The GT300/Fermi Thread - Part 2!

[SKYMTL]

I realize that with squirrel cage coolers the goal is pressure, not max CFM. But IMO, 11.3CFM isn't going to be sufficient to cool a GPU.

Are you seriously arguing that ATI needs a ~10W cooler to cool a 5870 but that Nvidia found a way to get by with ~3W on a hotter GTX480?

It all comes down to efficiency of the heatsink. I'm just pointing you to the specs of the fan itself.

As for the HD 5870, doesn't it use a Globe Fan blower-style unit rated at less than 5W? I could be wrong though....

[trinibwoy]

Therefore, doesn't matter if you got Fermi, G200, or G92, if "only" 90% of code can be parallelized, anything beyond 10 SP or CPU wont be faster. Thus with 512 "cores", maximum efficiency requires 99.8% parallelization.

The proportion is also affected by the speed of the CPU. i.e the faster the CPU, the smaller the serial part becomes. Yes, it's just a fancy way of saying that as you remove the CPU bottleneck the proportion of parallelizable work tends toward 100% of the total. But even on the GPU a lot of stuff is serialized - geometry setup for example. Hence the multiple setup engines in Fermi. The more stuff you parallelize, the easier it is to scale performance.

[Chumbucket843]

Plz correct if I'm wrong..

Amdahl's law states that if P is the proportion of a program that can be made parallel (i.e. benefit from parallelization), and (1 − P) is the proportion that cannot be parallelized (remains serial), then the maximum speedup that can be achieved by using N processors is

( (1-P) + P/N )^-1

Therefore, doesn't matter if you got Fermi, G200, or G92, if "only" 90% of code can be parallelized, anything beyond 10 SP or CPU wont be faster.
Thus with 512 "cores", maximum efficiency requires 99.8% parallelization.

Seeing how vast majority of programs, and even games are barely even optimized for 2 cores, you can imagine how much ingenuity this requires on GPU side.

not much, gpu applications like graphics are embarassingly parallel and this parallelism increases with problem size. i.e. if you double the pixel count you double parallelism. this law sounds very grim but truthfully its not. gpu's are already running thousands of threads to hide latency.

[Solus Corvus]

It all comes down to efficiency of the heatsink. I'm just pointing you to the specs of the fan itself.

As for the HD 5870, doesn't it use a Globe Fan blower-style unit rated at less than 5W? I could be wrong though....

http://pic.xfastest.com/sxs112/ASUS/...SUS5870-33.jpg

[saaya]

That would be totally awesome, 480 / 5x70 extreme edition.

xtreme edition
id def pay 100$ ,or even more, extra to have a card that is fully unlocked and has access to dividers and lets me enable and disable different blocks

You can't safely push more than 10W through a standard GPU fan header.

theres a standard for gpu fan headers?
even if, how would you know if the 470 or 480 have a higher spec for fans? what IF they would consume a lot of power? nvidia would run into an issue cause according to the spec they cant use a powerful fan to cool the card?

Not to mention that this is only a 75mm fan which means 21W could translate into 23,000+ RPMs.

huh? how do you get to 23k rpm? i dont think rpm scale linearily...

I don't know where that 1.8A comes from but it is either a misprint or some random number thrown in.

i used to test a lot of fans and measured their amps, and those ratings were always higher than what i actually meassured, regardless of the voltage i fed them and if they were spinning up or running full speed... still, 1.8A is a very high rating for a fan, and while it may mean nothing, it MAY point towards higher power consumption... a bigger heatsink costs a lot more than just a more powerful fan that increases the amount of heat it can remove...

and if you look at that pic of the 470 heatsink again, how exactly could it be any larger? the only way to make it longer would be to move the fan to the right, which isnt possible because the holes in the pcb would be very close to the right edge of the pcb, meaning there wouldnt be enough space to put components on there. and the only way to make it higher would be to go above atx case specs... and that wouldnt help much cause the air channel fed by the axial fan isnt as wide as the entire card, so... that heatsink is the biggest exhaust heatsink you can fit on a 470 card...

Deimos, you proved that its possible to write terrible code that performs incredibly slow, so what?
i think nvidia is pretty good at getting efficiency out of their asics... they designed them! its a custom instruction set asic, instructions that THEY thought up and implemented... if anybody knows how to use that design efficiently its nvidia themselves

but maybe im just assuming a lot here... we all know how sad things look in many companies when you peek behind the pr facade

[kemo]

http://pic.xfastest.com/sxs112/ASUS/...SUS5870-33.jpg

That fan can push at least 110CFM and it is 92x92x38mm and looking at the 470GTX with fan of 75X75X30 it does make sense to have higher Amps to push the same CFM but still both ratings are very high .. and with 1.8A that is definitely too much rpm


Edit: anybody has any info about the 285/280GTX fan

[Mungri]

I bought the highest performance Delta fan I could find for my PC once, just looking at the RPM and CFM numbers.

I think that could be what caused my hearing loss

I think I still have it somewhere. I dont care how well or bad the Fermi performs, but I'm sure that no one wants another FX 5800 paintdryer.

[Aerwidh]

Just 2-3 years ago, your stupidity for trying to play a game on IGP would be rewarded with startup crashes.[..] Due to the wrong moon phase.

Yeah, those are particularly annoying

[Solus Corvus]

That fan can push at least 110CFM and it is 92x92x38mm and looking at the 470GTX with fan of 75X75X30 it does make sense to have higher Amps to push the same CFM but still both ratings are very high .. and with 1.8A that is definitely too much rpm


Edit: anybody has any info about the 285/280GTX fan

1.8A does seem like a lot. But We don't know anything about the thermals of these chips either. And the fans won't be spinning 100% all the time. You can't draw any conclusion just from the max power draw, you have to know the fan profile as well.

For you and SKYMTL to dismiss it as a typo because you can't imagine a cooler needing a fan with that much power is a simple argument from incredulity - a common logical fallacy.

[saaya]

actually, id prefer a card with a fan that CAN push a lot of air even if its noisy over one that cant...
but would i want it to be noisy at stock speeds under load? of course not....

about the fx comments... well maybe noise wise, but that would be a really stupid mistake from nvidia...
perf wise gf100 wont be able to fail as bad as nv30... no way... you guys dont remember how bad geforce fx really was... it was slower than its predecessor, something that is simply (luckily) impossible with gf100 as the architectures of gt200 and gf100 are too similar...

[SKYMTL]

For you and SKYMTL to dismiss it as a typo because you can't imagine a cooler needing a fan with that much power is a simple argument from incredulity - a common logical fallacy.

It's a fact. Just got off the phone with a rep at Delta. Listed spec is the CSA maximum for the input amperage that the fan motor will support. Info on their site is the "typical" usage parameters.

However, we will see when the reviews are released and comments are made about how loud / quiet the fan is.

[-Sweeper_]

any idea on GTX480 PCB size?

[Solus Corvus]

It's a fact. Just got off the phone with a rep at Delta. Listed spec is the CSA maximum for the input amperage that the fan motor will support. Info on their site is the "typical" usage parameters.

However, we will see when the reviews are released and comments are made about how loud / quiet the fan is.

So the number on the fan is the maximum power draw and the number on the website are "typical" usage?

[LiquidReactor]

So is everyone yet starting to realize that Fermi will be smaller, cooler and faster then what everyone expected?

So the number on the fan is the maximum power draw and the number on the website are "typical" usage?

You want to call Delta yourself? Give it up already

any idea on GTX480 PCB size?

Smaller then 5870 and similar to GTX 285/280.

[Solus Corvus]

You want to call Delta yourself? Give it up already

Give what up? I'm just trying to clarify his statement.

But thanks for trying to discourage further discussion about one of the few solid pieces of information we have.

[SKYMTL]

So the number on the fan is the maximum power draw and the number on the website are "typical" usage?

No worries.

Amperage rating on the fan: maximum amperage that the motor will support within CSA guidelines (not sure what that means). Needed for certification.

Info on their website: Typical operating RPMs, etc.

[Luka_Aveiro]

Even if the fan is 1.8A, 100% could not translate into 1.8A.

I remember w1zzard talking about something related to fan control through BIOS, so, even if the fan CAN pull 1.8A, BIOS might be able to not let it happen.

Right?

[Solus Corvus]

No worries.

Amperage rating on the fan: maximum amperage that the motor will support within CSA guidelines (not sure what that means). Needed for certification.

Info on their website: Typical operating RPMs, etc.

That makes sense. Did he say what the F in the model number stood for? They have a PDF that explains their naming scheme, but the F didn't make any sense according to it.

It seems like a rather powerful fan and a decent cooler. But we really can't draw any conclusion about how hot the chip is without knowing what the fan profile is. We will have to see how loud it is when it is released, but personally I wish they had used a bigger fan.

[mindfury]

2900XT fan:

G80 and GTX280 fan:

[SKYMTL]

That makes sense. Did he say what the F in the model number stood for? They have a PDF that explains their naming scheme, but the F didn't make any sense according to it.

He wasn't quite sure since there are specific and custom models that Sales has access to but he alluded to the fact that it could be a designation for the fan without the blower housing it usually comes in.

[-Sweeper_]

looks like GTX470 SLI...

anybody with two HD5870s to compare?

[Yvese]

looks like GTX470 SLI...

anybody with two HD5870s to compare?

Why bother? The 5870 is obviously going to lose since it has less memory.

Show some benches of 2GB 5870, or use 1920x1200 res.

[mao5]

480:

http://bbs.pczilla.net/attachments/m...ee82e7a341.jpg

5870 1GHz/1250MHz (core/mem) same setting as 480 in that video with Tessellation on

http://bbs.pczilla.net/attachments/m...994258a635.jpg

5870 1GHz/1250MHz (core/mem) same setting as 480 in that video only Tessellation off

http://bbs.pczilla.net/attachments/m...94d96c3616.jpg

5870 1GHz/1250MHz (core/mem) Tessellation on/off comparing

http://bbs.pczilla.net/attachments/m...c75a2f937a.gif

[annihilat0r]

And? 1ghz 5870 performs same as GTX 480?

[-Sweeper_]

480:

http://bbs.pczilla.net/attachments/m...ee82e7a341.jpg

5870 1GHz/1250MHz (core/mem) same setting as 480 in that video with Tessellation on

http://bbs.pczilla.net/attachments/m...994258a635.jpg

5870 1GHz/1250MHz (core/mem) same setting as 480 in that video only Tessellation off

http://bbs.pczilla.net/attachments/m...94d96c3616.jpg

5870 1GHz/1250MHz (core/mem) Tessellation on/off comparing

http://bbs.pczilla.net/attachments/m...c75a2f937a.gif

there's very little tessellation on these scenes..

show me a framerate comparison at the dragon now