The GT300/Fermi Thread

[Sam_oslo]

It is too early to be too sure about the performance, but based on current info, this seams to be very like the 880GTX vs 3870x2 scenario.

Nvidia fights back with a new architecture. A single GF100 GPU which seams to be superior to the single ATi, and "almost" as good as the double GPU ATi.

[JohnMike]

Nvidia fights back with a new architecture. A single GF100 GPU which seams to be superior to the single ATi, and "almost" as good as the double GPU ATi.

Well...... ATI has the chance and time to strike back once GT300 arrives. Right now ATI 5### gen is leading and there's space to improve.

For nVidia is 'run and don't miss....or else!'

[OverShocked]

Not sure if anyones posted this yet, but here ya go...

http://www.maximumpc.com/article/fea...dias_gf100_gpu

[Sam_oslo]

Yeah, I hope ATi manages to put up a good fight. at least with refresh if not a new GPU.
Otherwise is is going to be bad news for the price of the GF100. A superior Nvidia GPU has a tradition to become expensive.

[xVeinx]

why people think that 5870 must compete with gf100 ?

Eric: This isn't directed at you, more the idea. Please don't take this personal. I'm not bashing you in any way.

Attempting to break things down into what comparison is appropriate becomes subjective at best (including my past attempts at the same, for which I apologize ). Every card needs to be compared to every other card currently available. Let the chips fall where they may, and spend the time comparing feature sets or AA methods instead or worrying about what card should be compared to what card (and thereby who "wins" at any given level). If your bent on high FPS and like a certain feature set, then you buy a 5970 (assuming that the GF100-GTX380 variant is indeed slower). Vice Versa if the GTX380 is faster. Yes, you want the better product, and people may disagree at times which one that is. Attempting to figure out what comparison's are "legal" or "right" though gets to be too much. I don't think it's because we're all a bunch of rabid fanboys that we do this, but more that we assume that this is some sort of competition which has a complex set of rules. It's not our competition. It's up to ATI and Nvidia to compete, in a number of categories, and the rules are somewhat simple. ATI and Nvidia get to feed off of our obsessions with the latest benchmarks and hardware, tossing a graph or benchmark result our way, and letting the fur fly as a result. We don't win anything in the competition (except the right to spend lots of money on hardware we rarely use to it's full capacity), so it's a little silly to get too defensive of this.

[Nedjo]

Can't wait to see the availability on this one, people thought 5870 was hard to find.

well it'll come down to the level of compromises and cost of binning.

regardless of this... err... I really don't know how to call it - PR specs stunt, maybe... we still don't know plethora of really important questions, so we can continue with speculation like we did so far!

There are couple of scenarios with different probabilities:

a) highly unlikely - NV launches GTX380 w/ 512 CCs (Cuda Cores) with scarce availability and hig price tag in line with 5970, but still slower!

b) probably - NV launches GTX 380 w/ 448 (512-2x32) CCs with top five partners and tight availability but with ~15% higher performance compared to 5870 (and in line of 2x GT200 promise) /w $499 pricetag!

"b" scenario leaves place for GTX390 in Q3 /w 512 CCs (six months should be sufficient time to gather as much as possible 512 CC working chips).

Also in line with "b" scenario is assumption that demand can be serviced with 384 CCs (448-2x32) part called GTX360 that would go against 5850 and most probably be with it in line when it comes to performance.


But this salvaging process isn't cheap, and it is time consuming, so it cant' be long term answer! Also it makes you think wouldn't NVIDIA be batter with smaller chip on time, compared to butchered big (apparently somewhere between 550 and 570 mm^2) chip that isn't economical justifiable without shrinking in 28nm tech!??

[annihilat0r]

let's say GF100 IS A LOT FASTER THAN HD5970!! BIG DEAL!!! we'll now play games with 250 fps instead of 200 FPS!!!!

each passing day I care a little less about gaming performance. ONE 5870 is more than enough for everything save eyefinity.

This is why you gotta applaud Nvidia for at least trying to do something different. I hope their past track record is indicative of the pressure they are now going to apply on the game developers to use more complex geometry / tessellation.

But of course, we know that before the next generation of consoles, any new graphics technology advance will be just a gimmick.

Console kids. You have killed gaming.

[bfar]

Hats off to Nvidia for developing a brand new architecture. GF100 may not end up being as quick as we'd like, but remember the R600 - that was no great shakes, but it laid out a path for 3 fantastict generations of ATI technology. The future could be bright.....

[omar little]

has it been established whether fermi uses hardware or software tesselation?

[omar little]

Eric: This isn't directed at you, more the idea. Please don't take this personal. I'm not bashing you in any way.

Attempting to break things down into what comparison is appropriate becomes subjective at best (including my past attempts at the same, for which I apologize ). Every card needs to be compared to every other card currently available. Let the chips fall where they may, and spend the time comparing feature sets or AA methods instead or worrying about what card should be compared to what card (and thereby who "wins" at any given level). If your bent on high FPS and like a certain feature set, then you buy a 5970 (assuming that the GF100-GTX380 variant is indeed slower). Vice Versa if the GTX380 is faster. Yes, you want the better product, and people may disagree at times which one that is. Attempting to figure out what comparison's are "legal" or "right" though gets to be too much. I don't think it's because we're all a bunch of rabid fanboys that we do this, but more that we assume that this is some sort of competition which has a complex set of rules. It's not our competition. It's up to ATI and Nvidia to compete, in a number of categories, and the rules are somewhat simple. ATI and Nvidia get to feed off of our obsessions with the latest benchmarks and hardware, tossing a graph or benchmark result our way, and letting the fur fly as a result. We don't win anything in the competition (except the right to spend lots of money on hardware we rarely use to it's full capacity), so it's a little silly to get too defensive of this.

those are a lot of words when the answer is quite simple: they will compete based on price point
the vast majority of potential buyers couldnt give a rats ass about eyefinity or that silly 3d thing that requires dorky glasses and an uberexpensive display

[zerazax]

this whole thing sounds like its gonna be another 8800gtx vs 3870x2. except this time, the ati card was released first.
and as with the 8800gtx, nvidia has basically set itself up with a good architecture that will be seen over and over in future gpu's, while the sense im getting from ati is that their 5000 series will need a complete overhaul.

Then you'll be happy to know that the 5xxx was the last generation (barring a refresh) to be based on the R600. The next generation is due to be a new microarchitecture

But that's why I think it's likely ATI has a refresh on 40nm within the next 6 months based on RV870 and then releases a 6xxx later, probably on 28nm after they do a RV740-esque run on 28nm first

[tbone8ty]

anybody confirm this test results yet with the 5870?

wonder what settings there were at

http://en.expreview.com/2010/01/18/g...thed/6475.html

[fiskov]

Exactly what i wanted by the looks of it, as good as the 280 GTX in SLI, with a DELICIOUS lack of Catalyst Control Center.
Sign me up, i'll take 2 please.

[Teemax]

GF100 beating HD 5870 by 60%?

Oh plllleeeeeaaaase.

[BababooeyHTJ]

why if they cost the same.. or even less... lol

do people not understand...

5970 was ALWAYS ati's card to fight against nvidia's TOP SINGLE card.. in price AND performance.....

so it's fair... as thats what has been known for like a year now..

anyone who says different is just pissed cos nvidia didnt do aswell as they hoped.

I'm sorry but you can not compare a dual gpu card to a single gpu card. Performance does not always scale. I had a 4870x2 and it's performance compared to my 280 on average was very disappointing. There were games that ran fine on my 280 especially with mods that just ran horribly on the x2. I can't think of any game where it was the other way around. Sure the x2 performed better in some games but never by that much and usually in games where I didn't really need it and mods seemed to make things worse for the x2. Drivers make or break a multi gpu setup and with a single gpu you don't need to worry about that as much.

That said I can't wait to see what price that the 5870 hit since it seems like a solid card but I still wouldn't touch a 5970 with a ten foot pole.

Edit: I forgot to mention the 4870x2's (and from what I hear hemlock) disappointing overclocking potential. My 280 hits about a 25% overclock on the core on air, I don't know of too many x2s that can pull that off (or anywhere near it) on water.

[Sam_oslo]

GF100 beating HD 5870 by 60%?

Oh plllleeeeeaaaase.

Yep, you bet. At least that's what these early results indicating

[LinkinParkBoy]

[LordEC911]

Including architecture scaling, GAMING PERFORMANCE and a whole wack of other stuff.....


RIGHT HERE!!!!!

Did you ask on the slides of synthetic tests how they used a 5870?
According to ChrisRay, some of those test/demos/benchmarks use PhysX...

[BenchZowner]

GF100 beating HD 5870 by 60%?

Oh plllleeeeeaaaase.

Surely not true!
But wait... is it illegal to be true ?
Or is nVIDIA a new company that cannot touch AMD/ATi ?

[SKYMTL]

i7 920 @ 3.3GHz used. 5870 is 34.64% faster than 285.



i7 960 @ 3.2GHz used.



5870 is just way too low (just 27.69% faster than 285), NV is BSing us. 5870's score should in fact be even higher with updated drivers, anand's review is ancient - add 5-15%, not totally sure about the accurate figure.

No.

You can see that our "simulated" results for the HD 5870 are actually within a few FPS of AT's. So, vindication right there.

The thought that new drivers added up to 15% performance to Far Cry 2 is absolutely false as well.

[SKYMTL]

Did you ask on the slides of synthetic tests how they used a 5870?
According to ChrisRay, many of those test/demos/benchmarks use PhysX...

Which synthetic tests? Unigine, Far Cry 2, HawX, DX11 Toolkit, etc. don't use PhysX at all.

[zalbard]

GF100 should in fact be much faster than 5870 in tessellation due to its architecture, not surprising here. Same for general geometry. Talking about the rest, though... it doesn't seem nearly as bright.

[tbone8ty]

anybody confirm this test results yet with the 5870?

wonder what settings there were at

http://en.expreview.com/2010/01/18/g...thed/6475.html

[InCredible]

I'm sorry but you can not compare a dual gpu card to a single gpu card. Performance does not always scale. I had a 4870x2 and it's performance compared to my 280 on average was very disappointing. There were games that ran fine on my 280 especially with mods that just ran horribly on the x2. I can't think of any game where it was the other way around. Sure the x2 performed better in some games but never by that much and usually in games where I didn't really need it and mods seemed to make things worse for the x2. Drivers make or break a multi gpu setup and with a single gpu you don't need to worry about that as much.

That said I can't wait to see what price that the 5870 hit since it seems like a solid card but I still wouldn't touch a 5970 with a ten foot pole.

i understand what you're saying..but overall you compare things based on performance and cost. simple as that..if you want to know what the fastest card you can get for X amount of dollars then thats how you compare..regardless if its xfire/sli/single gpu or dual gpu..

theres pros and cons to these things of course and sometimes its better to deal with 1 gpu vs 2..but thats something for the consumer to decide

fastest card is fastest card period..
this went over when 5870 came out and all the nvidia guys were like its still not the fastest card blah blah blah..and ati guys were saying yea but id rather have a single gpu vs dual..

same thing will be said now just flop the brands around
and i do agree id rather have a single gpu vs two if its competitive
however id wait to see just how competitive it is and how well it will overclock under water..since we know the 5970 can handle a good overclock.

also itll be nice to see what ati does about fermi(refreshes and price drops)

[zalbard]

You can see that our "simulated" results for the HD 5870 are actually within a few FPS of AT's. So, vindication right there.

On AT's graph 5870 is faster and 285 is slower, so can't agree with you here, Nvidia shamelessly shaved those 7%.

The thought that new drivers added up to 15% performance to Far Cry 2 is absolutely false as well.

Yeah, OK, I checked some driver comparison tests and I take it back. I said I wasn't sure anyway. Seems to be within 5% in most cases, rarely up to 10.