The GT300/Fermi Thread

[Florinmocanu]

FINALLY someone hit the nail on the head....

I just wish I had said that myself.

Well, than, your simulated result is not anymore relevant than the result with the 1080 core 5870.

we have to see results made with the same identical test-bed/settings in order to check the level of performance. Simulating the same setup is practically impossible, there are always things which change the outcome.

[hennyo]

I know I'm going to have a hell of a time if what I saw and my other info is wrong, but here it comes:
The GTX 380 will be nearing the 5970 in most games, and will have a significant advantage over the 5870.
Regarding the price... well, if you're a realist and live under no illusions, then it's easy to say that it will be priced similarly to the 5970, normally 50 bucks lower or so.

This information sounds about right for the performance, but I couldn't even venture a guess at the price point myself. There is something deceiving here tho in the way you are describing the performance here. What should be noted is yes the GTX380 part will be trading blows with the 5970 based on what we know and it will most likely slightly lose in most scenarios, but what isn't mentioned here is what "most scenarios" are, or as you put it "most games".

The fact of the matter is, that "most games" the 5970 plays at frame rates that are far higher than what is needed to get the absolute most out of them. It is the few games and scenarios where the 5970 is not enough power is where the GTX 380 will out shine the former. Things such as high IQ Crysis settings, extreme resolution, high usage of DX11, and things such as texture mods. Another important difference is the GTX 380 will be night and day in minimum frame rates in places that can bring these GPU's to their knees. All these factors combined will paint a much lighter picture for the GTX 380 in real world use over the 5970 than traditional benchmarks will let on to.

[jaredpace]

So far they have AI on in all these tests... don't see any difference

price i guess $550-650

[Chumbucket843]

Or they could simply beeline for HD6870...
If AMD plans the release of HD6870 one year after HD5870, then HD6870 would be as much a competitor for Fermi as Fermi would for HD5870.

are you implying that ATi is a full generation behind? we dont even know anything about r900 but if it is a new architecture on a new process then it will take a while to launch. that's being a little risky. there is a lot that could go wrong. see delay of fermi as an example.

[grimREEFER]

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.

[SKYMTL]

Well, than, your simulated result is not anymore relevant than the result with the 1080 core 5870.

we have to see results made with the same identical test-bed/settings in order to check the level of performance. Simulating the same setup is practically impossible, there are always things which change the outcome.

Hence why it is "simulated".

The results were never presented as definintive in any way, shape or form regardless of which way people are taking them. However, considering that I was able to replicate their GTX 285 benchmarks quite well, it goes to reason that the results are a good approximation of overall performance.

No, there isn't any way to replicate any driver tweaks, Windows tweaks or anything on the software side from the GF100 system but once again: that's why the results were simulated.

Heck, for all we know the GF100 in that system was running balls-out overclocks that won't be seen in the production samples...all I did was put the results people videotaped into charts and added a HD 5870 running in a near-identical system. It isn't rocket science nor do I have any reason to hobble the ATI card. If the ATI card perfromed as good as the GF100, we would probably be getting even more hits.

[LesGrossman]

I didn't say the price doesn't matter for "me", it's just that when you qualify the perfomance of the cards and compare them you should compare the same cards like 1vs1 or 2vs2 gpus ... Is my english so bad I have to repeat this simple idea every time? sheesh

So comparing a 334mm chip with a 550mm chip with more 40% transistors is fair? And price doesn't matter? Since when?

I guess the drama goes on and the "deep dive revelations" only made it worse.

[Setix]

Ok so I think I missed something. When the NDA lifted yesterday, what information was able to be released?...I didnt really see anything...

[phsinc1]

http://www.hardware.fr/articles/782-...ometrique.html

[Sam_oslo]

FINALLY someone hit the nail on the head....

I just wish I had said that myself.

However, it begs the question: if the HD 5870 is CPU-bound in the tests, why isn't the GF100 as well?

Good question is bugging you.
Maybe because the GF100 is "almost" a CPU by itself? Maybe this new architecture is capable of running more of those stuff that traditionally was running on CPU?

[Jamesrt2004]

FINALLY someone hit the nail on the head....

I just wish I had said that myself.

However, it begs the question: if the HD 5870 is CPU-bound in the tests, why isn't the GF100 as well?

you ran it on a better processor lol.. jk i dunno what you did but im guessing intel i7 so yeah.

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.

yeah this shrunk on 28nm will be like a frigging rocket, will be amazing..

and Your correct on ATI thats why they are preparing a new arc to launch H2 this year probably more towards Q4.

[SKYMTL]

you ran it on a better processor lol.. jk i dunno what you did but im guessing intel i7 so yeah.

NVIDIA = i7 960 + 6GB + ASUS Rampage II Extreme

Me = i7 960 + 6GB + Gigabyte X58-UD5


Basically the same setup minus the motherboard which doesn't do all that much anyways when you get to high IQ settings.

[Sam_oslo]

Same CPU, RAM and the same X58 MB, then it has to be the new GPU- architecture making this differences.

This is really interesting and needs further investigation, But
I put my bet on the new F100-architecture, which is more capable of running CPU-stuff on the GPU.

[KingOfsorroW]

So comparing a 334mm chip with a 550mm chip with more 40% transistors is fair? And price doesn't matter? Since when?

I guess the drama goes on and the "deep dive revelations" only made it worse.

Oh sure - it's all about drama here About the transistors - it doesn't matter, as long as it is 1vs1 gpu comparison. The technogy inside the gpu is the very competion. When you have two separate gpus on a single card - then it's a different story. Then, you look for the same 2headed graphcis card at the competitor camp and compare them. That's how I see it personally - the fair comparison.
Then there is price of course, which you may take into consideration or your decision may be driven solely by it. that would depend on your budget of course.

[h0bbes]

Seems like fermi will not be as good as it was supposed to be and not as bad as it was hoped to be (by some)

[ajaidev]

NVIDIA = i7 960 + 6GB + ASUS Rampage II Extreme

Me = i7 960 + 6GB + Gigabyte X58-UD5


Basically the same setup minus the motherboard which doesn't do all that much anyways when you get to high IQ settings.

Can you test the 5870, 5870 @ 1ghz and 5970 with the exact settings as the GF100.

Ya all will be simulated but then again its a good side scale

[zalbard]

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.
Assuming the original 34.64% performance improvement over 285 we already get 67.77fps.
And assuming drivers improved the performance by 10% (yea, that's where BS starts, obviously, I suppose it's less due to CPU bottlebeck? and if there's no bottleneck it's probably more) we get 74.54fps. Stock clocks.



And one more edit: who is going to run these cards with under-4GHz processor anyway? So there should be no CPU bottleneck, but that requires a whole new test I am afraid.

[NickF]

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

[eric66]

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.

over dramatizing much ? 5870 isn't 3870 there is maximum 25 percent difference between 5870 and gf100 in fc 2 bench plus 5970 isn't 5870 in cf its more like 5850 cf so no you fail sorry

[Solus Corvus]

And? I didn't bother scrolling back through that whole conversation between you all, but if the tests were all done with the same hardware (CPU, Mobo and RAM) and only changing out the GPUs, what does that have to do with anything?

Didn't they all have a common baseline?

No, the Xbit numbers were on a completely different system. Comparing the minimum framerates xbit got for 5970 with what nvidia got for fermi is meaningless.

FC2 is very CPU limited and I'm pretty sure they use either a 3.8Ghz or 4Ghz i7 which in my tests could equate a 15-18% improvement in the Ranch Small benchmark over a chip clocked at 3.2Ghz.

You need to pay attention...
They left A.I. ( Artificial Intelligence ) enabled, thus the benchmark becomes CPU limited and differences between differently clocked or even maker CPUs arise.
Considering that in a sum of games the 5870 needs an i7 clocked at 3.7GHz to be CPU-limit-free, and there you have it.

FINALLY someone hit the nail on the head....

I just wish I had said that myself.

However, it begs the question: if the HD 5870 is CPU-bound in the tests, why isn't the GF100 as well?

Then why are AT's numbers for 285 and 5870 lower at 1920x1200 but higher at 2560x1600? All I'm saying is that the numbers look a bit off, IMO, and we need a proper review to know for sure.

[Sam_oslo]

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

http://images.anandtech.com/graphs/rv870_092209122344/20103.png[/IMG]

i7 960 @ 3.2GHz used.

http://images.hardwarecanucks.com/image//skymtl/GPU/GF100/GF100-42.jpg[/IMG]

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.

5-15%? I'm afraid it won't help. You see if these numbers should be off even with 20% , then the GF100 is still so superior that you need at least 30% on top av that to consider the 5870 as a competitor.

[eric66]

why people think that 5870 must compete with gf100 ? atis strategy was about small efficient gpus right ? i don't remember them saying that we will try to create the worlds fastest single gpu since 2000 disaster so they learned there lesson no more power hungry expensive gpus. new strategy is dual gpu as high end so you actually must compare gf100 to 5970 to be fair in both power consumption heat and price gf100 will be much more than 5870 and close to 5970

[NickF]

Yeah, I just drank some of charlie's hatorade. I have to say, I think his analysis seems to have a bit more technical support and rational than nVidias comparison with a chip its not intended to compete against.

Edit, also I'd like to point out that even the 5850 managed to defeat the GTX 285. So having a 67% FPS increase isn't really too astonishing.

[zalbard]

to be fair in both power consumption heat and price gf100 will be much more than 5870 and close to 5970

But not performance wise it seems!

[ajaidev]

Raw method very useful in things of a eyefinity nature

vs



Very Intelligent method CSAAis a mix of 8x MSAA and 24 coverage samples/pixel also useful in things of a eyefinity nature

I dont think 32x CSAA will be of much use in modern games, anyways here is the sampling for CSAA and SSAA.

"As for consumption and noise, they should instead be placed in category Radeon HD 5970 than 5870 which is sure to fuel further debate. "

http://translate.google.co.in/transl...ial%26hs%3D73v

This is what i have been saying compare 5970/5950 to GF100