Support NeoGAF

[Hajaz]

Summer will bring a GPU war

Analysis PR teams gear up for carnage

By Charlie Demerjian: Friday, 02 May 2008, 5:05 PM

THIS SUMMER WE are going to see an large scale GPU war, something we haven't really seen for a few years now. Nvidia may have all the headlines today , but ATI has been plotting and rebuilding for the 7xx series launch.

Last year, the mere suggestion of ATI doing well was laughed at, but the firm took the outright lead with the X2 cards and forced NV into the reactionary GX2. ATI can do three and four-way adequately, albeit with the the Broken OS, while NV can only do it in name. Same with the bucket called 'hybrid' for both power and frame rate.

ATI has been plotting a comeback, and the R770/R700 parts should take the outright lead once again. The trick to the cards is what we told you almost two years ago, no more big GPUs.

You saw a little of that with the 3870X2, but the bridge was a simple PCIe switch. The real magic this time is a bridge that shares memory, GDDR5 in this case. Yup, you will have 2 GPUs with one set of memory.

This simplifies designs, lowers chip cost, and speeds time to market. You get two full variants for the design cost of 1.25, and you are on the happy end of the cost/area curve for fabbing silicon. While the early word on GT200 is that it is again 500mm^2+, ATI will have 2x chips that are much smaller, which translates into a huge cost advantage.

The other nice thing is that the bridge should keep the GPUs hidden from the system. This has a disadvantage of hard-wiring in the Crossfire modes leaving a little performance on the table, but when you have two of them in the system, it looks like two GPUs, not four. One look at the 1 -> 2 -> 4 scaling rates will show what a win that is.

What it comes down to in the end is that ATI looks to have built up a technological lead that NV is reacting to. It is the same thing that happened about the time when NV first released SLI, it had all the answers and ATI had to catch up. Now the tables are turned.

Another interesting change is on the PR side. NV has been mouthing off to anyone who won't run away fast enough while ATI has been silent. Some minor reshuffling at the ATI PR camp says that they are prepping for an outright hot war. As you know, it is always the silent ones that are the most trouble.

This summer we will see both trends come to a head, a new resurgent ATI technological lineup with a PR team willing to beat heads to get the job done. NV will be playing catchup with a long list of paper technologies and a few real ones.

The shouting should be pretty intense, and with Intel flipping sides, things will be all the more interesting.

yes, its the inq

 

[Frankfurter]

yes, its the inq

I'm no so much into this kind of stuff, but does that mean that this is just some kind of speculation?

 

[Juice]

Maybe I'm just old, but has ATI started re-using both code and product numbers to identify its products?

How does a model 7600 turn around to be considerably worse than a 4800?

Gruntmaster 6000, a slimmed down version of the Gruntmaster 9000, but just as fun!

 

[artist]

I know this post will get lost in this melee, but whatever.

The high-end PC graphics market doesn't matter very much - ATI having contracts for the 360 and Wii this gen will keep them going for a long, long, long time. What does matter is this interesting bit:

Nvidia has no CPU line.

Here's where we will be a few years down the road:

- Intel and AMD will have comparable integrated CPU/GPU solutions that offer substantial entry-level performance; mostly just an extension of where we are now, but with AMD truly integrating the entire solution into one chipset.
- Via will be in the process of owning them both in the embedded box and ultra-small PC market with its Pico-ITX and low power/price/performance CPUs.
- Nvidia will likely have a good GPU lineup, but any CPU designs they might be working on will have the same problems gaining traction in the market as Via's initial CPUs did a few years ago. Nvidia might even be shut out of chipsets for Intel and AMD at this point if they choose to design a CPU, just like Via suffered with Intel a few years ago.

Nvidia's in a hard spot right now, since all they can do without running into problems is keep the GPU lines cranking and keep doing chipsets. All of their competitors have CPU lines, which adds additional revenue. If anyone in this game will go bankrupt (lol right), it would be Nvidia.

Same reason why I think Nvidia was better off buying VIA over that shithole AEGIA.

Maybe I'm just old, but has ATI started re-using both code and product numbers to identify its products?

How does a model 7600 turn around to be considerably worse than a 4800?

Gruntmaster 6000, a slimmed down version of the Gruntmaster 9000, but just as fun!

Nvidia will face the same situation pretty soon. AMD started reusing the numbers but with a 'HD' differenciation.

I'm no so much into this kind of stuff, but does that mean that this is just some kind of speculation?

The rumors are suggesting the same thing, shared memory pool for AMD's next X2 card. However no one knows what kind of performance numbers that translates to in real world games.

 

[avaya]

I know this post will get lost in this melee, but whatever.

The high-end PC graphics market doesn't matter very much - ATI having contracts for the 360 and Wii this gen will keep them going for a long, long, long time. What does matter is this interesting bit:

Nvidia has no CPU line.

Here's where we will be a few years down the road:

- Intel and AMD will have comparable integrated CPU/GPU solutions that offer substantial entry-level performance; mostly just an extension of where we are now, but with AMD truly integrating the entire solution into one chipset.
- Via will be in the process of owning them both in the embedded box and ultra-small PC market with its Pico-ITX and low power/price/performance CPUs.
- Nvidia will likely have a good GPU lineup, but any CPU designs they might be working on will have the same problems gaining traction in the market as Via's initial CPUs did a few years ago. Nvidia might even be shut out of chipsets for Intel and AMD at this point if they choose to design a CPU, just like Via suffered with Intel a few years ago.

Nvidia's in a hard spot right now, since all they can do without running into problems is keep the GPU lines cranking and keep doing chipsets. All of their competitors have CPU lines, which adds additional revenue. If anyone in this game will go bankrupt (lol right), it would be Nvidia.

You missed the 800-pound gorilla. Mobile. Nokia, Samsung & SonyEricsson will want embedded solutions for the handsets of the future. Nvidia already has links with Sony.

People overstate the importance of integrated graphics chipsets. Enough of them will be redundant, paired with a discrete GPU each time that it will provide a sufficient market. Moreover the margin on them is miniscule. As long as Nvidia can prove that it is the leader in the discrete space then Nvidia will be fine. Nvidia won't be stupid enough to get involved with CPU design in the x86 market.

No one is going bankrupt. This isn't 3dfx small time anymore.

 

[artist]

You missed the 800-pound gorilla. Mobile. Nokia, Samsung & SonyEricsson will want embedded solutions for the handsets of the future. Nvidia already has links with Sony.

People overstate the importance of integrated graphics chipsets. Enough of them will be redundant, paired with a discrete GPU each time that it will provide a sufficient market. Moreover the margin on them is miniscule. As long as Nvidia can prove that it is the leader in the discrete space then Nvidia will be fine. Nvidia won't be stupid enough to get involved with CPU design in the x86 market.

No one is going bankrupt. This isn't 3dfx small time anymore.

Marketshare:
Nokia - 40%+
Sony Ericcsson - 9%+

It is difficult for a tech company to thrive on a single pronged approach, they either become niche or will have to diversify.

 

[avaya]

Marketshare:
Nokia - 40%+
Sony Ericcsson - 9%+

It is difficult for a tech company to thrive on a single pronged approach, they either become niche or will have to diversify.

Nokia's marketshare is built off it's low-end handsets, which are not the one's in question. When you go up the market the situation is fractured. SE and Nokia have close ties and move together on things like this. It's still a nascent market for higher end GPU's.

However my point was to contest this notion that Nvidia don't have a CPU product means they will be in serious trouble. That is a leap of logic that does not fit with what we have seen before. A technology company should stick to what it does best. Mobile and embedded soltuions provide enough basis for growth. Intel and AMD have no native advantage in that arena, their area of expertise is still solely in x86. Nvidia has ample opportunity to provide graphics solutions for high end handsets.

 

[SRG01]

I know this post will get lost in this melee, but whatever.

The high-end PC graphics market doesn't matter very much - ATI having contracts for the 360 and Wii this gen will keep them going for a long, long, long time. What does matter is this interesting bit:

Nvidia has no CPU line.

Here's where we will be a few years down the road:

- Intel and AMD will have comparable integrated CPU/GPU solutions that offer substantial entry-level performance; mostly just an extension of where we are now, but with AMD truly integrating the entire solution into one chipset.
- Via will be in the process of owning them both in the embedded box and ultra-small PC market with its Pico-ITX and low power/price/performance CPUs.
- Nvidia will likely have a good GPU lineup, but any CPU designs they might be working on will have the same problems gaining traction in the market as Via's initial CPUs did a few years ago. Nvidia might even be shut out of chipsets for Intel and AMD at this point if they choose to design a CPU, just like Via suffered with Intel a few years ago.

Nvidia's in a hard spot right now, since all they can do without running into problems is keep the GPU lines cranking and keep doing chipsets. All of their competitors have CPU lines, which adds additional revenue. If anyone in this game will go bankrupt (lol right), it would be Nvidia.

AMD's 730G(??)/Puma platform for notebooks will be extremely attractive to entry-level notebook or enterprise consumers. It's due to come out later this year. Larrabee, on the other hand, I am still skeptical about, since Intel does not have a good track record with their GPU lines despite their PR. On the other hand, Puma may be coming out too late for it to matter, as it is slated to come out later this year or early the next.

VIA isn't going up or down. It has no great technological advances to carry them into the next generation, even with their pico-ITX or their next-gen chip. The embedded market is quite literally owned by ARM and ultramobiles will most likely go to Intel's Atom, as many people do not have high hopes for their new chip.

As for nVidia, they do not need a CPU line to survive. It has been demonstrated year after year that they make a significant amount of money through chipsets and video hardware. As long as they make performance hardware, public perception of their brand will not change and people will still buy their product, much like how Intel commands a large portion of the market at the moment.

If anything, I expect next year to be more of the status quo: Intel is still on top, AMD is behind yet again with private companies/governments throwing money at them, and nVidia to outperform AMD/ATI significantly. Markets are not disrupted easily, unless there is a massive setback like with Intel several years ago when AMD took the lead.

 

[SRG01]

Marketshare:
Nokia - 40%+
Sony Ericcsson - 9%+

It is difficult for a tech company to thrive on a single pronged approach, they either become niche or will have to diversify.

Er, what? SE should have more than 9% of the global market...

 

[avaya]

Er, what? SE should have more than 9% of the global market...

No 9% is correct. SE's real share is mid to high end. They do not really compete in the low end where the major numbers are.

 

[camineet]

Did JSS say that they'll achieve Teraflops with a single card? For a SLI system yes. Check his wording again..

He didn't say wether it would be single card SLI via dual GPU (ala current GX2) or SLI with 2 cards. If GT200 is one TeraFlop, and I have no reason to think it's not, then 2 TFLOPs could be achived with single card, dual GPU SLI. If AMD ATI wanted to they could get 4 TFLOPs from using two, dual-GPU R700 cards

It contradicts your own theory. Do you honestly think a dual GPU board is possible where each gpu is more than 400mm2 ? Keep dreaming.

Well its quite clear which side of the fence you tip towards.:lol I'm saying nothing in regards to who will win the top crown.[/QUOTE]

What I am saying is, AMD/ATI and Nvidia will be very competitive this summer and this fall.

This kind of stuff should be entirely possible by Q1 2009.

 

[zoku88]

This might not be legit at all, but here are some unofficial benchmarks comparing the 512MB 4870 to the 1GB 9800 GX2..

http://www.bilgiustam.com/512mb-ati...orce-9800-gx2-crysis-ve-3dmark-2006-testleri/

 

[Bidermaier]

This might not be legit at all, but here are some unofficial benchmarks comparing the 512MB 4870 to the 1GB 9800 GX2..

http://www.bilgiustam.com/512mb-ati...orce-9800-gx2-crysis-ve-3dmark-2006-testleri/

I want to believe

 

[artist]

Nokia's marketshare is built off it's low-end handsets, which are not the one's in question. When you go up the market the situation is fractured. SE and Nokia have close ties and move together on things like this. It's still a nascent market for higher end GPU's.

http://press.nokia.com/PR/200605/1049032_5.html

However my point was to contest this notion that Nvidia don't have a CPU product means they will be in serious trouble. That is a leap of logic that does not fit with what we have seen before. A technology company should stick to what it does best. Mobile and embedded soltuions provide enough basis for growth. Intel and AMD have no native advantage in that arena, their area of expertise is still solely in x86. Nvidia has ample opportunity to provide graphics solutions for high end handsets.

What? AMD has larger design wins when it comes to mobile solutions. And we'll revisit this topic again when Intel also joins the party in a big way ..

Also the mobile market isnt nearly as significant as you are making it out to be. Both Nvidia and ATI (AMD) have been hyping it for the last 4-5 years and its all been hot air.

He didn't say wether it would be single card SLI via dual GPU (ala current GX2) or SLI with 2 cards. If GT200 is one TeraFlop, and I have no reason to think it's not, then 2 TFLOPs could be achived with single card, dual GPU SLI. If AMD ATI wanted to they could get 4 TFLOPs from using two, dual-GPU R700 cards

Which proves my point exactly. He didnt commit to anything specific. Even if he had specifically said that they would achieve 2 TFlops with a dual board single card, we all know it doesnt necessarily pan out. Didnt Nvidia repeatedly mention that they would have a 1TFlop back in 2007? :lol

What I am saying is, AMD/ATI and Nvidia will be very competitive this summer and this fall.
This kind of stuff should be entirely possible by Q1 2009.

Doesnt sound right when you dismiss one over the other months before the launch of either of them.

Those 4870 benchmarks dont look believable to me.

 

[iamblades]

AMD's 730G(??)/Puma platform for notebooks will be extremely attractive to entry-level notebook or enterprise consumers. It's due to come out later this year. Larrabee, on the other hand, I am still skeptical about, since Intel does not have a good track record with their GPU lines despite their PR. On the other hand, Puma may be coming out too late for it to matter, as it is slated to come out later this year or early the next.

VIA isn't going up or down. It has no great technological advances to carry them into the next generation, even with their pico-ITX or their next-gen chip. The embedded market is quite literally owned by ARM and ultramobiles will most likely go to Intel's Atom, as many people do not have high hopes for their new chip.

As for nVidia, they do not need a CPU line to survive. It has been demonstrated year after year that they make a significant amount of money through chipsets and video hardware. As long as they make performance hardware, public perception of their brand will not change and people will still buy their product, much like how Intel commands a large portion of the market at the moment.

If anything, I expect next year to be more of the status quo: Intel is still on top, AMD is behind yet again with private companies/governments throwing money at them, and nVidia to outperform AMD/ATI significantly. Markets are not disrupted easily, unless there is a massive setback like with Intel several years ago when AMD took the lead.

Intel's graphics designs are really the wildcard here. If they design everything well, they have the process advantage to slaughter everyone else. I don't see them making a P4/netburst/rambus type design clusterfuck anytime soon, but I could easily see the graphics side of their platform being subpar for the first couple generations, which could allow AMD time to catch up on the process front. Nvidia is a huge question mark for me atm. They have great designs right now, but they are competing against 2 larger companies that have the capability to design and produce an all in one CPU+chipset+GPU solution with all the benefits that entails(if done properly anyway). Can Nvidia design chipsets and GPUs that are good enough to justify not using the ones that were designed from the ground up to complement the CPU? It's possible they could pull it out, but you never know.

 

[godhandiscen]

I want to believe

OH SHIT!
Still not getting until X2 version.

 

[TaeOH]

I want to believe

That won't be enough to take on Nvidia's next chip coming in July. Once again ATI will be too little too late.

 

[artist]

That won't be enough to take on Nvidia's next chip coming in July. Once again ATI will be too little too late.

Typically that should be enough, next-gen GPU is twice as fast as previous-gen GPU.

However, those 4870 numbers are highly doubtful in the first case.

 

[godhandiscen]

That won't be enough to take on Nvidia's next chip coming in July. Once again ATI will be too little too late.

Are you crazy, those numbers are too fucking good to be true. Even at the price of $550 (current price of a decent GX2) it would be a steal.

That benchmark is bullshit though seeing as ATI is a canadian company and the blog appears to only be trying to get hits.

 

[Bidermaier]

That won't be enough to take on Nvidia's next chip coming in July. Once again ATI will be too little too late.

So what? I dont give a shit about Ati wining the pointless fastest card race. If the numbers are true i am happy to see Ati making a single chip that beats the fastest dual gpu card right now. If nvidia can go futher that is perfect. Competition drives tech development.

The only reason why the 8800GTX is not obsolete yet is because the 2900 was a bluff. We all lose when there is no competition.

 

[camineet]

If the small RV770 GPU is even competitive with 9800 GX2, I'll be pleased, but not surprised.

Remember even an 8800 Ultra beats the 9800 GX2 in some areas.

 

[avaya]

http://press.nokia.com/PR/200605/1049032_5.html

Close ties as in common spec roadmap. See transition from 3.5G to >4G.

What? AMD has larger design wins when it comes to mobile solutions. And we'll revisit this topic again when Intel also joins the party in a big way ..

Also the mobile market isnt nearly as significant as you are making it out to be. Both Nvidia and ATI (AMD) have been hyping it for the last 4-5 years and its all been hot air.

This is like saying the PC era will never come about in 1970. The mobile market will dwarf the current market in the years to come. Slowly but surely it's getting to the stage where real GPU solutions will be necessary to bring the differentiator between products. I don't think Intel will win big in that market. I don't think anyone will establish dominance.

 

[artist]

Close ties as in common spec roadmap. See transition from 3.5G to >4G.

Yes and undoubtedly Nokia is the bigger fish.

This is like saying the PC era will never come about in 1970. The mobile market will dwarf the current market in the years to come. Slowly but surely it's getting to the stage where real GPU solutions will be necessary to bring the differentiator between products. I don't think Intel will win big in that market. I don't think anyone will establish dominance.

What! we'll have Crysis on our mobile phones? :lol :lol :lol Seriously, in the next 5-10 years mobiles will start relying on ARM/Atom like cpus to drive them. You just got to look at the iPhone/clones and apply it to the mainstream market in the next 5 years ..

 

[TaeOH]

So what? I dont give a shit about Ati wining the pointless fastest card race. If the numbers are true i am happy to see Ati making a single chip that beats the fastest dual gpu card right now. If nvidia can go futher that is perfect. Competition drives tech development.

The only reason why the 8800GTX is not obsolete yet is because the 2900 was a bluff. We all lose when there is no competition.

I don't care either who has the best card. I went from a X1900 to an 8800GTX. The GTX is the best card I have ever bought though, still after a year and a half one of the top cards on the market.

The reason I said I did not think that would be enough is because of the rumors of Nvidia's next release being a true next gen chip and not a refresh. So it should match or exceed the numbers posted for the 4800. ATI needs another 9700 otherwise it will be hard for them to gain market share back from Nvidia. If the 4800 and the GT200 (or whatever Nvidia next chip is called) are on par with each other, I would likely stay with Nvidia because the stock coolers they use are quieter then anything I have ever gotten from ATI.

 

[proposition]

I don't care either who has the best card. I went from a X1900 to an 8800GTX. The GTX is the best card I have ever bought though, still after a year and a half one of the top cards on the market.

The reason I said I did not think that would be enough is because of the rumors of Nvidia's next release being a true next gen chip and not a refresh. So it should match or exceed the numbers posted for the 4800. ATI needs another 9700 otherwise it will be hard for them to gain market share back from Nvidia. If the 4800 and the GT200 (or whatever Nvidia next chip is called) are on par with each other, I would likely stay with Nvidia because the stock coolers they use are quieter then anything I have ever gotten from ATI.

Don't forget, the 9800 GX2 is two 9800 chips on a single card, not one.

 

[TaeOH]

Don't forget, the 9800 GX2 is two 9800 chips on a single card, not one.

But it is only approximately 20-30% faster than an 8800GTX Ultra in the most generous benchmarks, and not in all games since it depends upon SLI, which we all know has to be coded for.

Hopefully both ATI and Nvidia WILL ship single chip solutions this year with better than GX2 performance.

 

[artist]

ATI needs another 9700 otherwise it will be hard for them to gain market share back from Nvidia.

Not necessarily. RV670 won them quite a chunk of market share (back) and it wasnt the top card. Another example is the FX5200 from Nvidia, their saviour during the NV30 fiasco.

Having the top dog definitely has advantages; halo effect, better stock trading, dev support etc. The last two depend on other factors like die size, PR team etc.

 

[artist]

Not sure if this was mentioned already:

http://messages.finance.yahoo.com/Stocks_(A_to_Z)/Stocks_N/threadview?m=tm&bn=13029&tid=435392&mid=435402&tof=7&rt=2&frt=2&off=1

240 cores... 10 clusters x 3x8 SPs/cluster? I'm not exactly impressed to say the least unless each SP is more powerful than G80's though, or overall chip efficiency was somehow improved. The reason why I bother posting this is that it's an excerpt from Barron's, which apparently met with Jen-Hsun last week.

That seems less than twice as much as G92/G80. Still it should easily top the 9800GX2.

 

[Ryoma-Echizen]

Let the massive ownage begin.

http://www.dailytech.com/article.aspx?newsid=12185

http://www.hardwarecanucks.com/foru...eon-hd4870-512mb-graphics-card-review-16.html
http://www.hothardware.com/articles/ATI_Radeon_HD_4850_and_4870_RV770_Has_Arrived/
http://www.elitebastards.com/cms/index.php?option=com_content&task=view&id=584&Itemid=31


GTX260 is dead and the GTX280 in S.O.S

 

[Vaporak]

Nice bump. :lol But I think were all pleasantly surprised to see ATI/AMD back in the game.

 

[godhandiscen]

Let the massive ownage begin.

http://www.dailytech.com/article.aspx?newsid=12185

http://www.hardwarecanucks.com/foru...eon-hd4870-512mb-graphics-card-review-16.html
http://www.hothardware.com/articles/ATI_Radeon_HD_4850_and_4870_RV770_Has_Arrived/
http://www.elitebastards.com/cms/index.php?option=com_content&task=view&id=584&Itemid=31


GTX260 is dead and the GTX280 in S.O.S

One more time, we are going to celebrate....

Anandtech review http://www.anandtech.com/video/showdoc.aspx?i=3341&p=1

 

[zoku88]

Why did you bump this thread? We already have a suitable GPU topic XD

Spoiler

and the topic title is wrong, though.)

 

[godhandiscen]

Why did you bump this thread? We already have a suitable GPU topic XD

Spoiler

and the topic title is wrong, though.)

Spoiler

Yeah it should be RV700. Also, I didn't bump, it. I found it on first page.

 

[zoku88]

Yeah it should be RV700. Also, I didn't bump, it. I found it on first page.

Lol, I didn't mean you >.>

(and technically, the 280 is the performance king. It just has a sucky price.) >.>

 

[godhandiscen]

Lol, I didn't mean you >.>

(and technically, the 280 is the performance king. It just has a sucky price.) >.>

Lamest king ever then. When for $400 you can Cross Fire 2 4850s and beat the "king" then you know there is no kingdom. Also, the 4870X2 will pwn it from here to the stars. Just looking at the 4870CF benches you can already tell how badly Nvidia f'd up this gen.

Also, $570 4870X2 BELIEVE!

 

[zoku88]

Lamest king ever then. When for $400 you can Cross Fire 2 4850s and beat the "king" then you know there is no kingdom. Also, the 4870X2 will pwn it from here to the stars. Just looking at the 4870CF benches you can already tell how badly Nvidia f'd up this gen.

Also, $570 4870X2 BELIEVE!

but thats no longer a single gpu

 

[godhandiscen]

but thats no longer a single gpu

We could go forever. The reality is that the GTX280 isnt the performance king because there are superior alternatives at cheaper prices. Now, the only benchmark at which the GTX280 could excell at is Tri-SLI, and that will run around $1800 on the cards alone. Even then 2*4870X2 in CF come awfully close for $700 less. This is a bad generation for Nvidia, period.

 

[godhandiscen]

dp

 

[zoku88]

We could go forever. The reality is that the GTX280 isnt the performance king because there are superior alternatives at cheaper prices. Now, the only benchmark at which the GTX280 could excell at is Tri-SLI, and that will run around $1800 on the cards alone. Even then 2*4870X2 in CF come awfully close for $700 less. This is a bad generation for Nvidia, period.

That's not performance, that's price/performance. Disregarding price, since we're only talking about performance, an nvidia configuration has the top performance.

And note, I wasn't even talking about a configuration. I was talking GPU (not even card, GPU.)

 

[godhandiscen]

That's not performance, that's price/performance. Disregarding price, since we're only talking about performance, an nvidia configuration has the top performance.

And note, I wasn't even talking about a configuration. I was talking GPU (not even card, GPU.)

Ok, sorry then Nvidia is the performance king when it comes to a single GPU. I wonder if they are happy though. I'll just keep listening to Daft Punk.

 

[bathala]

well since this thread of 4850 was bumped.

Any impression from owners especially running stock fan?
Howz Crysis?

 

[zoku88]

well since this thread of 4850 was bumped.

Any impression from owners especially running stock fan?
Howz Crysis?

There are impressions in the other thread. We should probably just stop writing in this one, since it's pretty useless now...

 

[camineet]

R700 Inter-GPU Connection Discussion



R700 ~ 4870X2 board pics?

http://img.hexus.net/v2/graphics_cards/ati/4850X/R7001-big.jpg

http://img440.imageshack.us/img440/7695/2008062821e8649056b17c3ba0.jpg

http://farm4.static.flickr.com/3121/2617131693_5ddd77d49b_b.jpg


CNET Article

For ATI, the execution of this chip-ganging strategy is the key. And this is where ATI appears to have been successful. "The inter-processor communications. Getting that to work has been the trick. This is what ATI has done. They've come up with this stellar way of doing inter-processor communications so they can in fact get the scaling," according to Peddie.

AMD-ATI's upcoming R700 (rumored to be called the 4870 X2) two-chip graphics board will be the ultimate test of this strategy.

"It's a new proprietary inter-processor communication technology. If they put these two chips on one board and it does scale properly, then they have pulled off a coup," he said.

"When you gang up graphics chips (using the traditional Scalable Link Interface or CrossFire technologies) they roll off pretty fast. ("Roll off" implies that performance doesn't scale up well.) "So when you put two boards in, you don't get twice the performance but you (only) get one and a half. You put four boards in and you (only) get about 1.7, 1.8. What ATI is saying is that with two chips using (their) proprietary inter-bus, they will get 1.8 (the performance) with two chips. If that's true, you can expect to see four of them giving you something around 2.5."

Getting 2.5 times the performance from four boards would be a masterstroke for ATI.

The previous ATI dual-chip solution was very different, Peddie said. "The HD 3870 X2 was not a proprietary bus but a CrossFire connection. The CrossFire connection and the SLI connection are at the very, very end of the pipeline. Not the most efficient place to do an inter-processor communication. That's one of the reasons ATI has abandoned it."

So it's still not 1:1 scaling, but apparently significantly better than CrossFire or SLI.

 

[Crovax33]

In the forum discussion it says they reworked CrossFire to allow 1.8 scalability. Did I understand this correctly?

 

[Diablos]

^ What a stupid thing to say. nVidia's the first choice for "hardcore gamers" *currently* because they have the best parts out now. The end.

Exactly. Remember when everyone went crazy over teh RADEON 9700 AND 9800 PRO OMG.

It IS nice to see AMD/ATI catching up, though.

 

[zoku88]

Why are we bumping this thread again when we already have a more recent thread.... :-/

 

[camineet]

1.) It seems the new inter-connect chip between the two RV770 GPUs on the 4870 X2 board will allow 1.8x the performance of a single RV770, without any form of CrossFire whatsoever.

2.) Seperately, it seems they're also improving CrossFire as well, since people will still want to CrossFire two 4870 X2 boards, as well as single-GPU 4850 and 4870 cards.

3.) Still need to do more reading, don't understand what they've done to CrossFire yet.

 

[Crovax33]

Why are we bumping this thread again when we already have a more recent thread.... :-/

Is this info in the other thread? I just want to upgrade soon, so I am discussing about parts wherever I can.