Yes, Nvidia. Despite the 3x market cap advantage and the fact Nvidia's Geforce 9 series is coming out very soon, I believe they will be the losers over the next two years.
Firstly, their chipset business is in ruins. They are forcing manufacturers to release an entirely new series of expensive motherboards (7xx) with just a bugfix to make them run what everyone expected to run on their 6xx series: 45nm Intel processors. The dissipation of their chipsets is extremely high, so much so that Intel is annoyed at having to include two of their chips in Skulltrail. Nvidia's chipsets no longer offer significant advantages over Intel's or AMD's in price or features, especially as Crossfire matures.
Secondly, the Geforce 9 series is just a rebadged Geforce 8 series. The third highest projected SKU, the 9800GT, is rumoured to be just an 8800GT with Tri-SLI support. The second highest, the 9800GXT, will be an 8800GTS with slightly higher clocks and be no more than 15% better than the 8800GTX released well over a year ago. The top SKU, the 9800GX2, will suffer from all of the problems of SLI (Nvidia have shown no commitment to improving SLI, wheras AMD has shown their dual card, the HD3870X2, can provide nearly transparent Crossfire.), have an extremely high dissipation dueto the 65nm technology and idle-state power reduction being behind AMD's. Also due to the 65nm process and a lesser degree of integration between the cores than HD3870X2, the production is more expensive which could put the retail cost of the 9800GX2 $100 or $150 above the HD3870X2. There is a high chance of supply issues too.
Thirdly, you haven't seen AMD's real high-end yet. The HD3870X2 is out, but the CrossfireX driver that will provide even better performance and transparency than the already impressive current Crossfire driver isn't out yet.
Anandtech's CrossfireX preview:
Configuration HL2 UT3 Bioshock CoD4 Crysis
2-way CF Improvement over 1 card 83% 80% 71% 98% 87%
3-way CF improvement over 2 cards 30% 34% 37% 44% 0%
4-way CF improvement over 3 cards 10% 3% 7% 29% 4%
4-way CF improvement over 1 card 160% 150% 151% 268% 98%
Those aren't even final numbers - that's just a testing build. When the final driver is released in March, the HD3870X2 has the potential to improve in benchmarks by 20% due to Crossfire improvements. This doesn't account for single-card monthly driver improvements which AMD will have had three or more months of by the time the 9 series is established.
Fourthly, there has been plenty of independent rumours about R700. Final silicon has been returned, and some potential model numbers have been leaked. It is looking set for a late Q2/early Q3 release, according to some sources, and with the native multi-core architecture it will show the power of AMD's new Crossfire and be cheaper to manufacture with better yields than a big single-core card. In contrast, there have been very few rumours about Nvidia's next generation - instead we hear that some low-end 9 series cards could be released in May and June, almost ruling out a big new launch immediately after. I think R700 will be on its own in the field for a quarter of so, and being a new generation it will destroy both G9x and RV670 in all the benchmarks.
Fifthly, standards support on the Nvidia side is lacking. AMD were the first to DirectX 10.1 (and Nvidia doesn't have it in the Geforce 9 series either), the first to PCI-E 2.0, the first to DisplayPort and the first to have double-precision for GPGPU computing.
Sixthly, take a look at this article: http://www.3dprofessor.org/Reviews%20Folder%20Pages/FireGLV8650/FireGLV8650P1.htm
AMD, with the 80nm "failed"R600 architecture-based FireGLs, has taken back the performance leadership in the workstation market while being significantly cheaper than Quadro FX. With RV670 at 55nm and then R700 at 55nm or 45nm, think how much more of a performance lead AMD could get with no real improvements forecast on the Nvidia side for six months or more. That was previously a very profitable market for Nvidia; could their market share fall?
Finally, with Intel's new GPU architecture Larrabee and AMD's Fusion project integrating 45nm revised Phenoms with GPU cores, Nvidia won't be able to offer as much value in their products in the longer term. I can't make solid predictions beyond 2008 though, so perhaps Nvidia will have some initiative of their own: Geforce as a PPU, perhaps? AMD have also opened up their GPU specifications and are supporting an open-source driver project. This could end up with a better quality driver and more support from the growing group of free software advocates (I bought an AMD card for my Ubuntu computer for that reason).
So, Nvidia doesn't have much of a chance in the next few years. I think they will try and offer better price/performance as AMD did in 2007 - with the corresponding reduction in revenue.
Firstly, their chipset business is in ruins. They are forcing manufacturers to release an entirely new series of expensive motherboards (7xx) with just a bugfix to make them run what everyone expected to run on their 6xx series: 45nm Intel processors. The dissipation of their chipsets is extremely high, so much so that Intel is annoyed at having to include two of their chips in Skulltrail. Nvidia's chipsets no longer offer significant advantages over Intel's or AMD's in price or features, especially as Crossfire matures.
Secondly, the Geforce 9 series is just a rebadged Geforce 8 series. The third highest projected SKU, the 9800GT, is rumoured to be just an 8800GT with Tri-SLI support. The second highest, the 9800GXT, will be an 8800GTS with slightly higher clocks and be no more than 15% better than the 8800GTX released well over a year ago. The top SKU, the 9800GX2, will suffer from all of the problems of SLI (Nvidia have shown no commitment to improving SLI, wheras AMD has shown their dual card, the HD3870X2, can provide nearly transparent Crossfire.), have an extremely high dissipation dueto the 65nm technology and idle-state power reduction being behind AMD's. Also due to the 65nm process and a lesser degree of integration between the cores than HD3870X2, the production is more expensive which could put the retail cost of the 9800GX2 $100 or $150 above the HD3870X2. There is a high chance of supply issues too.
Thirdly, you haven't seen AMD's real high-end yet. The HD3870X2 is out, but the CrossfireX driver that will provide even better performance and transparency than the already impressive current Crossfire driver isn't out yet.
Anandtech's CrossfireX preview:
Configuration HL2 UT3 Bioshock CoD4 Crysis
2-way CF Improvement over 1 card 83% 80% 71% 98% 87%
3-way CF improvement over 2 cards 30% 34% 37% 44% 0%
4-way CF improvement over 3 cards 10% 3% 7% 29% 4%
4-way CF improvement over 1 card 160% 150% 151% 268% 98%
Those aren't even final numbers - that's just a testing build. When the final driver is released in March, the HD3870X2 has the potential to improve in benchmarks by 20% due to Crossfire improvements. This doesn't account for single-card monthly driver improvements which AMD will have had three or more months of by the time the 9 series is established.
Fourthly, there has been plenty of independent rumours about R700. Final silicon has been returned, and some potential model numbers have been leaked. It is looking set for a late Q2/early Q3 release, according to some sources, and with the native multi-core architecture it will show the power of AMD's new Crossfire and be cheaper to manufacture with better yields than a big single-core card. In contrast, there have been very few rumours about Nvidia's next generation - instead we hear that some low-end 9 series cards could be released in May and June, almost ruling out a big new launch immediately after. I think R700 will be on its own in the field for a quarter of so, and being a new generation it will destroy both G9x and RV670 in all the benchmarks.
Fifthly, standards support on the Nvidia side is lacking. AMD were the first to DirectX 10.1 (and Nvidia doesn't have it in the Geforce 9 series either), the first to PCI-E 2.0, the first to DisplayPort and the first to have double-precision for GPGPU computing.
Sixthly, take a look at this article: http://www.3dprofessor.org/Reviews%20Folder%20Pages/FireGLV8650/FireGLV8650P1.htm
AMD, with the 80nm "failed"R600 architecture-based FireGLs, has taken back the performance leadership in the workstation market while being significantly cheaper than Quadro FX. With RV670 at 55nm and then R700 at 55nm or 45nm, think how much more of a performance lead AMD could get with no real improvements forecast on the Nvidia side for six months or more. That was previously a very profitable market for Nvidia; could their market share fall?
Finally, with Intel's new GPU architecture Larrabee and AMD's Fusion project integrating 45nm revised Phenoms with GPU cores, Nvidia won't be able to offer as much value in their products in the longer term. I can't make solid predictions beyond 2008 though, so perhaps Nvidia will have some initiative of their own: Geforce as a PPU, perhaps? AMD have also opened up their GPU specifications and are supporting an open-source driver project. This could end up with a better quality driver and more support from the growing group of free software advocates (I bought an AMD card for my Ubuntu computer for that reason).
So, Nvidia doesn't have much of a chance in the next few years. I think they will try and offer better price/performance as AMD did in 2007 - with the corresponding reduction in revenue.
Facebook
Twitter
Instagram