ATi 4870 X2 (R700) thread

[ShadowOfMyself]

I believe it... People said the same about the 48xx series pricing... "If AMD is pricing them that low, it means they will not deliver", well, you got to eat your own words... So whats surprising about yet another bargain product from AMD?

 

[AmberClad]

Fud is confirming that it will be 2GB. Codename is "Spartan" (someone at ATI is apparently a fan of 300).

 

[taltamir]

Originally posted by: chizow

Originally posted by: toslat
On being CPU limited:
A simple test using two quads at two different clocks (all else same) should put this to rest

I've been linking them throughout, but its *very* obvious in any bench that uses 4GHz and some of the exotic configs out there, like Tri-SLI or CrossFireX. Anyways, here's two really good examples from Tweaktown:

4870 Crossfire @ 3GHz

4870 Crossfire @ 4GHz

Summary: Up to 1920, there is very little difference in performance between 4870CF and 4850CF and much less scaling compared to a single 4870 to 4870CF vs 4850 to 4850CF. Scaling at 2560 is less consistent, but you still see big gains meaning you are not completely GPU bottlenecked. Tweaktown also has 4850CF @ 3GHz and 4GHz and GTX 280 in SLI/Tri-SLI that echoes similar results. What you will notice is that the GTX 280 also scales very well as a single-GPU depending on CPU speed, meaning it is also CPU bottlenecked up to 1920. I'd expect similar from a faster 4870 variant up until it capped out about the same as the GTX 280.

On PLX 'bridge' and shared PCIe:
The PLX chips on x870X2 do not split lanes i.e. 16x switched ~= 8x:8x. It is wrong to use 8x:8x performance of 4850 CF to predict performance of 4870X2

IIRC the 3870X2 has a 48 lane 3 port v1.1 PEX8547 switch (not a bridge) and thus each GPU gets full 16x access when it is granted and not 8x. The advantage of this is that since the PCIe connection to the north bridge carries bursty traffic (actual traffic pattern depends on the game), the impact of latency is reduced and bandwidth preserved, as opposed to the bandwidth bottlenecking that would occur with a fixed 8x:8x split.

In the 4870X2, the new switch (possibly PEX8648) will support v2.0 which will double the available data rate (though with slight latency increase).

Also its is rumored that inter GPU traffic will improve. Not sure if inter GPU traffic went through the NB (and memory) in the 3870X2 since the switch functionally could route traffic between the GPUs. So I expect, at the least, that inter GPU traffic is switched at the PEX chip, and if possible they might have a common/duplicate memory area.

So there is no basis yet to state that the 4870X2 will be bottlenecked by the 'bridge' unless you can provide stats that show that the traffic pattern on the bus is sustained at >50% for a 4870 (single or CF),and even then show that ( 2x individual data - common data) exceeds 100% of 16x in CF

That does look to be true about the PLX switch, but that inherently assumes a normal PCIE card will be using the bus at less than 50% efficiency plus you add latency into the equation over a straight split. Here's a pretty good write-up on the 3870 and the switch, with diagram (about 1/2 way down) Digital-Daily 3870X2 Bridge Chip. It actually seems like there's more overhead with a switch rather than a splitter regardless of available bandwidth.

that is not "very little difference" on the CPU Speed... going from 3 to 4ghz (33% OC CPU)
1280x1024: from 120 to 150 fps on 1280x1024 resolution... 25%
1920x1200: 112 to 165 = 47% increase in FPS
2560x1600: 97 to 120 = 23% increase in FPS

It seems to me like the CPU is a major limit there if increasing it by 33% increases FPS by 23 to 47% depending on resolution. (with max gains being at 1920x1200, yum!)

 

[sxr7171]

Originally posted by: Astrallite
SLi performance is a bit misleading. While the average framerate ends up looking great, the issue is the minimum framerate doesn't tend to go up that much. You get maybe a 20-30fps average fps gain, but the minimum fps goes up from 18 to 21.

Then you have the problem where occasionally SLI will have worse minimum framerates due to latency between cards, but the average still looks much higher. The game will be choppier but the numbers look better.

Now this usually is circumvented in x2 cards but not always.

Yeah that's the $64,000 question. I'll wait for performance numbers. Isn't ATI working on a technology to reduce latencies compared to older generation x2 cards?

 

[sxr7171]

Originally posted by: SteelSix

Originally posted by: taltamir
mmm... we kept on hearing it WILL be 1GB but shared across both GPUs.... supposedly THAT will eliminate microstutter somehow...

But now it says 2x1GB?

A 2GB DDR5 board for under $500?? It would be a major trump card if true. I just can't believe it though. Need proof. Damn I'm so ready to buy this thing if that's true..

In what situations are we graphics memory limited? At least I hope you have a 30" 2560x1600 display.

 

[sxr7171]

Originally posted by: dasracht
New confirmation on the 2x1gb ram amount!

Digitimes: AMD Radeon HD 4870 X2 ready by August

From the article:

Evaluation samples are schedule to be available in mid-July, and AMD will begin shipping reference design boards with 2GB GDDR5 memory at the end of the month, the sources detailed.

HD 4870 X2 graphics cards are expected to be priced around US$499, the sources detailed.

2x1GB is bad news. 1x2GB is good news.

 

[Compddd]

When will we know if the memory pool is shared or not on the X2?

 

[sxr7171]

Originally posted by: taltamir

Originally posted by: chizow

Originally posted by: toslat
On being CPU limited:
A simple test using two quads at two different clocks (all else same) should put this to rest

I've been linking them throughout, but its *very* obvious in any bench that uses 4GHz and some of the exotic configs out there, like Tri-SLI or CrossFireX. Anyways, here's two really good examples from Tweaktown:

4870 Crossfire @ 3GHz

4870 Crossfire @ 4GHz

Summary: Up to 1920, there is very little difference in performance between 4870CF and 4850CF and much less scaling compared to a single 4870 to 4870CF vs 4850 to 4850CF. Scaling at 2560 is less consistent, but you still see big gains meaning you are not completely GPU bottlenecked. Tweaktown also has 4850CF @ 3GHz and 4GHz and GTX 280 in SLI/Tri-SLI that echoes similar results. What you will notice is that the GTX 280 also scales very well as a single-GPU depending on CPU speed, meaning it is also CPU bottlenecked up to 1920. I'd expect similar from a faster 4870 variant up until it capped out about the same as the GTX 280.

On PLX 'bridge' and shared PCIe:
The PLX chips on x870X2 do not split lanes i.e. 16x switched ~= 8x:8x. It is wrong to use 8x:8x performance of 4850 CF to predict performance of 4870X2

IIRC the 3870X2 has a 48 lane 3 port v1.1 PEX8547 switch (not a bridge) and thus each GPU gets full 16x access when it is granted and not 8x. The advantage of this is that since the PCIe connection to the north bridge carries bursty traffic (actual traffic pattern depends on the game), the impact of latency is reduced and bandwidth preserved, as opposed to the bandwidth bottlenecking that would occur with a fixed 8x:8x split.

In the 4870X2, the new switch (possibly PEX8648) will support v2.0 which will double the available data rate (though with slight latency increase).

Also its is rumored that inter GPU traffic will improve. Not sure if inter GPU traffic went through the NB (and memory) in the 3870X2 since the switch functionally could route traffic between the GPUs. So I expect, at the least, that inter GPU traffic is switched at the PEX chip, and if possible they might have a common/duplicate memory area.

So there is no basis yet to state that the 4870X2 will be bottlenecked by the 'bridge' unless you can provide stats that show that the traffic pattern on the bus is sustained at >50% for a 4870 (single or CF),and even then show that ( 2x individual data - common data) exceeds 100% of 16x in CF

That does look to be true about the PLX switch, but that inherently assumes a normal PCIE card will be using the bus at less than 50% efficiency plus you add latency into the equation over a straight split. Here's a pretty good write-up on the 3870 and the switch, with diagram (about 1/2 way down) Digital-Daily 3870X2 Bridge Chip. It actually seems like there's more overhead with a switch rather than a splitter regardless of available bandwidth.

that is not "very little difference" on the CPU Speed... going from 3 to 4ghz (33% OC CPU)
1280x1024: from 120 to 150 fps on 1280x1024 resolution... 25%
1920x1200: 112 to 165 = 47% increase in FPS
2560x1600: 97 to 120 = 23% increase in FPS

It seems to me like the CPU is a major limit there if increasing it by 33% increases FPS by 23 to 47% depending on resolution. (with max gains being at 1920x1200, yum!)

If your game is at 97FPS does it matter if you can take it to 120fps? Your LCD refreshes at 60Hz. Do the comparison on a game where you go from non-playable 20fps to a playable 28fps. Now that would be CPU limiting worth a damn (if present and I doubt that).

 

[taltamir]

that is the game HE CHOSE TO LINK TO, so he is arguing that the CPU doesn't matter and links directly to a test of UT3 that is clearly CPU limited...

You want a low FPS game? World in Conflict... (just picked it because i know it is intensive, i didn't check every single test in that article, some might fare better, some worse)
interesting you said 22 to 28 is significant...
http://www.tweaktown.com/revie...n_crossfire/index.html
http://www.tweaktown.com/artic...rex_at_4ghz/index.html

Min FPS change:
1280x1024: 24 to 42
1920x1200: 23 to 33
2560x1600: 23 to 28
Almost the 22 to 28 example you gave, but not quite...

Average FPS:
1280x1024: 42 to 76 = 81%
1920x1200: 41 to 77 = 88%
2560x1600: 40 to 64 = 60%

Anyways, 60 to 88% increase in average FPS due to a 33% CPU speed increase. WIC in definitely CPU limited there.

 

[Sentry2]

2x1GB is bad news. 1x2GB is good news.

It's still better than 2x512MB....

 

[chizow]

Originally posted by: taltamir

Originally posted by: chizow

Originally posted by: toslat
On being CPU limited:
A simple test using two quads at two different clocks (all else same) should put this to rest

I've been linking them throughout, but its *very* obvious in any bench that uses 4GHz and some of the exotic configs out there, like Tri-SLI or CrossFireX. Anyways, here's two really good examples from Tweaktown:

4870 Crossfire @ 3GHz

4870 Crossfire @ 4GHz

Summary: Up to 1920, there is very little difference in performance between 4870CF and 4850CF and much less scaling compared to a single 4870 to 4870CF vs 4850 to 4850CF. Scaling at 2560 is less consistent, but you still see big gains meaning you are not completely GPU bottlenecked. Tweaktown also has 4850CF @ 3GHz and 4GHz and GTX 280 in SLI/Tri-SLI that echoes similar results. What you will notice is that the GTX 280 also scales very well as a single-GPU depending on CPU speed, meaning it is also CPU bottlenecked up to 1920. I'd expect similar from a faster 4870 variant up until it capped out about the same as the GTX 280.

On PLX 'bridge' and shared PCIe:
The PLX chips on x870X2 do not split lanes i.e. 16x switched ~= 8x:8x. It is wrong to use 8x:8x performance of 4850 CF to predict performance of 4870X2

IIRC the 3870X2 has a 48 lane 3 port v1.1 PEX8547 switch (not a bridge) and thus each GPU gets full 16x access when it is granted and not 8x. The advantage of this is that since the PCIe connection to the north bridge carries bursty traffic (actual traffic pattern depends on the game), the impact of latency is reduced and bandwidth preserved, as opposed to the bandwidth bottlenecking that would occur with a fixed 8x:8x split.

In the 4870X2, the new switch (possibly PEX8648) will support v2.0 which will double the available data rate (though with slight latency increase).

Also its is rumored that inter GPU traffic will improve. Not sure if inter GPU traffic went through the NB (and memory) in the 3870X2 since the switch functionally could route traffic between the GPUs. So I expect, at the least, that inter GPU traffic is switched at the PEX chip, and if possible they might have a common/duplicate memory area.

So there is no basis yet to state that the 4870X2 will be bottlenecked by the 'bridge' unless you can provide stats that show that the traffic pattern on the bus is sustained at >50% for a 4870 (single or CF),and even then show that ( 2x individual data - common data) exceeds 100% of 16x in CF

That does look to be true about the PLX switch, but that inherently assumes a normal PCIE card will be using the bus at less than 50% efficiency plus you add latency into the equation over a straight split. Here's a pretty good write-up on the 3870 and the switch, with diagram (about 1/2 way down) Digital-Daily 3870X2 Bridge Chip. It actually seems like there's more overhead with a switch rather than a splitter regardless of available bandwidth.

that is not "very little difference" on the CPU Speed... going from 3 to 4ghz (33% OC CPU)
1280x1024: from 120 to 150 fps on 1280x1024 resolution... 25%
1920x1200: 112 to 165 = 47% increase in FPS
2560x1600: 97 to 120 = 23% increase in FPS

It seems to me like the CPU is a major limit there if increasing it by 33% increases FPS by 23 to 47% depending on resolution. (with max gains being at 1920x1200, yum!)

I guess I should've clarified, that should read there is very little difference on a 3GHz CPU, which is what I've been saying this entire time and the point of linking 3GHz vs 4GHz benches, that the CPU you need to see differences in the high-end parts available today most likely has not been released yet.

 

[chizow]

Originally posted by: Compddd
I really hope this thing blows everyone and everything away, just so Chizow will hush up.

And if it doesn't, will you just crawl back under your bridge until ATI rises again? Nothing wrong with hoping I suppose, but based on 4870CF and GTX 280 SLI I don't think we'll see massive gains with the 4870X2 over a single 4870 or even 4850CF.

 

[toslat]

Originally posted by: chizow

Originally posted by: toslat
On being CPU limited:
A simple test using two quads at two different clocks (all else same) should put this to rest

I've been linking them throughout, but its *very* obvious in any bench that uses 4GHz and some of the exotic configs out there, like Tri-SLI or CrossFireX. Anyways, here's two really good examples from Tweaktown:

4870 Crossfire @ 3GHz

4870 Crossfire @ 4GHz

Summary: Up to 1920, there is very little difference in performance between 4870CF and 4850CF and much less scaling compared to a single 4870 to 4870CF vs 4850 to 4850CF. Scaling at 2560 is less consistent, but you still see big gains meaning you are not completely GPU bottlenecked. Tweaktown also has 4850CF @ 3GHz and 4GHz and GTX 280 in SLI/Tri-SLI that echoes similar results. What you will notice is that the GTX 280 also scales very well as a single-GPU depending on CPU speed, meaning it is also CPU bottlenecked up to 1920. I'd expect similar from a faster 4870 variant up until it capped out about the same as the GTX 280.

As much as you can use those results to argue that you are not GPU limited, they really dont say that you are CPU limited either, as those were two different systems (CPU, memory, mobo etc) used in the tests. Something like QX9xxx at two different multipliers was what I had in mind.

On PLX 'bridge' and shared PCIe:
The PLX chips on x870X2 do not split lanes i.e. 16x switched ~= 8x:8x. It is wrong to use 8x:8x performance of 4850 CF to predict performance of 4870X2

That does look to be true about the PLX switch, but that inherently assumes a normal PCIE card will be using the bus at less than 50% efficiency plus you add latency into the equation over a straight split. Here's a pretty good write-up on the 3870 and the switch, with diagram (about 1/2 way down) Digital-Daily 3870X2 Bridge Chip. It actually seems like there's more overhead with a switch rather than a splitter regardless of available bandwidth.

The latency involved for the PEX 8547 and PEX 85648 are 110ns and 140ns max respectively.

I am not assuming the efficiency of the PCIe bus is less than 50% (though quite likely). What am saying is that the argument of the x16 being a bottle neck, or not, depends heavily on the traffic pattern on the PCIe.

The PCIe is used by the video card to communicate with the CPU, memory, another GPU, etc. It is used to get new data,or control, that are not available in the onboard memory. How often the GPU uses the bus depend on the GPU in question, onboard memory, and the game. Some games will give the GPU large chunks of data occasionally and feed it small chunks in between while some are more even in their use. This traffic pattern argument actually goes down to the PCIe v2.0 x v1.1 argument, and given that you can see improvements in going to v2.0 for high end cards, I dont think that most games would have a sustained >50% usage. BTW at 50% usage on v2.0 16x, the GPU will be moving about 4GB/s which is like a overwriting an entire 512MB memory 8 times every second.

The switch vs splitter argument also depends on traffic pattern. In v2.0, 8x will transfer 4GB in 1s and 5GB in 2s, while a 16x will transfer both in 1s i.e 8x will result in doubled latency for the 5GB case. The 8x guarantees the availability of bandwidth but usually doesn't do you much good if both GPU have a common target which can only process one request at a time. Hence in such single target environments, switching is usually more advantageous since one of the GPUs would have to wait so why not let the active GPU maximize resource use. Ofcourse the latency overhead of switching cannot be ignored but it is a fixed overhead similar to what could be introduced by any other chip on the data path. If your traffic is low speed but more or less constant rate then you should use a splitter, while a switch will benefit you more if its bursty.

In summary, am not saying that the 16x will not be a bottleneck, but that you cant reach that conclusion based on 8x:8x performance as they are two different scenarios.

 

[sxr7171]

Originally posted by: Sentry2

2x1GB is bad news. 1x2GB is good news.

It's still better than 2x512MB....

Yeah but honestly what's the point of 2 GPUs on one card if they can't address Minimum framerates and microstuttering?

 

[sxr7171]

Originally posted by: taltamir
that is the game HE CHOSE TO LINK TO, so he is arguing that the CPU doesn't matter and links directly to a test of UT3 that is clearly CPU limited...

You want a low FPS game? World in Conflict... (just picked it because i know it is intensive, i didn't check every single test in that article, some might fare better, some worse)
interesting you said 22 to 28 is significant...
http://www.tweaktown.com/revie...n_crossfire/index.html
http://www.tweaktown.com/artic...rex_at_4ghz/index.html

Min FPS change:
1280x1024: 24 to 42
1920x1200: 23 to 33
2560x1600: 23 to 28
Almost the 22 to 28 example you gave, but not quite...

Average FPS:
1280x1024: 42 to 76 = 81%
1920x1200: 41 to 77 = 88%
2560x1600: 40 to 64 = 60%

Anyways, 60 to 88% increase in average FPS due to a 33% CPU speed increase. WIC in definitely CPU limited there.

Okay, definitely CPU limited there. Can't wait for Nehalem.

 

[chizow]

Originally posted by: toslat
As much as you can use those results to argue that you are not GPU limited, they really dont say that you are CPU limited either, as those were two different systems (CPU, memory, mobo etc) used in the tests. Something like QX9xxx at two different multipliers was what I had in mind.

That's only true when drawing comparisons at 4GHz. The 3GHz results are on the same platform and show virtually no difference in performance between 4850CF and 4870CF and also very similar to older SLI solutions (9800GX2). That's textbook CPU bottlenecking. I know its a bit of a shock to people since it was assumed a 3GHz C2D was good enough, but I think this last round shows its clearly not. This is just one review site btw, any site you see comparing exotic 2-3 way SLI/CF or using a slow CPU will show almost no scaling.

In summary, am not saying that the 16x will not be a bottleneck, but that you cant reach that conclusion based on 8x:8x performance as they are two different scenarios.

Fair enough, but I think its also obvious that you can't get around the fact you'll be trying to pass ~2x data over a single bus with an X2 compared to dual-slot CF. The Tom's guide does clearly show PCI 1.1 x8 isn't enough with last gen parts, but there's no noticeable difference between PCI 1.1 x16 and PCI 2.0 x16. We'll see if that holds true on current parts.

 

[MyLeftNut]

Originally posted by: sxr7171

Originally posted by: taltamir
that is the game HE CHOSE TO LINK TO, so he is arguing that the CPU doesn't matter and links directly to a test of UT3 that is clearly CPU limited...

You want a low FPS game? World in Conflict... (just picked it because i know it is intensive, i didn't check every single test in that article, some might fare better, some worse)
interesting you said 22 to 28 is significant...
http://www.tweaktown.com/revie...n_crossfire/index.html
http://www.tweaktown.com/artic...rex_at_4ghz/index.html

Min FPS change:
1280x1024: 24 to 42
1920x1200: 23 to 33
2560x1600: 23 to 28
Almost the 22 to 28 example you gave, but not quite...

Average FPS:
1280x1024: 42 to 76 = 81%
1920x1200: 41 to 77 = 88%
2560x1600: 40 to 64 = 60%

Anyways, 60 to 88% increase in average FPS due to a 33% CPU speed increase. WIC in definitely CPU limited there.

Okay, definitely CPU limited there. Can't wait for Nehalem.

Does the amount of RAM used on each of those test systems affect the performance in games like WiC? The 3Ghz system only had 2gb, which apparently they ran tests in both XP and Vista. Whereas, the 4Ghz system had 4gb using Vista. Also, how do we know which tests were run in XP and which in Vista on the 2gb system?

 

[toslat]

Originally posted by: chizow
That's only true when drawing comparisons at 4GHz. The 3GHz results are on the same platform and show virtually no difference in performance between 4850CF and 4870CF and also very similar to older SLI solutions (9800GX2). That's textbook CPU bottlenecking. I know its a bit of a shock to people since it was assumed a 3GHz C2D was good enough, but I think this last round shows its clearly not. This is just one review site btw, any site you see comparing exotic 2-3 way SLI/CF or using a slow CPU will show almost no scaling.

You make it seem like once you are not GPU limited then you must be CPU limited, as if those were the only two factors to consider.

Agreed the improved scores point to not being GPU limited, but to show that its CPU limited I would expect that you change only the CPU and keep all others, including FSB, memory, Motherboard and hard drive, constant.

In the tests you linked to
FSB: 333MHz vs 400MHz
Memory: 2 x 1GB 800MHz DDR2 vs 2x2GB 1666MHz DDR3
HDD: 250GB 7200RPM vs 150GB 1000RPM

Obviously the 4GHz system is superior in more places than just CPU, and thus its not conclusive evidence of CPU bottlenecking.

Again, am not arguing otherwise, just saying that the data provided do not support the argument strongly enough.

 

[nib95]

2GB! Damn. Considering holding off on my 4870 purchase now lol.

 

[Hauk]

nm

 

[taltamir]

Originally posted by: MyLeftNut

Originally posted by: sxr7171

Originally posted by: taltamir
that is the game HE CHOSE TO LINK TO, so he is arguing that the CPU doesn't matter and links directly to a test of UT3 that is clearly CPU limited...

You want a low FPS game? World in Conflict... (just picked it because i know it is intensive, i didn't check every single test in that article, some might fare better, some worse)
interesting you said 22 to 28 is significant...
http://www.tweaktown.com/revie...n_crossfire/index.html
http://www.tweaktown.com/artic...rex_at_4ghz/index.html

Min FPS change:
1280x1024: 24 to 42
1920x1200: 23 to 33
2560x1600: 23 to 28
Almost the 22 to 28 example you gave, but not quite...

Average FPS:
1280x1024: 42 to 76 = 81%
1920x1200: 41 to 77 = 88%
2560x1600: 40 to 64 = 60%

Anyways, 60 to 88% increase in average FPS due to a 33% CPU speed increase. WIC in definitely CPU limited there.

Okay, definitely CPU limited there. Can't wait for Nehalem.

Does the amount of RAM used on each of those test systems affect the performance in games like WiC? The 3Ghz system only had 2gb, which apparently they ran tests in both XP and Vista. Whereas, the 4Ghz system had 4gb using Vista. Also, how do we know which tests were run in XP and which in Vista on the 2gb system?

I missed that.. I don't consider ANY test ran with 2GB of ram valid today. 4GB is the absolute minimum for me to even consider a test as valid. I have been using 4GB since company of heroes... I think that was september of 06. it had a bar showing that it needs more ram then I have at max settings. When tested, I got 1-2 fps on max everything on my X2 3800+ with a 7900GS and 2GB of DDR2-800 ram, and 30-40fps with 4GB of ram.
I immidiatly bought 2 more gigs of ram and never looked back.. I think it was supreme command took 2.1GB of ram BY ITSELF on my vista 64bit machine. (+1.3GB for OS and background stuff... 3.4GB total).

That also would explain an 88% FPS increase on a 33% OC. Maybe the OC only accounts for 10% and the other 78% come from the extra ram.

 

[JPB]

R700 Spartan has 2GB of memory

Comes in late August

We just got a final confirmation that R700 uses two RV770XT GPUs and that it supports massive 2GB of GDDR5 memory.

Naturally, the card uses 256 bit memory interface and it looks that the card can ship in last days of August.

ATI used the codename Spartan for R700, as someone got hot for the movie. At this time, there are no talks about any clock speeds but the brand should settle at Radeon HD 4870 X2 area.

We will confirm this as soon as we can.

 

[Compddd]

Originally posted by: chizow

Originally posted by: Compddd
I really hope this thing blows everyone and everything away, just so Chizow will hush up.

And if it doesn't, will you just crawl back under your bridge until ATI rises again? Nothing wrong with hoping I suppose, but based on 4870CF and GTX 280 SLI I don't think we'll see massive gains with the 4870X2 over a single 4870 or even 4850CF.

Why don't you take your own advice then and crawl back under your bridge until nVidia rises again sir?

 

[nonameo]

You know, I wonder if they are delaying so that they can have extra time to beef up crossfire some more.

 

[dasracht]

Theres a high chance its because of gddr5 shortages. They ran into shortages with the 4870, and it uses 1/4 of the ram the x2 will use.

Late august is quite the delay. I think the expectations were late July at the earliest, early August at the latest.