Advanced Micro Devices - Moderated (AMD) Message Board

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD)

AMD 222.17

+0.5%

2:01 PM EST

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next

To: pgerassi who wrote (255265)	8/5/2008 7:25:59 PM
From: eracer	Read Replies (3) of 275872

Re: That graph that was presented to show nearly linear scaling between 8 and 32 means that the performance at 8 cores must be very bad. No, the 780G chipset getting 30FPS at 800x600 in Half-Life 2 is very bad. A cut-down and underclocked version of Larrabee (8 cores @ 1GHz) averaging around 60FPS at 1600x1200 and 4xAA as the paper shows would be quite good in comparison. Most modern high performance GPUs like the 3870 and 4870 scale at 40-50% at the top end. Scaling that poor would have nothing to do the the GPU and more to do with the lack of memory, memory bandwidth, CPU performance or other system/software limitations. A single Radeon 4850 already outperforms a Radeon 3870 by 50% at higher resolutions with AA/AF despite the fact that Radeon 4850 is at a frequency disadvantage. The Radeon 3870 core is clocked 24% higher (775MHz vs. 625MHz) than the Radeon 4850 and has 13% higher memory bandwidth, yet is far behind in performance. Do you also believe Crossfired Radeon 4850 cards would only be about 50% faster than a single Radeon 3870 at higher resolutions because of your knee theory? To see a GPU where the infrastructure was overbuilt, just look at the original R600 (Radeon 2900). Leave it to the guy who believes in the 8 watt Swift and who calculated the 256-bit memory controller on the Radeon 3850/3870 GPUs takes up ~160mm^2 of 192mm^2 of die space to get the 2900XT analysis totally wrong. In comparison to the memory controller the rest of the 2900XT GPU is underbuilt. It couldn't take advantage of the bandwidth because the rest of the core wasn't powerful enough. That is why a single Radeon 4870, not to mention Crossfired Radeon 4870s, annihilates the 2900XT in many benchmarks. They both have about the same bandwidth, but the 4870 core is just that much more powerful.

Report TOU Violation

Share This Post

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next