Fastest DX11 GPU On The Planet From Nvidia - GTX580? [via Engadget]

This is the first I heard of this card being cooler and quiter, I thought they were just squeezing more juice out of their usual Fermi.
 
This is the first I heard of this card being cooler and quiter, I thought they were just squeezing more juice out of their usual Fermi.

We've known it was supposed to be cooler and quieter for several days already... this is the first I'd heard of the vapor-chamber HSF for them stock however.
 
Same thermal solution as ATi used on the reference 5970s...

5970_400w.jpg
 
I'll only upgrade to nVidia's "flagship" cards when they get their heads out of their asses and put more than 1.5GB video ram on them for us surround users.
 
I'll only upgrade to nVidia's "flagship" cards when they get their heads out of their asses and put more than 1.5GB video ram on them for us surround users.

You say upgrade, so I can only assume you are using a lower end card with less memory. Why then would you complain about only 1.5gb being memory bound but stick with something even more constrained? I haven't heard anyone else complain about the 1.5gb being a bottleneck, have you actually monitored your ram utilization during gameplay to verify this is a limiting factor? If it is, just turn down AA a notch.

Having said that, its a bit surprising there isn't a Zotac version with the doubled 3gb out. This makes me even more skeptical that you are actually memory constrained since they have GTX 460's with 2gb of ram and would certainly jump on customising the higher margin flagship if there were a market.
 
Question ofcourse being will it stay that way past Nov 22nd.

Regardless I want the GTX 560. 384 Cuda core Fermi sounds like my next card.
 
Yawn, how the in the world is this anything post worth when it's been in every NVidia 580 rumor thread posted directly from the source. It's starting to get really annoying having threads on some article based on forum rumors that started here or were posted here already. Talk about snowballing..
 
Always the smaller online retailers that end up listing cards first. When will it show up on Egg? That's when I know it's totally legit.
 
I think the one coming afterwards will be the fastest. It would be crazy for AMD to not buy 2 or 3 GTX 580s and overclock then and test them and set their clocks to outperform them. Even in a bad world where Cayman's tech wasn't up to par I would have to believe it would come very close to GTX 580.

Any card that comes later has that advantage. However nothing can stop nvidia from claiming the fastest when they launch since they will more than likely launch first (11/9/2010)

The fun part is nothing is stopping AMD from then making the same statement when they release their Cayman (11/22/2010) if it's not delayed like all the rumors are now saying.
 
Well since this GPU was paper launched in 2009,
1263608214xxTstzDnsd_1_3_l.gif

I hope it actually makes it out this year.

Anyone ever remember a paper launch that lasted more than a year?
 
It would be crazy for AMD to not buy 2 or 3 GTX 580s and overclock then and test them and set their clocks to outperform them.


Why would AMD be crazy for not having their cards outperform overclocked 580s?

First, the overclocks will vary. Should a stock 6970 beat the most overclocked 580 in the world?
 
Why would AMD be crazy for not having their cards outperform overclocked 580s?

First, the overclocks will vary. Should a stock 6970 beat the most overclocked 580 in the world?

No. However since they are launching after they would get alot of flak if they are slower. So why not buy a few GTX 580s bench them and test the reference card and set your reference speed to be a faster card. That's all I'm saying. Every company who comes out with the product after has that advantage. Why not take advantage of that.
 
Last edited:
Well since this GPU was paper launched in 2009,
1263608214xxTstzDnsd_1_3_l.gif

I hope it actually makes it out this year.

Anyone ever remember a paper launch that lasted more than a year?

The card the market is getting this year is 128 Texture units though. So that card from last year, has paper launched and will never see the light of day. ;)
 
Last edited:
No. However since they are launching after they would get alot of flak if they are slower. So why not buy a few GTX 580s bench them and test the reference card and set your reference speed to be a faster card. That's all I'm saying. Every company who comes out with the product after has that advantage. Why not take advantage of that.

Why would AMD do that? They haven't done that in the past. Why would they do it now? The 4870 was launched after the 280 and it was still slower.

But it doesn't matter to me. It all depends on what it is priced at. If the 6970 is 5% slower than the 580 but costs $100 less I know what card I'm going to get. These last few years AMD has gone for the sweet spot; that is small and affordable, but efficient and fast chips. They are going for perf/price.
 
No. However since they are launching after they would get alot of flak if they are slower. So why not buy a few GTX 580s bench them and test the reference card and set your reference speed to be a faster card. That's all I'm saying. Every company who comes out with the product after has that advantage. Why not take advantage of that.


Faster makes sense but if you look at your original post you said they should overclock them then test them and make it beat theirs...
 
Well since this GPU was paper launched in 2009,

I hope it actually makes it out this year.

Anyone ever remember a paper launch that lasted more than a year?

The GPU in GTX 580 is not the same as GF100. GF100 will never see the light of day @ 40 nm, in any retail products.

Also, a "paper launch" applies to a product, not a GPU. GF100 is not a product.
 
The card the market is getting this year is 128 Texture units though. So that card from last year, has paper launched and will never see the light of day. ;)

It's doubtful that GF110 has 128 TMUs, but we will see once the reviews start showing up. GF110 seems to be a "thinner" GF100, without HPC specific features, which reduces the number of transistors and die size.
 
So what are the specs?

Trying to piece it together. This is what I'm getting so far:

772Mhz GPU clock (source)
512 cuda cores
64 texture units
1.5Gb of gddr5
192.4 Gb/s bandwdith
384-bit wide memory bus

Maybe we can put that in a table compared to the 480, 460, 285 and try to extrapolate its performance.
 
Last edited:
They do specifically say 'Fastest GPU' and not Video Card, so I think we can assume they're not aiming it higher than the 5970.
 
Faster makes sense but if you look at your original post you said they should overclock them then test them and make it beat theirs...

Correct, it would make sense for them to overclock it to see how the GF110 Scales no? I mean why have the card and not find out everything you can about it. Perhaps even examine the cooler, and see if they can apply a similar cooler in time to allow cayman to run even cooler for an even higher stock clock. These are the types of strategic things AMD would have the luxury of doing since the GTX 580 will probably be out on the retail channel for over a week and a half before Cayman is due (nov 22) if it's(cayman) not delayed of course, as current rumors are saying.

So yes have the Cayman 5970 run faster than GTX 580 at stock vs stock, but also see how it overclocks to know what the architecture and overclocked counterparts to be sold will be capable of.
 
It's not a finished review....I don't even think people are supposed to have access to it.
 
It's down. Anyone get a mirror up? By now someone must have managed to score one.
 
No. However since they are launching after they would get alot of flak if they are slower. So why not buy a few GTX 580s bench them and test the reference card and set your reference speed to be a faster card. That's all I'm saying. Every company who comes out with the product after has that advantage. Why not take advantage of that.

The 580 was just announced around a month ago, I think Cayman was too far along in the process to be making any changes without incurring steep costs.
 
The 580 was just announced around a month ago, I think Cayman was too far along in the process to be making any changes without incurring steep costs.

I'm talking about raising their core/shader/memory clocks, not retweaking the architecture. by getting 580 they can see what it's capable of and set their clocks to beat fermi 580 or ensure what they already have will beat it. They can also get a good idea as to what the pre overclocked 580s will bring to the table. They'll also know if their cooler will cut it for those clocks and if they need a newer cooler.

etc.. etc.. etc...
 
Back
Top