Fusion Microsoft = graphics/gaming pwnage?

[cbn]

i think it's because computer bottleneck is really excited about Llano and hopes one day that we will have wii-sized APU systems that can run ps3/xbox/windows stuff smoothly. except it's just not fast enough.

I think Llano will make some nice cheap systems, but I am hoping the Bulldozer spec APUs will be much better. (re: 32nm stars cpu could end up being much weaker than dual core sandy bridge.)

 

[alyarb]

i'll be honest since it's late and i'm having a drink: ever since anand's article (nov 30 my god has it been that long) about AMD's plan with bulldozer i'm less excited. AMD has us believing that we were getting an eight core part, but we aren't getting that. it's going to basically be a doublewide quad core whose halves posses a dedicated scheduler and data cache so that you get 8 physical threads, but the FPU units are not duplicated. although they are 256-bit and they can do FMA, you're not getting an 8 core's worth of them. this is a problem because AMD sticks with their architectures and processes for too long. they need to be focused on an enthusiast platform for magny cours similar to X58 but they haven't given us any sign of that. sandy bridge is bringing 8-threaded quads with 256-bit AVX and an IGP to our segment as well as native hex-cores without IGP (hopefully on 1366), so AMD is going to lose to intel there, and you're right, a 4-threaded Llano will lose to a 4-threaded sandy bridge, but Llano will probably have a better IGP.

cell processor on the other hand is already in the 200+ gigaflops class of GPU. synthetically it can probably do more but for ps3 games let just say 200, and the chip in the xbox is closer to 100. these are small CPUs that have had more FPU in 2006 than intel and AMD will have in 2011. sandy bridge, though 32nm, will have 5 times as many transistors as a cell processor and consume 3 times as much power. it's a little unfair because of the sandy IGP but it's close enough. these chips are just better for games, and v8envy makes a good point with the h.264 encoding as well. the huge problem with consoles right now is not enough RAM. not enough video RAM either. not sure what they were thinking.

 

[Kuzi]

I think Llano will make some nice cheap systems, but I am hoping the Bulldozer spec APUs will be much better. (re: 32nm stars cpu could end up being much weaker than dual core sandy bridge.)

APUs should do really well in the notebook/netbook space and small form factor desktops. I'm expecting over 50% power draw reduction on all 32nm AMD processors, they are adding many features that will help them with that. You have the shrink from 45nm to 32nm, power gating, and the use of HKMG among other features.

It's interesting because Intel and AMD are taking slightly different approaches with their APUs. On one hand we have Sandy offering very high IPC crunching power, with anemic graphics performance. And on the other hand, we got Llano, having average IPC coupled with a comparatively powerful GPU.

 

[cbn]

APUs should do really well in the notebook/netbook space and small form factor desktops. I'm expecting over 50% power draw reduction on all 32nm AMD processors, they are adding many features that will help them with that. You have the shrink from 45nm to 32nm, power gating, and the use of HKMG among other features.

It's interesting because Intel and AMD are taking slightly different approaches with their APUs. On one hand we have Sandy offering very high IPC crunching power, with anemic graphics performance. And on the other hand, we got Llano, having average IPC coupled with a comparatively powerful GPU.

I do like the power saving features on the 32nm Llano. The design essentially looks like the pinnacle of Phenom II design (sans L3 cache).

I guess it depends on how much stronger Sandy Bridge dual core ends up being in comparison to Clarkdale? (With weaker than expected memory performance on Clarkdale, Sandy Bridge still has some good room for improvement). Also according to a recent Fudzilla article (link coming) Sandy Bridge dual core is supposed to be a good deal faster frequency wise.

EDIT: My mistake the fudzilla article is not comparing desktop CPUs. It is comparing mobile sandy bridge vs arrandale. Here is the lnk--->http://www.fudzilla.com/content/view/17713/66/

 

[alyarb]

the mobile aspect doesn't really matter. arrandale also has an off-die, high latency memory controller and sandy bridge is going to murder it, especially at 4 ghz and a lower tdp.

 

[grimpr]

Dont judge Bulldozer by Intels metrics and standards, judge the design in its future potential, at 22nm nodes and below, its uncore capabilities (which we dont have a clue), its potential fusion with GPU's, its scalability, throughput, and power/performance/manufacturing costs when all is revealed. BD was designed to overcome future manufacturing burdens, their costs and mixing up with GPUs.

 

[Kuzi]

i'll be honest since it's late and i'm having a drink: ever since anand's article (nov 30 my god has it been that long) about AMD's plan with bulldozer i'm less excited. AMD has us believing that we were getting an eight core part, but we aren't getting that. it's going to basically be a doublewide quad core whose halves posses a dedicated scheduler and data cache so that you get 8 physical threads, but the FPU units are not duplicated. although they are 256-bit and they can do FMA, you're not getting an 8 core's worth of them. this is a problem because AMD sticks with their architectures and processes for too long. they need to be focused on an enthusiast platform for magny cours similar to X58 but they haven't given us any sign of that. sandy bridge is bringing 8-threaded quads with 256-bit AVX and an IGP to our segment as well as native hex-cores without IGP (hopefully on 1366), so AMD is going to lose to intel there, and you're right, a 4-threaded Llano will lose to a 4-threaded sandy bridge, but Llano will probably have a better IGP.

While Sandy is going to have the most powerful x86 CPU core, it will still use Hyperthreading which is not the most elegant multi-threading approach. Each Sandy core will have it's own 256bit wide FPU, but when a second thread needs to run that FPU (and integer units) still need to be shared.

A Bulldozer model has two integer units with a large shared 256bit FPU, and AMD is going to equate each Bulldozer model with one Sandy core. So for example if we take a Quad-core Sandy, it will have 4x256bit FPU units, and can run 8 threads with HT. A 4 Model Bulldozer would also have 4x256bit FPU units, and can run 8 threads on 8 "real" cores, not shared like what Sandy does.

For highly threaded software that use over 4 threads, Bulldozer "should" actually have an edge there. And the higher the thread count, the better Bulldozer will look. Right now we still don't have any IPC numbers, but if I was to make a guess, Sandy would offer ~10% more performance over i7, of course if AVX is used the gain would be much higher.

Bulldozer needs to offer 15-20% just to catch up to the i5/i7 per core IPC, and that's not an easy task. AMD has their work cut out for them, so it remains to be seen how they do, but I have high hopes for Bulldozer.

 

[Kuzi]

EDIT: My mistake the fudzilla article is not comparing desktop CPUs. It is comparing mobile sandy bridge vs arrandale. Here is the lnk--->http://www.fudzilla.com/content/view/17713/66/

They are counting the clock speed advantage too, actual IPC per clock may not be that high. Look at i7 as an example, it was only about 8% faster than Core2 per clock.

 

[ZimZum]

Mobile gaming was long dominated by Nintendo with Sony coming in a distant second. That was until Apple stormed onto the scene with its iPhone and stole the show. Today the iPhone is arguably mobile gaming's most influential platform.

Microsoft hopes to steal Apple's thunder when it releases Windows Phone 7 Series phones later this year. This week Microsoft shared some details about what gaming on the new operating system will look like.

Unlike the Zune and Courier prototype, which are both powered by NVIDIA's Tegra, most Windows Phone 7 Series phones will use a Qualcomm SnapDragon chipset that contains a GPU from NVIDIA-rival AMD. The GPU, the AMD Z430, features a unified pixel & vertex shader pipeline (based on the Xbox 360 Xenos GPU).

http://www.dailytech.com/Windows+Ph...ing+480x800+Xbox+Live+Titles/article17881.htm

 

[Acanthus]

Of course the Xbox 3 will be an ATi product.

There are only 2 games in town and MS and Nvidia have a bad relationship from the Xbox 1 contract wars.

 

[cbn]

Speaking of gaming on phones and Consoles.

Does anyone have speculation on what Steam (from Valve Corp) might be planning?

No doubt, Steam would like to see its market presence increase, not decrease.

 

[BenSkywalker]

I don't think it would be all that hard to compete with Nintendo on the game play front. As I find the vast majority of Wii games boring after about 5 minutes or so.

That just shows how out of touch you are with that market. MarioKart for the Wii is almost assured to outsell the best selling PC game this year, despite coming out years ago. It is a very simplistic pick up and play title and it resonates strongly with a certain demographic. It may not be what you or I are looking for in a game, but the level of refinement on the simplistic mechanics they have is something MS has never come close to demonstrating they were capable of replicating. I would assume this is why they bought Rare, unfortunately it seems that without Nintendo helping them along Rare isn't that good of a developer themselves.

It looks like with Natal MS wants to go after both markets. Casual and hardcore gamers.

MS is a distant third in the casual games market, not even close to Sony(Flower, LBP etc) who themselves aren't in Nin's league. Natal can try and be whatever they decide, without the games that appeal to the market it won't work- and so far they have demonstrated nothing that indicates they are capable of making those types of games. That shouldn't be considered a real slam on MS as so far noone has had any real success on the Wii outside of Nin. They are operating with a different mindset and have proven for decades that they are the best at what they do. MS has yet to prove it with a single game.

Does anyone have speculation on what Steam (from Valve Corp) might be planning?

On the consoles? If anything, getting laughed at. MS/Sony/Nin own all the cards in that market, they won't allow anything like Steam to compete with their own services, it would just be stupid business.

 

[grimpr]

Globalfoundries/Dresden will soon ramp up 32nm production on AMD's Fusion Llano MPU/GPU chip.

3-D interconnects on research, possibility in 2012.

http://www.semiconductor.net/article/453306-Dresden_Fab_Transformation_Underway.php

 

[crazylegs]

Speaking of gaming on phones and Consoles.

Does anyone have speculation on what Steam (from Valve Corp) might be planning?

No doubt, Steam would like to see its market presence increase, not decrease.

Totally off topic but do you really need 7 double spaced thread links in your sig? All of which go to posts on this very forum...

 

[Headfoot]

emulating x86 on POWER is not fast and you can't virtualize it on nonnative architectures.

Exactly. They won't break backwards compatibility with 360 games if they can help it. The old Xbox used a 700 mhz celeron, and moving architectures to an IBM 3 core derivative for the 360 hurt compatibility. More than likely they will just move on to the latest POWER architecture from IBM (or some derivative thereof) so that players can keep their old 360 games working (and keep Natal working too).

The entire reason Microsoft is successful is due in large part to their legacy support.

 

[Headfoot]

That just shows how out of touch you are with that market. MarioKart for the Wii is almost assured to outsell the best selling PC game this year, despite coming out years ago. It is a very simplistic pick up and play title and it resonates strongly with a certain demographic. It may not be what you or I are looking for in a game, but the level of refinement on the simplistic mechanics they have is something MS has never come close to demonstrating they were capable of replicating. I would assume this is why they bought Rare, unfortunately it seems that without Nintendo helping them along Rare isn't that good of a developer themselves.

Not necessarily true. The Wii may have casual gaming down in one sense, but they are NOT the only players in that market. One very, very important factor to consider is the recent surge in cheap indie titles on both PC (via Steam) and Xbox 360 (via Marketplace). Many of these indie titles rely entirely on gameplay (much like nintendo) but aren't very "nintendo" like (more mature for a lack of a better word).

For example, games like Braid, World of Goo, Plants vs Zombies and Audiosurf were hugely popular on PC because of their gameplay. I can see that style of cheap and easy gameplay used to capture the casual market away from Nintendo.

I like casual games but I dont like having some flavor of Mario in every single casual game I own

 

[sxr7171]

The PS2 is still in its lifetime(it sold millions of units last year and had some games chart too). Just because the 360 has an expected lifetime of ~8 years, doesn't mean the 720 won't come out much sooner then that(see PS2). I figured MS would be looking at 2012 at the latest, and it seems to me that 2011 may be more likely. Going with DVD9 helped MS out in the short run, but it is seriously showing its' age at this point. By 2013 DVD9 will be looking like a floppy.



If we were to assume that is correct, it still wouldn't be enough to include backwards compatability. Little endian processors do not emulate big endian very well, moving back to x86 would be a big step backwards(as a point of reference, the fastest i7 chip still can't go toe to toe with the PS3's CPU for SP FP- not scaled or adjusted for build process or what it would take to emulate it, straight out head to head running perfect optimized code for both).

x86 is a very bad architecture for extremely low power useage and an extremely bad processor for specialized purposes. It is extremely good at being a flexible multi purpose solution.



Tegra2 utterly trounces AMD currently, Intel has from Atom to Itanium. While neither Intel or nV may provide ideal top to bottom solutions, between the two of them it is hard to imagine AMD becoming superior on both the high end and low side and everything in between when they are so far behind on so many different fronts right now. AMD had a really good launch for a graphics card with nV shipping quite a bit behind them. Outside of that they haven't excelled at anything in quite a while. While I'm sure your conjecture plays out exactly how AMD would love to see things happen, I don't see AMD having the engineering resources to come close to making that a viable reality within the next few years(not that they won't release what they are saying they will, but it being superior to what nV/Intel combined can offer in every segment).

NV and Intel aren't friends really. If anything will Intel just buy NV already?

 

[sxr7171]

Not necessarily true. The Wii may have casual gaming down in one sense, but they are NOT the only players in that market. One very, very important factor to consider is the recent surge in cheap indie titles on both PC (via Steam) and Xbox 360 (via Marketplace). Many of these indie titles rely entirely on gameplay (much like nintendo) but aren't very "nintendo" like (more mature for a lack of a better word).

For example, games like Braid, World of Goo, Plants vs Zombies and Audiosurf were hugely popular on PC because of their gameplay. I can see that style of cheap and easy gameplay used to capture the casual market away from Nintendo.

I like casual games but I dont like having some flavor of Mario in every single casual game I own

We are in the minority judging by console sales. While I think personally Nintendo failed me this generation that thing is still selling like crazy at $200 - the same price as the 360 arcade.

People like little cute boxes and cute little games I guess.

 

[BenSkywalker]

Not necessarily true. The Wii may have casual gaming down in one sense, but they are NOT the only players in that market. One very, very important factor to consider is the recent surge in cheap indie titles on both PC (via Steam) and Xbox 360 (via Marketplace). Many of these indie titles rely entirely on gameplay (much like nintendo) but aren't very "nintendo" like (more mature for a lack of a better word).

Just for perspective-

Today, the firm released a new report on the Xbox Live Arcade marketplace, which FADE says generated $103.4 million in 2009, which is a 34% increase year-over-year.

By comparison, Mario Kart Wii has cleared $1Billion in revenue generation. Not all the Mario games, just the singular title. Yes, I would say that Nintendo has an absolute lock on that market just as Windows does on the desktop OS space. No, it isn't 100%, but it's close enough that the discussion revolves around marginal exceptions.

I can see that style of cheap and easy gameplay used to capture the casual market away from Nintendo.

I've hear all sorts of speculation about how Mac OS and Linux will do the same to Windows. No matter how valid certain points are- nothing has changed much. It is the same in the casual game space. Yes, flash based on line games are certainly increasing in popularity, but their rise seems to be in lock step with the Wii's rise. Casual gaming is certainly a rapidly growing market, it just happens to be one that Nintendo owns.

I have a Wii for my kids, and sure it is a cool concept, but if I'm not playing with my kids it doesn't get turned on(and even they have moved almost entirely over to PC/PS3/360 gaming at this point). I don't like the type of games they are making, been hugely let down by the direction they have taken with Zelda and Metroid and pretty much lost interest in anything they make outside of party game uses(and even then, GuitarHero/RockBand are increasingly filling that slot) but hard numbers are difficult to dispute. Nintendo dominates that market segment.

 

[GodisanAtheist]

Intel timeline shows it having their own fusion well ahead of AMD. Intel has already released i3 and i5 CPUs with on chip (but not on die) CPUs. They are about a year and half ahead of AMD in terms of "fusion".
Nvidia recently bought VIA (after years of hiring CPU hardware designers), when you buy a via nano netbook with its via made x86 CPU, you are now buying an nvidia CPU. It is only a matter of time before nvidia comes to market with their own fusion.

-WTF Nvidia bought VIA?! Why the hell haven't I heard of this anywhere?! Nothing's coming though the google searches.

 

[taltamir]

interesting. I cant find it either... closes I have come is this: http://www.pcper.com/comments.php?nid=6960

I remember an article that said it outright bought it very recently, but maybe it was a hoax since I cannot verify it.

 

[taltamir]

NV and Intel aren't friends really. If anything will Intel just buy NV already?

intel has enough legal heat without buying out their competitors (monopoly is illegal you know)

 

[GodisanAtheist]

interesting. I cant find it either... closes I have come is this: http://www.pcper.com/comments.php?nid=6960

I remember an article that said it outright bought it very recently, but maybe it was a hoax since I cannot verify it.

-Yeah I found a bunch of articles, some even dating back to 2008 about Nvidia thinking of buying VIA. However, I don't know how much good it would do them as their x86 license with Intel is supposed to be VERY non-transferable, might be why the whole thing fell through.

 

[taltamir]

well, according to this the FTC is setting it up to be transferable
http://www.pcper.com/comments.php?nid=8143

 

[GodisanAtheist]

well, according to this the FTC is setting it up to be transferable
http://www.pcper.com/comments.php?nid=8143

-Yeah, Intel's had the x86 license for what, almost 25 years now? Its about time it got opened up a bit (best case scenario Nvidia could acquire one without having to pick up VIA's dead weight).