The Fermi Thread - Part 3

[ToxAvenger]

If by 5870 killer you mean a card on average 10-15% faster ,consuming more power and OCing worse while costing much more to make(huge die),then yes, it's a 5870 killer alright.

Your sarcasm skill are finely polished. I personally will wait for proper reviews before I pass judgment, although I must admit that Fermi may very well be shaping up to be a letdown.

[eric66]

yep amd did exactly the same they cherry picked res where gtx 295 ran out of memory and made us believe that they had gtx 295 killer when in reality it got spanked by gtx 295 nearly in all res except 2560 !

[informal]

Your sarcasm skill are finely polished. I personally will wait for proper reviews before I pass judgment, although I must admit that Fermi may very well be shaping up to be a letdown.

Well I'm not saying it will be a letdown "per se" ,but it is indeed very late to the market and not promising big increases in gaming performance compared to what we have today in high-end segment(I'm not talking about GPGPU though,in this segment it will rock,I'm sure).

[Chumbucket843]

If by 5870 killer you mean a card on average 10-15% faster ,consuming more power and OCing worse while costing much more to make(huge die),then yes, it's a 5870 killer alright.

5870 is a great card but that doesnt mean that the gtx480 will be bad. if power rumors are true that probably means good overclocking. the high power consumption is from leakage power which is great for clockspeed but bad for efficiency. keep in mind that enthusiasts want peak performance. they want 32x AA and 16x AF on the latest titles on their 2560x1600 monitor.

how do you know how 400 series costs much more? you should do some research first. even if the die costs 2x more the other costs will lower difference in BOM.

[Sn0wm@n]

leakage doesnt mean high oc ...... it only means bad transistor....


are the c2 revision phenom better then the c3 revision ???? nope ... they got a better process etc... less leakage .... see where im going ....

wich all result in a higher and more stable oc

[ToxAvenger]

5870 is a great card but that doesnt mean that the gtx480 will be bad. if power rumors are true that probably means good overclocking. the high power consumption is from leakage power which is great for clockspeed but bad for efficiency.

Does this apply to TSMC? I seem to remember the higher leakage parts being better overclockers on the AMD on Intel side, this is also true of GPUs?

[informal]

leakage doesnt mean high oc ...... it only means bad transistor....


are the c2 revision phenom better then the c3 revision ???? nope ... they got a better process etc... less leakage .... see where im going ....

wich all result in a higher and more stable oc

You are (mostly) correct .The high power of the 480 model indicates NV needed to boost core voltage beyond their initial targets(low voltage was the goal from the start). By raising the core voltage they entered the unfortunate skyrocketing of the power draw because the already leaky trans. in already castrated die (512SP->480/448)began to leak even more. The OC headroom will largely depend on one's ability to cool this monster so water cooling may be the best option for enthusiasts who will buy this card and OC it. For stock/default users, heat shouldn't be a problem at all (if one likes hot piece of electronic equipment radiating heat in their case ).

[Solus Corvus]

I agree that there isn't enough objective data to make any conclusions about GTX480 right now. I'll wait for a real review. It could even be a 5970 killer or whatever, it's not impossible it just seems unlikely. We will see the reality soon enough. I just hope there is price competition. It'd be nice to have GTX470 be about as fast as a 5870 and then GTX480 would probably be right between 5870 and 5970.

Anyway, my stuff is just an honest guess and I am more then willing admit as such. But with all the conflicting information being spread as fact, some people will soon have to recall the counter-factual information they disseminated. It really was shameful to watch so many people claiming to have inside sources and then spreading different information. And yet we still don't have any real solid performance information. But that's ok, that's what a real review is for anyway. I'm sticking by my GTX480 on average 30% faster then 5870 prediction because I made it in good faith. I'd note that it was with full shaders and clocks at the time. Anyway, on with the reviews.

[Chumbucket843]

Does this apply to TSMC? I seem to remember the higher leakage parts being better overclockers on the AMD on Intel side, this is also true of GPUs?

it applies to any process. that's why 45nm AMD phenoms are almost hitting 7GHz under ln2. r600 was good under ln2 because it was on leaky 80nm process which was chosen for its raw performance although having an unfinished architecture didnt help.

if the rumors about leaky xtors are true then 480 should be more power efficient by going for higher clocks with less shaders.

[Sn0wm@n]

You are (mostly) correct .The high power of the 480 model indicates NV needed to boost core voltage beyond their initial targets(low voltage was the goal from the start). By raising the core voltage they entered the unfortunate skyrocketing of the power draw because the already leaky trans. in already castrated die (512SP->480/448)began to leak even more. The OC headroom will largely depend on one's ability to cool this monster so water cooling may be the best option for enthusiasts who will buy this card and OC it. For stock/default users, heat shouldn't be a problem at all (if one likes hot piece of electronic equipment radiating heat in their case ).

so cooling your card will limit the leakage to a bare minimum ???? and thats the reason top oc'er claim high leakage part oc better ????

but get temps closer to absolute zero and you get hyperconductivity wich would allow voltage to flow more freely ????? so maybe there is a reason they like high leakage part on ln2 and liquid hellium????

[informal]

so cooling your card will limit the leakage to a bare minimum ???? and thats the reason top oc'er claim high leakage part oc better ????

but get temps closer to absolute zero and you get hyperconductivity wich would allow voltage to flow more freely ????? so maybe there is a reason they like high leakage part on ln2 and liquid hellium????

Won't limit the leakage but will get the uncontrollable temperature (under stock cooling) to manageable levels ,so it will not hamper the functionality of the device. Remember ,the parts may leak as hell,but if you cool them down they will work (for how long it's unknown as it is with any MPU that you drive out of specs).

[Sn0wm@n]

thanks for the explanation

[Your_Boss]

These are supposed to be new.

I don't like showing pics over and over but these are few and far between so here they are again on a new page for you.(so you don't have to keep searching back in the thread)

Does anyone know what ATI driver version they used? 10.3a?, 10.2? 10.1? ..maybe 11.9? Who know?

Also, can anyone confirm these Farcry2 and Dirt2 scores with their own 5870? What are others getting at these same rezs? Put settings on highest, but without AA, except for the one that mentions AA. Thanks.

[migunani]

what is power draw of 480 ?

depends on your mood ... and the color of your pant...

[ajaidev]

I made this using what hardware-infos say's is the final specs and those prices are launch ones not retail.

[Mungri]

That table seems to look correct. Although any of the information so far could still be false, I think a lot of it is right now.

[eric66]

5870 eyefinity edition draws more nearly 210 watt or something and price is 449

[onethreehill]

[IMG]http://img717.imageshack.us/img717/2629/uploadee.png[IMG]

I made this using what hardware-infos say's is the final specs and those prices are launch ones not retail.

Hexus has a similar table
http://www.hexus.net/content/item.php?item=23008

[Aerwidh]

I made this using what hardware-infos say's is the final specs and those prices are launch ones not retail.

Why use different prefixes for pixels/second and flops on the Nvidia cards, compared to the ATI cards

[ToxAvenger]

Why use different prefixes for pixels/second and flops on the Nvidia cards, compared to the ATI cards

Maybe Nvidia lists it that way? And shouldn't the 5870 EF6 have a higher power consumption due to doubling up on memory? The AMD slide I remember stated 225 watts, or there abouts IIRC.

[Frontl1ne]

I don't like showing pics over and over but these are few and far between so here they are again on a new page for you.(so you don't have to keep searching back in the thread)

Does anyone know what ATI driver version they used? 10.3a?, 10.2? 10.1? ..maybe 11.9? Who know?

Also, can anyone confirm these Farcry2 and Dirt2 scores with their own 5870? What are others getting at these same rezs? Put settings on highest, but without AA, except for the one that mentions AA. Thanks.

Not the best system in the world, but I ran FarCry 2 just then:

System:

3x ATi 4870 (stock clocks)
AMD Phenom II X4 B50 @ 3.7GHz
4GB DDR3 1333
Windows 7 x64

FC2 Settings:

Settings: Demo(Ranch Small), 1920x1200 (59Hz), D3D10, Fixed Time Step(No), Disable Artificial Intelligence(No), Full Screen, Anti-Aliasing(None), VSync(No), Overall Quality(Ultra High), Vegetation(Very High), Shading(Ultra High), Terrain(Ultra High), Geometry(Ultra High), Post FX(High), Texture(Ultra High), Shadow(Ultra High), Ambient(High), Hdr(Yes), Bloom(Yes), Fire(Very High), Physics(Very High), RealTrees(Very High)

FPS:

GTX480 (From graph):

Min: 72
Avg: 92

5870 (From graph):

Min: 47
Avg: 66

Mine:

Min: 66
Avg: 88

One thing you guys should note though, my average graphics load was 50 - 60% So yeah, they were held back a bit

[Your_Boss]

Thanks for the benches.

but I was hoping for someone with a 5870. Where are ya? come out of the woodworks and bench with 10.3a.(if ya got 'em)

[Vardant]

I was basing my estimates on a full 512 shader gpu at rumored target clocks. If GTX480 is indeed cut down or with significantly lower clocks then I'd estimate even lower. As for beating a 5970, you can believe that if you like but I doubt it. Even Nvidia's estimates, with full shaders and target clocks, was 40% faster then 5870. That's not enough to match 5970.

40% faster than 5870 is exactly where the 5970 is, lol

[GoldenTiger]

40% faster than 5870 is exactly where the 5970 is, lol

Yeah, pretty much exactly.

[btdvox]

Will there be a 2+ GB version of the GTX 480?