DX ten point one.nvidia am fail.
I heard that detail of Larrabee will be available in next IDF.
Nvidia GT300 core: Speculation
- Thread starter Shtal
- Start date
- Status
I heard that detail of Larrabee will be available in next IDF.
Mintmaster
Veteran
No, they should look to their own strategy of how to be a highly profitable business. With the exception of G80, they've never gone for much more expensive GPUs than their competition in any generation except NV30, and they've been most successful when they had substantially better products than their competition at any given production cost (G7x, G8x, G9x before RV770).
Anyway, these specs look rather stupid given that they aren't self-consistent.
No mention of DP...
What do you guys think, will this generation (consumer cards) have double precision FP, as the current rumour is, that the GT200b will not have it?
I find the current Nvidia DP implementation with the 1:12 speed penalty quite useless anyway, as it can in most cases even emulated faster.
What do you guys think, will this generation (consumer cards) have double precision FP, as the current rumour is, that the GT200b will not have it?
I find the current Nvidia DP implementation with the 1:12 speed penalty quite useless anyway, as it can in most cases even emulated faster.
Chances this is real: zero
reason: die size is IDENTICAL to GT200
common sense...
reason: die size is IDENTICAL to GT200
common sense...
Yeah, just like unification and HDR+MSAA :smile:What? The Nvidia focus group members tell me DX 10.1 is pointless.
Paul choose to absent himself from discussions at RWT, he was not banned. He is welcome back at any time as long as he can be polite.
Large IPF chips cost Intel relatively little to produce because:
1. They have a lot of cache with redundancy, so point defects don't reduce ASP
2. They are produced on older process technology that is fully depreciated
In contrast, GPUs are mostly logic where point defects will cause a chip to be binned down and lower the ASP.
GPUs are also produced on leading edge process technology which TSMC is depreciating through the prices charged to AMD/NV (so they have to pay indirectly).
But let's suppose that a 500mm2 chip costs around $150 to produce. That's fine when you sell the chip for $600, you've still got nice margins.
But when you sell a whole system (board, cooler, chip, DRAM) for $450, the margins are much much worse. The DRAM is probably $100 alone, the cooler is going to cost a lot simply due to weight (say $20), the PCB has >10 layers and probably costs about 30 or 50, so right there your margins have been cut down tremendously.
DK
Always good to hear from you DK, thanks for stopping by! Can you comment any further on Paul's decision to leave? As stubborn as the guy was, I probably learned more from reading his articles on RWT than just about anywhere else (Johan de Gelas' work on Chip-Architect being the other).
Sorry for the OT, and thanks for the IPF analysis! Without Paul around to do it for us it's good to have someone else step up to the plate
Just a quick note on graphics card costs though:
You missed a couple steps of the chain, however - NV buys GT200 "kits" (DRAM, GPU, PCB, cooler) for perhaps $200, then sells them to AIB partners for perhaps $300, who then sell to resellers/distributors/OEMs for perhaps $400, who then resell for $450-$600. Note: all numbers are my rough estimates. I could be off by +/- $50 at any point of the chain.
Sorry for the OT, and thanks for the IPF analysis! Without Paul around to do it for us it's good to have someone else step up to the plate
Just a quick note on graphics card costs though:
You missed a couple steps of the chain, however - NV buys GT200 "kits" (DRAM, GPU, PCB, cooler) for perhaps $200, then sells them to AIB partners for perhaps $300, who then sell to resellers/distributors/OEMs for perhaps $400, who then resell for $450-$600. Note: all numbers are my rough estimates. I could be off by +/- $50 at any point of the chain.
Last edited by a moderator:
Maybe but at this point I don't think DX10.1 is expected from them. If it's indeed the case that the base G8x architecture doesn't lend itself well to certain DX10.1 functionality there really isn't a compelling reason for them to make substantial changes until DX11.
GT200 @ 65nm 576 sq.mm die area
GT200 @ 55nm 470 sq.mm die area
GT300 @ 55nm 576 sq.mm die area
I was wondering about how would Nvidia be able to squeeze into 106 sq.mm die area - extra 240 stream processors for Total 480 stream processors into 576 sq.mm die area.
http://www.techpowerup.com/index.php?64193
I've always been curious as to what functions those really are? Nvidia - no wonder - is very tight lipped about it and no one ever mentioned which and foremost why DX10.1 should be giving them such a hard time.
Dave B(TotalVR)
Regular
If you ask me, the 'GT300' (if indeed that is how it will be named) will simply be a 55nm die shrink of GT200 with no major differences other than the obvious, power consumption, clock speeds and YIELDS.
There is a good chance they might include GDDR5, however, allowing them to lower the number of memory controllers and vastly increasing the memory clock speed. I'd sooner expect this on a 8800GTX -> 9800GTX style makeover though.
There is a good chance they might include GDDR5, however, allowing them to lower the number of memory controllers and vastly increasing the memory clock speed. I'd sooner expect this on a 8800GTX -> 9800GTX style makeover though.
Reducing the number of crossbar MC channels would also require a reduction in the number of ROP partitions. NV would have to re-architect their ROP partitions to keep the same fillrates.
Didin't you just describe GT200b?
Assuming the abundant fillrate of GT200 is of any use to begin with.......
True, but the implication is to reduce to 4 or 6 ROP partitions, (G92 or G80 level, respectively) and I would think that would be quite the step backward...
edited because eye kant spel
Last edited by a moderator:
Dave B(TotalVR)
Regular
- Status