NVIDIA GF100 & Friends speculation
- Thread starter Arty
- Start date
Depends, what are the target clocks for both given products. Clearly ATI was aiming for high clocks, 850Mhz+ for cores. Nvidia is in teh 400-600Mhz Core clock range with domain clocks 1.2-1.4Ghz but dont be surprised to see 1.5-1.6 omain clock for 512SP parts.
Also, if that was the case why should AMD clock the 5830 even higher (assuming the rumored 750 MHz are correct)? And why do almost all 5850s overclock so well? Doesn't make much sense to me...
Not only must all SPs be working, but they will also clock higher than the crippled parts?
Doesn't seem very likely, methinks, unless the available number of chips will be extremely small.
Consider:
* Only a subset of chips will have all SPs functional.
* Only a subset of chips will be able to clock higher than the norm.
* When you combine a subset of a subset, you end up with a tiny fraction of the whole.
...So, don't hope for too much, you could end up getting disappointed.
It is not only a question of what version of the Gf100 they can get out of production, but also what they can sell.
Vendors won´t be to eager to stock up numbers of 550+ $ 480 cards instead of 350-400$ 470 cards.
It is the same basic idea ATI is showing with the 5970. Keep the supply low to keep prices up. If the market is flooded with 480 the resulting price drop affects the whole line-up below the top card.
Could ATI sell a 5870 for its current price if 5970 comes down below 400$. Hardly.
Vendors won´t be to eager to stock up numbers of 550+ $ 480 cards instead of 350-400$ 470 cards.
It is the same basic idea ATI is showing with the 5970. Keep the supply low to keep prices up. If the market is flooded with 480 the resulting price drop affects the whole line-up below the top card.
Could ATI sell a 5870 for its current price if 5970 comes down below 400$. Hardly.
A) castrated SMs don't save much power unless you use power gating, and there have been no signs nv uses them in fermi.
B) IF 512 SM part consumes more power than tolerable, that would mean bad engineering on nv's part. Power and heat are one of the biggest constraints today and if you cannot keep them in check while you bust the reticle's limits, then something somewhere went very wrong.
Haven't said anything of that.
512 SPs at 1700 Mhz Hot Clock 280 TDP
448 SPs at 1700 Mhz Hot Clock nearly 280 TDP
448 SPs at 1500 Mhz Hot Clock 240 TDP
More power than tolerable?! That can only be judged when knowing the overall performance and features.
1700 MHz hot clock. No way, GTX280 had 1300 MHz shader clock. I think they wont reach even 1500 MHz.. Tesla has max 1400 MHz (with still questionable oficial specs range from 1.2 GHz to 1.4 GHz
) and tesla cards had always higher clocks than the graphic versions.Groo The Wanderer
Regular
If it were true, that would be the case. It is not. Then again, he said that Fermi yields were 40% and AMD's smaller Cypress was yielding less. I don't think the boy quite understands the words he is using.
-Charlie
Groo The Wanderer
Regular
If it isn't broken, where are the 750MHz 512SP parts? Are you trying to tell me they didn't actually ship last November?
-Charlie
1700 MHz hot clock. No way, GTX280 had 1300 MHz shader clock. I think they wont reach even 1500 MHz.. Tesla has max 1400 MHz (with still questionable oficial specs range from 1.2 GHz to 1.4 GHz
For the obligatory hair splitting 1296MHz for the 280
Highest clocked first generation Tesla was the 1070 at 1.44GHz which was based on GT200b@55nm and not GT200@65nm. Peak frequency for a single chip GeForce GT200b variant is at 1476MHz for the GTX285. Where's the higher clock for the Tesla exactly?
Target frequency was slightly lower than that and that frequency stood for core clock and not half the hot clock.
"Tolerable" is application dependent.
Relatively few individuals are prepared to tolerate very high power draws regardless of performance and features. Of course, you could say that within its niche this is less of a problem, but that is just another way of saying that its niche is quite narrow, and that power draw helps define it.
It has been strange to see the rest of the industry move towards smaller, cooler, quieter, cheaper, portable and the graphics chips grow ever larger and more power hungry. The Radeon 9700 was the game-changer that broke the 25W power envelope of AGP. I can't help but wonder what it would have meant for the relevance of 3D graphics for personal computers as a whole if the IHVs had stayed within a more modest envelope and developed their architectures from that. The present situation has felt like we're looking at an evolution into a dead end.
Not that I haven't personally thought about it, but on a humorous note NV will re-invent the office chair turning it into a power generating bicycle. You feed that way the PC with some of its needed power and you do your health a favor by not allowing your butt to grow too much
The rest of the industry can go smaller, cooler, quieter, cheaper as even a 2y old dual core is enough for almost everything on a computer. The graphics on the other hand have still a long way to movie like quality.
I disagree. Power draw is an incidental thing that only becomes an issue if it results in a loud cooler or you're super green and count every watt. I'm not sure anyone would care whether they're plugging in two 8-pin connectors instead of two 6-pin. As with most products, price is what defines the niche and there are so many other external factors that it's really hard to point the finger at power consumption. The increasing adequacy of cheaper graphics cards at running console ports probably is doing a lot more to harm the high end market.
I meant tolerable from nv's POV, not consumers.
Tolerable as in within nv's power budget.
For the love of god people, this...is...not...a...Charlie...thread...
Here's a good idea, before you hit the post button, see if your post contains something in relation to Charlie the person. If there is, it's not on topic. You can disagree with his articles/posts/ideas, but try to relate it back to the topic (i.e. don't dig up some irrelevant article and say "HAY LOOK YOU WERE WRONG LAST TIME LOLZ").
Put him on your ignore list if you feel he's too controversial and adds nothing to the discussion.
Here's a good idea, before you hit the post button, see if your post contains something in relation to Charlie the person. If there is, it's not on topic. You can disagree with his articles/posts/ideas, but try to relate it back to the topic (i.e. don't dig up some irrelevant article and say "HAY LOOK YOU WERE WRONG LAST TIME LOLZ").
Put him on your ignore list if you feel he's too controversial and adds nothing to the discussion.
Similar threads
- Deleted member 2197
- Architecture and Products
