Silicon Investor (SI) -- The First Internet Community

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) -- Ignore unavailable to you. Want to Upgrade?

To: eracer who wrote (255286)	8/6/2008 5:10:43 PM
From: pgerassi	Read Replies (1) \| Respond to of 275872

Eracer: To answer your PS, no not from AMD, but if you would have googled it ("Fusion 40nm SOI"), dvhardware.net was the first one listed, when I just did that. I didn't think anyone would have thought 8 core Larrabee would be all that fast. If it has a R600 like infrastructure (512 bit memory and bi directional ring bus plus a 16x PCIe interface), it would also have R600 like die size and power consumption making it extremely expensive given the performance. But since like the RV610 has 1/8th the stream processors and memory/ring bus width of the R600, so would the 8 core Larrabee compared to its 64 core brother. At least we can test the RV610 (740G, 780G and 790GX) with real games, real drivers and real systems. Larrabee exists only on paper. As far as graphics performance is concerned, initial Intel estimates of what its GPUs can do are far above what they actually can do when they hit the market, both in performance and in functionality. So when they put out a graph with very little information to relate it to the real world, look for them to use every iota of give to make it look better. Thus if they say four samples, they likely really mean 4AF which games allow AF far more than they allow AA. Its also easier to do AF than AA on GPUs especially given tile based rendering and software rasterization. They also likely used the easiest 25 sequential frames to render in each of those games. The quality settings could be default or lower again because they didn't specify them. Lastly because they didn't profile the ordinary serial code, just the parts that get executed alot, any Amdahl like restrictions on real frame rates are removed (the parts that make scaling bad because they are not parallelizable). Those parts are in the real tests above. Pete