NVIDIA GF100 & Friends speculation
- Thread starter Arty
- Start date
Until amd is making less money than nv from it's graphics side, I wouldn't worry about it in the slightest. And afaik, nv makes ~2.5x more revenue per quarter.
What isn't good?
Makes you wonder whether Charlie with his claimed 275W hit the nail originally on its head (with the original frequencies) and if the rumor mongerers at vrzone just try to make up excuses for the absurd 295W TDP.
It's always nice to see that if you plant somewhere freely invented hypothetical initial frequencies they show up the next day. Desperation for headlines must be high in such cases I guess and apparently way too many buy anything you feed them without at least trying to cross-verify anything. Since Alex loves that quote here it is again: to steal from one source is plagiarism to steal from many research.
LOL you so want this to be true, that you don't even bother to read exactly what's written there and that that is the exact same rumor from a while back, only copy-pasted to vr-zone, with some rumored frequencies in the mix.
Maybe you should read what Ailuros said after your post
itsmydamnation
Veteran
LOL you so want this to be true, that you don't even bother to read exactly what's written there and that that is the exact same rumor from a while back, only copy-pasted to vr-zone, with some rumored frequencies in the mix.
Maybe you should read what Ailuros said after your post
hi pot meet kettle....
There are about as many hints to that as there are to the opposite. I get that you like NV, but you can't just claim something is true because it's the rumor you happen to favor. Then again, consider this: 512 is just 6.67% more than 480. That's not much, you can easily make up for that with higher clocks, especially since GPUs usually scale a bit better with clocks that with additional units.
I think this isn't the case of GF100. Better clock scaling of previous GPUs is caused by boosting geometry performance. But geometry performance of GF100 is dependant not only on clocks, but on number of functional blocks, too.
Hopefully double-digit Fps.
You know that by now, nice package
or arn't you Carsten from pcgameshardware? (great site btw).You mean like many are claiming it's 480 also by selecting a rumor they favor ? I don't see you quoting and correcting them over that...
As for your "if it has 480 ALUs then it will most likely have higher clocks", thanks for proving my point earlier.
The post I quoted before that is getting me so much flak, is from someone claiming it will have 480 ALUs and that the TDP difference from 448 to 480 is only 25 watts, but from 480 to 512 it will be 45 watts, which then added that that would be due to higher clocks i.e. suggesting that the fully enabled chip would have higher clocks than the not-fully enabled chip with 480 ALUs and that is the exact opposite of what you, me and actually most of us expect.
Plus others added that the TDP difference could also be justifiable by the enabling of other units like TMUs and ROPs, with which I agree. But couldn't this hypothetical 480 ALUs version have everything else enabled, except that SM with 32 ALUs ? A definite possibility and in that case, the 45 watts difference in the rumored TDP, doesn't make sense.
Last edited by a moderator:
Based on the architecture specs, that's my opinion too. It makes more sense for the chip to be fully enabled and with rather modest clocks (given the problems they had), than disabling some units on the chip and clocking everything else much higher. This line of reasoning applies only to the high-end model of course.
Most people who claim it has 480 SP claim to know it from their own sources, not to choose the rumor they like best. And lately, those people (Fudo, Hardware-Infos, VR-Zone, I think Neliz too) seem to be more numerous than the ones saying 512.
As for the rest, there are reasons why there might be a larger TDP gap between 480 and 512 SP than between 448 and 480, and those reasons have already been discussed in this thread. The main argument was variability and the consequence of said variability on the minimum operating voltage.
That said, 295W does seem a little bit over the top.
For the record, I didn't say that "if it has 480 ALUs then it will most likely have higher clocks", my point was that whether it has 480 or 512 SPs, it's not going to make a very significant difference, but clocks are.
Finally, the whole point may be moot anyway, since it seems there might be very few GTX 480s compared to the number of 470s.
I've just ran unigine on my mates new 5870, with catalyst 10.3
Comparing it to http://www.youtube.com/watch?v=vpdPSZB8A8E (the nVidia vid), it's much closer at the part where it crosses the road and goes over the grass. The 5870 is 42/43 fps and rising compared to the 480's 44/45 and rising. I think it started closer to 39fps before 10.3.
The dragon never drops below 21 fps first time, and 20 fps second time which is quite an increase over Petersen's low point of 16 fps.
All settings are exactly the same except my mate has a Phenom II at 3.6 ghz. It does seem that the 10.3 catalyst has pulled back some performance, especially near the dragon.
Comparing it to http://www.youtube.com/watch?v=vpdPSZB8A8E (the nVidia vid), it's much closer at the part where it crosses the road and goes over the grass. The 5870 is 42/43 fps and rising compared to the 480's 44/45 and rising. I think it started closer to 39fps before 10.3.
The dragon never drops below 21 fps first time, and 20 fps second time which is quite an increase over Petersen's low point of 16 fps.
All settings are exactly the same except my mate has a Phenom II at 3.6 ghz. It does seem that the 10.3 catalyst has pulled back some performance, especially near the dragon.
I am & thanks!You know that by now, nice package
But believe it or not - I haven't played Metro 2033 for one second yet (regardless of GFX hardware). So I really don't know the answer.
Additionally, if they really ran into a power wall, higher clocks may be the less advisable choice unless they can really turn the disabled SM(s) completely off.
AFAIK, only intel can do it with bulk processes so far, and AMD can't/won't do it on 45nm or older SOI.
However, power-gating seems to be quite prevalent in the low power devices. I wonder how they do it if power gating transistors are hard to do if you are not Intel.
Similar threads
- Deleted member 2197
- Architecture and Products
