufacturers over the past three years with their lying about their gpu board problems that cost manufacturers hundreds of millions.
They have also borked up demos. a few years ago they were found to be cheating by cropping off-screen computations that were supposed to be done.
In addition, i trust Intel's statements and demos more than those from AMD. Coming off their performance a couple of years ago, I find it hard to believe anything they say or do.
In addition, i trust Intel's statements and demos more than those from AMD. Coming off their performance a couple of years ago, I find it hard to believe anything they say or do.
I have to ask exactly what you're referring to here.
The current iMac display could probably be considered a retina display already because you sit 30" away from it not 12" like a phone. So if you have 960 x 640 on a 3.5" screen at 12" away, you only need the iPad's 132ppi density for a computer. So a 21.5" would need 2400 x 1600, the 27" would need 3000 x 2000 and a 35" would need 3800 x 2500 or thereabouts. That's only about 15% more than we have now.
Sadly my ability to see pixels gets worst every year. That being said I still see room for improvement in desktop displays.
Quote:
You don't need to push that many pixels in a game though because the textures would have to be that high too to be worthwhile and the game itself would be huge. Given that consumer movies won't go beyond 1080p, that can be the benchmark for games too and the 5850 handles that easily with Bad Company 2 at maximum and 4xAA. There's not many places to go when a card can play a game like the following:
Right now 70% of all machines shipped are laptops, over 50% of all machines only have IGPs. That trend towards mobile is increasing. Some research suggests desktops won't drop to 10% in 2014 like I suggested but 19%:
The trend to mobile is there but it isn't towards laptops. With the advent of smart phones and tablets I could see many people moving back to desktop machines. The desktop being the primary computer and the smart phone or tablet being the mobile solution.
A portion of those machines will still push the higher end GPUs for the remaining enthusiasts but I don't think that market will be big enough to sustain the companies developing the dedicated GPUs. As the market lowers, the prices have to go up, which continues the market decline until they give up. AMD's tagline is "the future is fusion" - the only way to survive is if you have both CPU and GPU together. Sadly for NVidia, they only have one of the parts they need, Intel and AMD have both.
NVidia is in a very tough but not impossible situation. A better management team might save the company. As to high end GPUs demand may come to an end some day, I just think 2014 is a little early.
Quote:
I think there is a trend towards a plateau. A certain level of performance beyond which, there's no demand for it.
I think this is what I disagree with the most. In fact I would go so far as to say we need far more powerful computers to enable a new generation of software and capabilities.
Maybe I'm biased as my old 2008 MBP is to sluggish at times. Of course that could be OS/X but sometimes the machine seems to have a mind of it's own.
Quote:
Inevitably come the quotes from people who have said 'no one will ever need more than x' and have been proved wrong but we can see it today with the rise in popularity of ultra-mobile devices. You see it in TVs, which are at a point where you walk into a store and you have screens bigger than you are and you'd likely opt for one you can actually get into your car.
the interesting thing here is that computer performance does not equate with size. There may be a good reason for that large screen purchase. An ultra mobile device likewise serves specific needs. However a smart phone is not a substitute for a desktop computer.
At least not yet!
Quote:
When it comes to machines, it's a case of 'what's the cheapest machine I can get that does what I need'. If you're a casual gamer, you can get away with the lowest end Macbook Air:
Intel's IGP are crap, plain and simple. If they double performance every year for the next ten years, they will still be crap relative to all other options available.
If they double yearly for 10 years, that would mean the Intel IGP of 2020 will be 1024x more powerful (2^10) than the current IGP. I don't think that is an achievable improvement level, but if it is we all win
The major difference between mobile Sandy Bridge and its desktop countpart is all mobile SB launch SKUs have two graphics cores (12 EUs), while only some desktop parts have 12 EUs (it looks like the high-end K SKUs will have it).
You're completely wrong. Desktop version will have 6 SKUs, but mobile will have 12.
Strange (desktops are usually more powerful than mobile) but true!
I don't think it's that strange. If 70% of the computing market is mobile now, and there aren't exactly many aftermarket video options for laptops, then it makes perfect sense to make sure that the laptop version has as beefy a video option as possible. The desktop will be getting bypassed by people buying aftermarket videocard options in many cases.
I know that contradicts the near-term plateau I mentioned but it has to happen somewhere between 2014-2018 because they are reaching the fabrication limits for electronic components. They can make the die bigger/stacked perhaps, switch to optical transistors or find ways to ramp up the clock speeds without overheating the chips but they will run into issues.
It'll be interesting to see what happens when we reach that point because regardless of demand, they might not be able to make better chips.
Quote:
Originally Posted by FuturePastNow
I have to ask exactly what you're referring to here.
I was thinking the same thing. When the GMA 950 came to replace the Radeon 9200 etc, Intel and partners were trying to claim that IGPs weren't like IGPs of the past, which turned out to be wrong. They were slow, incompatible with poor feature support and far outperformed by the competition. This has happened with every GPU they've ever made.
With CPUs, they certainly outperform the competition though so I guess that's something but they have a lot of smoke and mirrors like their vanishing act with Larrabee. So much hype and it didn't amount to what they said it would (when they said it would anyway).
They made good moves with SSD but still under-delivered on sequential write. Light Peak will probably be good though.
Quote:
Originally Posted by wizard69
In fact I would go so far as to say we need far more powerful computers to enable a new generation of software and capabilities.
Web-based software certainly requires more capable hardware to be able to run Javascript and other content quickly enough so that will drive hardware forward.
Quote:
Originally Posted by wizard69
If they double yearly for 10 years, that would mean the Intel IGP of 2020 will be 1024x more powerful (2^10) than the current IGP. I don't think that is an achievable improvement level, but if it is we all win
They tend to double every 2 years so in 10 years, it's only 32x faster and the manufacturers will hit the manufacturing difficulties before then. It will fall short of current render farms:
but it should be plenty for photoreal rendering in real-time. The important drive needs to be on results and needs both software and hardware to be implemented smartly, not just with speedups on both sides. That's one impressive part about SB putting very fast fixed function architecture in there. Once people decide on the best algorithms, they should just put them in silicon to make them thousands of times faster. At first it seems like general purpose solutions would be better as those setups are more flexible but post-production rendering and games and mathematical algorithms are being refined to a point where people know what algorithms to run, they just need it done quickly.
As long as they don't exclude general purpose capability, I think that will end up being the best design.
Comments
The reason I am not posting is because John.B, nvidia2008 and nht are covering all the bullet points so well.
Intel's IGP are crap, plain and simple. If they double performance every year for the next ten years, they will still be crap relative to all other options available. "Sucking less than previous Intel IGPs" does not qualify as an 'accomplishment.' It qualifies as Moore's Law plus clockspeed increases.
Benchmarks that make Intel IGP look good are somehow borked.
Anand's benchmarks were for a desktop (higher clocked), 12 EU version (mobile will be only 6).
OpenCL has left a yawning chasm that AMD alone can fill. Let's hope Apple carpe the diem.
Intel's IGP are crap, plain and simple. If they double performance every year for the next ten years, they will still be crap relative to all other options available.
Where's the point where they become good enough that it doesn't matter though?
Right now, a 320M is about half of what an XBox 360 is capable of and it can play any game that's out today - that's about 3,500 graphically intense games. The highest end games push it to the limit but even on low quality, a game like Metro 2033 looks good:
http://www.youtube.com/watch?v=2hexGRRDZIY
Years ago, it used to matter a lot to get the highest GPU and CPU as their capability was so low. There also weren't that many decent games available. We're not at that stage any more.
If Intel manage to get in the region of the 320M with this revision, it's a side-step but Ivy Bridge could double it again within 12 months. Then you have XBox 360 performance in a laptop. Nobody complains about that quality.
You might say that in 2012, new consoles could arrive with new standards to meet but will the games really look much better than they do now and will PC versions be prevented from ramping down to suit lower hardware? I think not. From seeing the graphics quality that is in games right now compared to post-production graphics, I don't think we are much more than a couple of generations away from having enough visual quality that it doesn't matter if you only have a low-end GPU.
If you compare the entry iMac to the highest-end desktop card you can buy that is a single card and not two jammed into a single unit, the difference is just 2x. With the dual card models, it's still under 3x.
This means that you can either have a huge desktop tower consuming 200-300W running games at 50-75FPS with the fans blasting away or you have a Macbook Air running them at 25FPS consuming under 35W. The 320M stays extremely cool while playing.
Mafia 2 maximum quality + PhysX:
http://www.youtube.com/watch?v=N78L_iy7IZ8
Here's Mafia 2 medium quality out on a high-end GTX 480:
http://www.youtube.com/watch?v=iwGKpghgWoc
Here's the same game, same road on a Mac Mini 2010:
http://www.youtube.com/watch?v=xvOW3GzWAeo
Here's the iMac 4850 (similar to the 330M performance):
http://www.youtube.com/watch?v=MisekqMerEw
It's slightly hard to compare as the Mini one has an external camera but the experience is comparable.
Anand's benchmarks were for a desktop (higher clocked), 12 EU version (mobile will be only 6).
The desktop ones have the same clock as the comparative mobile ones and the mobile ones all have 12 EUs. It's the desktop models that come with either 6 or 12 EUs.
But your point about console gaming being the standard that everyone expects and enjoys is also true. There is a certain aesthetic and gameplay, once reached, that makes it acceptable and successful, despite technical deficiencies compared to full-fledged PC titles like Starcraft2.
One thing that kills me about consoles is the lack of proper antialiasing, which results in most cases causing the PC port to also lack antialiasing, even though mid range PC GPUs can easily chew on 2xAA or 4xAA at 1920x1080.
On another note 320M is impressive but I think 4x that is just about a nice setting to be "good enough for most people not to have to care anymore". I'm obviously biased but I think the ATI 5850 for gaming performance and Nvidia 460 for GPGPU etc is the "benchmark" for 2011-2015. Any next gen console would be killer using at least that level of graphics quality. Also, there is no getting by the fact that dedicated 1GB VRAM does absolute wonders for all kinds of stuff at 1080p.
I was initially skeptical about DX10 but implemented wisely and on decent hardware the shaders, aesthetic, realism, etc. is quite nice. DX11, while mainly focused on tessellation, is also quite impressive. Run UniEngine and some of the latest DX11 titles on a very high-end PC and it is like going from standard def to HD, you really might not want to go back to anything else.
Gameplay is of course an entirely different issue. Cut the Rope on my iPad has been really quite good. Not so much Angry Birds - that's just pure digital cocaine - you get the rush but the edginess too when you can't complete a level!
I like seeing the anger, disappointment etc in this thread. Feel my pain.
Which should be saved until something ships! Otherwise we have people speculating on what they think a processor is capable of. If Apple ships a bunch of Mac Laptops which show a step backwards in performance then I might very well join you in releasing anger. However I'm not going to jump the gun on this issue and will instead simply wait for Apple to debut something.
[quote]
The reason I am not posting is because John.B, nvidia2008 and nht are covering all the bullet points so well.
Intel's IGP are crap, plain and simple. If they double performance every year for the next ten years, they will still be crap relative to all other options available. "Sucking less than previous Intel IGPs" does not qualify as an 'accomplishment.' It qualifies as Moore's Law plus clockspeed increases.
[/quote[
The current intel IGPU's do suck, there is no way around that, All I'm suggesting is avoiding passing judgment upon the The coming SB chips. In the end you need an open mind until intel convinces you to close it.
Benchmarks that make Intel IGP look good are somehow borked.
Well I don't have much respect for the people doing the prerelease testing as I really think they are part of Intels marketing campaign.
However what are you going to say if some respectable people get the hardware, test it and post results showing better than 320M performance? It is pretty silly to think that it is impossible for Intel to do a good GPU.
Anand's benchmarks were for a desktop (higher clocked), 12 EU version (mobile will be only 6).
This is one of the guys I have trouble with.
OpenCL has left a yawning chasm that AMD alone can fill. Let's hope Apple carpe the diem.
This is again a questionable point of view. AMD definitely has a plan that may or may not pan out. Intel on the other hand has decided to keep us in the dak
There is a certain aesthetic and gameplay, once reached, that makes it acceptable and successful, despite technical deficiencies compared to full-fledged PC titles like Starcraft2.
One thing that kills me about consoles is the lack of proper antialiasing, which results in most cases causing the PC port to also lack antialiasing, even though mid range PC GPUs can easily chew on 2xAA or 4xAA at 1920x1080.
I agree entirely. I think consoles have badly affected PC gaming simply because that's where the focus is now and PCs gaming is optional. 007 Blood Stone is one of the worst game ports I've seen recently. Clearly it came to the PC as an afterthought.
But it's also to do with volume. The costs to make the game are so high ($100m) that to make a return on a $50 game, you have to sell over 2 million copies. Only big titles like CoD are doing this successfully. A lot of other studios are just dropping out. This is why the IGP market is so important because it makes up over 50% of all shipped computers (about 150 million target) vs a fraction of 30% for high-end gaming (probably 5% = 15 million people).
On another note 320M is impressive but I think 4x that is just about a nice setting to be "good enough for most people not to have to care anymore". I'm obviously biased but I think the ATI 5850 for gaming performance and Nvidia 460 for GPGPU etc is the "benchmark" for 2011-2015.
I would agree that 4x current mobile hardware would suffice for pretty much anything you need visually. This could be a reality with with Intel IGP in 3 years. We'll probably have 6 or even 8-core CPUs in the entry level by then.
Jan 2011: Sandy Bridge, dual-core, IGP = 320M, 32nm architecture
Jan 2012: Ivy Bridge, quad-core mobile, 2x IGP (24EUs) = GTX 460M, 22nm die-shrink
Jan 2013: Brickland, quad-core mobile, 3x IGP (24EUs) = GTS 250 desktop, 22nm architecture
Jan 2014: Brickland die-shrink, 6-core mobile (48EUs) = 5850/5870 desktop, 16nm die-shrink
The dedicated GPU will be dead in 2014 and I'd say that large towers will stop being manufactured for consumers. Obviously dedicated GPUs move on too but the visual bar that nobody needs the 300W tower for will ensure that desktop sales are eroded to under 10% and we'll all have MBAs with 1TB SSD a 6-core CPU and an Intel IGP.
Gameplay is of course an entirely different issue. Cut the Rope on my iPad has been really quite good. Not so much Angry Birds - that's just pure digital cocaine - you get the rush but the edginess too when you can't complete a level!
The good thing about the GPU plateau is that it allows more of a creative focus on content without having to worry that the game won't run on however many systems.
More concerning though is the well known lag issues due to something or other even in single player. To this date Treyarch has released a few updates and mentions that they are still "working on issues". Update and update later there is still some weird lagging as though the CPU is pegged and stuttering.
The state of PC gaming is abysmal. The hardware power is there. But the whole Windows PC model combined with game development economics and/or rampant greed and/or challenges with piracy makes PC gaming one big pile of rubbish to wade through.
I am still desperate enough though that I might struggle through COD Black Ops because there is a gaming experience there not available on Mac or iPad or iPhone. Same for NFS Hot Pursuit ... 30 minutes a day of brainless racing is tolerable.
I'm ready to throw the PC out the window and get an Xbox360. Worse thing is though, for who knows what reason it is not officially available in Malaysia so the grey market... Well it seems pretty scummy to me. Plus I have to get a HDMI converter or 1080p HDTV and so on.
AAAAARARARARARARRRGGGHHHHHHHHHH
I really swear, Apple is so close to moving into console gaming, they should just make the jump and do it. Put PC gaming well and truly out of its misery oh I just don't know anymore
I agree entirely. I think consoles have badly affected PC gaming simply because that's where the focus is now and PCs gaming is optional. 007 Blood Stone is one of the worst game ports I've seen recently. Clearly it came to the PC as an afterthought.
I'm not a gamer so I don't care from that perspective. At least I never was but Angery Birds and a few others on my iPhone are a big distraction. However I think this highlights at least on potential reason for Apples disinterest in gaming on the Mac, that is the consoles have won.
But it's also to do with volume. The costs to make the game are so high ($100m) that to make a return on a $50 game, you have to sell over 2 million copies. Only big titles like CoD are doing this successfully. A lot of other studios are just dropping out. This is why the IGP market is so important because it makes up over 50% of all shipped computers (about 150 million target) vs a fraction of 30% for high-end gaming (probably 5% = 15 million people).
the cost of production is likely another reason that gaming consoles are favored. Like it or not developers need to make a buck or two and on the PC there is no way to really do that.
I would agree that 4x current mobile hardware would suffice for pretty much anything you need visually. This could be a reality with with Intel IGP in 3 years. We'll probably have 6 or even 8-core CPUs in the entry level by then.
Display tech has a very long ways to go. Imagine a Retina like resolution on 35" monitor. Even at half of that you are still talking about a lot of pixels to push. I don't see the demand for faster GPUs stopping anytime soon.
Jan 2011: Sandy Bridge, dual-core, IGP = 320M, 32nm architecture
Jan 2012: Ivy Bridge, quad-core mobile, 2x IGP (24EUs) = GTX 460M, 22nm die-shrink
Jan 2013: Brickland, quad-core mobile, 3x IGP (24EUs) = GTS 250 desktop, 22nm architecture
Jan 2014: Brickland die-shrink, 6-core mobile (48EUs) = 5850/5870 desktop, 16nm die-shrink
AMD has indicated that they intend to rev the GPU sections of their Fusion processors faster than the CPU portion. I actually think we will see a race here to put as much GPU into these processors as is possible. The demand is certainly there.
The dedicated GPU will be dead in 2014 and I'd say that large towers will stop being manufactured for consumers. Obviously dedicated GPUs move on too but the visual bar that nobody needs the 300W tower for will ensure that desktop sales are eroded to under 10% and we'll all have MBAs with 1TB SSD a 6-core CPU and an Intel IGP.
I'm not sure I agree with the points above. For one the process shrinks that alllow for on die GPUs also allow for far more capable discrete GPUs. I just don't see a need for an end to discrete GPUs. They may become more specialized and less mainstream but I'd be shocked if they where gone by 2014.
As for towers there are lots of reasons to buy such. So I don't see them leaving either. On the other hand somthing like Apples Mini will be an extremely powerful platform when the 22nm processes hit. More so we are entering the age where a ten watt computer might actually be very useful. This is the flip side of the process improvements we have seen, having a device optimized for low power doesn't mean that it has to suck.
This brings us back to Apples laptops and the rumors of Sandy Bridge only GPUs in the MacBook. I know this might not be the case for many in this thread but a lower power chip, that is one that runs cooler and has a long battery life may be a very hot seller. The reality is many Mac Book owners don't care about 3D.
The good thing about the GPU plateau is that it allows more of a creative focus on content without having to worry that the game won't run on however many systems.
This is likely why Apple has had so much success with iOS devices. The capabilities from one model to the next are clearly defined. There is little mystery about how an iPhone 4 performs compared to an iPad. However I don't see a plateau as much as I see ckear and well defined rungs on a ladder. For example I'm expecting a clear jump in performance when iPad 2 comes out. This ends up being another incremental rung on the ladder to higher performing systems.
On the Macs, especially the Mac Books, I don't see a plateau either. The transition to Sandy Bridge IGP (if it actually happens) is just that a transition. It is somthing that has to happen and as you note there will be rapid advancements in performance afterward. Again this can be seen as a rung on a ladder but it is a new ladder going in a slightly different direction.
All of this hand wringing about the move to Intel integrated graphics will likely be over in a few weeks. We will either have or not have laptop systems based on Sandy Bridge, if we do we can get a clearer picture of just what the goods and bass are. Due to the new architecture I'm expecting both outstanding good points as well as a few bad points. Will the new processor make for a good gaming platform? I kinda doubt it but is that is not what most people are looking for.
Display tech has a very long ways to go. Imagine a Retina like resolution on 35" monitor. Even at half of that you are still talking about a lot of pixels to push. I don't see the demand for faster GPUs stopping anytime soon.
The current iMac display could probably be considered a retina display already because you sit 30" away from it not 12" like a phone. So if you have 960 x 640 on a 3.5" screen at 12" away, you only need the iPad's 132ppi density for a computer. So a 21.5" would need 2400 x 1600, the 27" would need 3000 x 2000 and a 35" would need 3800 x 2500 or thereabouts. That's only about 15% more than we have now.
You don't need to push that many pixels in a game though because the textures would have to be that high too to be worthwhile and the game itself would be huge. Given that consumer movies won't go beyond 1080p, that can be the benchmark for games too and the 5850 handles that easily with Bad Company 2 at maximum and 4xAA. There's not many places to go when a card can play a game like the following:
http://www.youtube.com/watch?v=7HGqsKjDOLc
I just don't see a need for an end to discrete GPUs. They may become more specialized and less mainstream but I'd be shocked if they where gone by 2014.
Right now 70% of all machines shipped are laptops, over 50% of all machines only have IGPs. That trend towards mobile is increasing. Some research suggests desktops won't drop to 10% in 2014 like I suggested but 19%:
http://techcrunch.com/2010/06/17/for...sell-netbooks/
A portion of those machines will still push the higher end GPUs for the remaining enthusiasts but I don't think that market will be big enough to sustain the companies developing the dedicated GPUs. As the market lowers, the prices have to go up, which continues the market decline until they give up. AMD's tagline is "the future is fusion" - the only way to survive is if you have both CPU and GPU together. Sadly for NVidia, they only have one of the parts they need, Intel and AMD have both.
On the Macs, especially the Mac Books, I don't see a plateau either. The transition to Sandy Bridge IGP (if it actually happens) is just that a transition. It is somthing that has to happen and as you note there will be rapid advancements in performance afterward. Again this can be seen as a rung on a ladder but it is a new ladder going in a slightly different direction.
I think there is a trend towards a plateau. A certain level of performance beyond which, there's no demand for it. Inevitably come the quotes from people who have said 'no one will ever need more than x' and have been proved wrong but we can see it today with the rise in popularity of ultra-mobile devices. You see it in TVs, which are at a point where you walk into a store and you have screens bigger than you are and you'd likely opt for one you can actually get into your car.
When it comes to machines, it's a case of 'what's the cheapest machine I can get that does what I need'. If you're a casual gamer, you can get away with the lowest end Macbook Air:
http://www.youtube.com/watch?v=RmtBXr9VH4s
http://www.youtube.com/watch?v=m-bqbFRsgcI
Will the new processor make for a good gaming platform? I kinda doubt it but is that is not what most people are looking for.
If it can play Starcraft 2 and Mass Effect 2 in the way they demoed, I'd say it will be a good gaming platform just like the 320M machines.
Since most Apple customers are not very technically sophisticated, this story is irrelevant for the majority.
Neither are most PC users.
That doesn't make sense.
If its too good to be true, it probably isn't true.
What I've been reading on this is that it should equal low end discreet gpu's. That would be very good, but it doesn't mean it will run OpenCL, which is even more important to Apple.
Intel won, not by being better but by being bigger and pushing them out of the way. It had to happen but NVidia don't deserve to go out this way.
NVidia has made plenty of mistakes. They didn't exactly endear themselves to manufacturers over the past three years with their lying about their gpu board problems that cost manufacturers hundreds of millions.
They have also borked up demos. a few years ago they were found to be cheating by cropping off-screen computations that were supposed to be done.
In addition, i trust Intel's statements and demos more than those from AMD. Coming off their performance a couple of years ago, I find it hard to believe anything they say or do.
If it can play Starcraft 2 and Mass Effect 2 in the way they demoed, I'd say it will be a good gaming platform just like the 320M machines.
Mass Efffect 3. Already done with 2.
I hope there's a Sandy Bridge mini...
In addition, i trust Intel's statements and demos more than those from AMD. Coming off their performance a couple of years ago, I find it hard to believe anything they say or do.
I have to ask exactly what you're referring to here.
The current iMac display could probably be considered a retina display already because you sit 30" away from it not 12" like a phone. So if you have 960 x 640 on a 3.5" screen at 12" away, you only need the iPad's 132ppi density for a computer. So a 21.5" would need 2400 x 1600, the 27" would need 3000 x 2000 and a 35" would need 3800 x 2500 or thereabouts. That's only about 15% more than we have now.
Sadly my ability to see pixels gets worst every year. That being said I still see room for improvement in desktop displays.
You don't need to push that many pixels in a game though because the textures would have to be that high too to be worthwhile and the game itself would be huge. Given that consumer movies won't go beyond 1080p, that can be the benchmark for games too and the 5850 handles that easily with Bad Company 2 at maximum and 4xAA. There's not many places to go when a card can play a game like the following:
http://www.youtube.com/watch?v=7HGqsKjDOLc
Right now 70% of all machines shipped are laptops, over 50% of all machines only have IGPs. That trend towards mobile is increasing. Some research suggests desktops won't drop to 10% in 2014 like I suggested but 19%:
The trend to mobile is there but it isn't towards laptops. With the advent of smart phones and tablets I could see many people moving back to desktop machines. The desktop being the primary computer and the smart phone or tablet being the mobile solution.
http://techcrunch.com/2010/06/17/for...sell-netbooks/
A portion of those machines will still push the higher end GPUs for the remaining enthusiasts but I don't think that market will be big enough to sustain the companies developing the dedicated GPUs. As the market lowers, the prices have to go up, which continues the market decline until they give up. AMD's tagline is "the future is fusion" - the only way to survive is if you have both CPU and GPU together. Sadly for NVidia, they only have one of the parts they need, Intel and AMD have both.
NVidia is in a very tough but not impossible situation. A better management team might save the company. As to high end GPUs demand may come to an end some day, I just think 2014 is a little early.
I think there is a trend towards a plateau. A certain level of performance beyond which, there's no demand for it.
I think this is what I disagree with the most. In fact I would go so far as to say we need far more powerful computers to enable a new generation of software and capabilities.
Maybe I'm biased as my old 2008 MBP is to sluggish at times. Of course that could be OS/X but sometimes the machine seems to have a mind of it's own.
Inevitably come the quotes from people who have said 'no one will ever need more than x' and have been proved wrong but we can see it today with the rise in popularity of ultra-mobile devices. You see it in TVs, which are at a point where you walk into a store and you have screens bigger than you are and you'd likely opt for one you can actually get into your car.
the interesting thing here is that computer performance does not equate with size. There may be a good reason for that large screen purchase. An ultra mobile device likewise serves specific needs. However a smart phone is not a substitute for a desktop computer.
At least not yet!
When it comes to machines, it's a case of 'what's the cheapest machine I can get that does what I need'. If you're a casual gamer, you can get away with the lowest end Macbook Air:
http://www.youtube.com/watch?v=RmtBXr9VH4s
http://www.youtube.com/watch?v=m-bqbFRsgcI
If it can play Starcraft 2 and Mass Effect 2 in the way they demoed, I'd say it will be a good gaming platform just like the 320M machines.
Well again I'm not much of a gamer so I have to take your word for it.
Intel's IGP are crap, plain and simple. If they double performance every year for the next ten years, they will still be crap relative to all other options available.
If they double yearly for 10 years, that would mean the Intel IGP of 2020 will be 1024x more powerful (2^10) than the current IGP. I don't think that is an achievable improvement level, but if it is we all win
Also they are 2xtimes more powerful than Arrandale's GPU.
Anand's benchmarks were for a desktop (higher clocked), 12 EU version (mobile will be only 6).
You're completely wrong. Desktop version will have 6 SKUs, but mobile will have 12.
Strange (desktops are usually more powerful than mobile) but true!
http://www.anandtech.com/show/3876/i...bridge-part-ii
The major difference between mobile Sandy Bridge and its desktop countpart is all mobile SB launch SKUs have two graphics cores (12 EUs), while only some desktop parts have 12 EUs (it looks like the high-end K SKUs will have it).
You're completely wrong. Desktop version will have 6 SKUs, but mobile will have 12.
Strange (desktops are usually more powerful than mobile) but true!
I don't think it's that strange. If 70% of the computing market is mobile now, and there aren't exactly many aftermarket video options for laptops, then it makes perfect sense to make sure that the laptop version has as beefy a video option as possible. The desktop will be getting bypassed by people buying aftermarket videocard options in many cases.
Mass Efffect 3. Already done with 2.
I hope there's a Sandy Bridge mini...
Same here, it would be nice if SB can run ME3 sufficiently. One day I'm sure games will look like the good trailers too:
http://www.youtube.com/watch?v=qZGFjBmD41Q
I know that contradicts the near-term plateau I mentioned but it has to happen somewhere between 2014-2018 because they are reaching the fabrication limits for electronic components. They can make the die bigger/stacked perhaps, switch to optical transistors or find ways to ramp up the clock speeds without overheating the chips but they will run into issues.
It'll be interesting to see what happens when we reach that point because regardless of demand, they might not be able to make better chips.
I have to ask exactly what you're referring to here.
With CPUs, they certainly outperform the competition though so I guess that's something but they have a lot of smoke and mirrors like their vanishing act with Larrabee. So much hype and it didn't amount to what they said it would (when they said it would anyway).
They made good moves with SSD but still under-delivered on sequential write. Light Peak will probably be good though.
In fact I would go so far as to say we need far more powerful computers to enable a new generation of software and capabilities.
Web-based software certainly requires more capable hardware to be able to run Javascript and other content quickly enough so that will drive hardware forward.
If they double yearly for 10 years, that would mean the Intel IGP of 2020 will be 1024x more powerful (2^10) than the current IGP. I don't think that is an achievable improvement level, but if it is we all win
They tend to double every 2 years so in 10 years, it's only 32x faster and the manufacturers will hit the manufacturing difficulties before then. It will fall short of current render farms:
http://www.slashfilm.com/cool-stuff-...s-renderfarms/
but it should be plenty for photoreal rendering in real-time. The important drive needs to be on results and needs both software and hardware to be implemented smartly, not just with speedups on both sides. That's one impressive part about SB putting very fast fixed function architecture in there. Once people decide on the best algorithms, they should just put them in silicon to make them thousands of times faster. At first it seems like general purpose solutions would be better as those setups are more flexible but post-production rendering and games and mathematical algorithms are being refined to a point where people know what algorithms to run, they just need it done quickly.
As long as they don't exclude general purpose capability, I think that will end up being the best design.