Comment Thread for GTX 480 and GTX 470 Reviews

[sideeffect]

Originally Posted by MonsterSound

So, how green your vid card is, is supposed to be the most important thing to judge when reviewing the latest most powerful vid card technology? Better get me a 5450.

What I was trying to highlight is that the architecture of the 5800 series can be clocked a lot higher than its current default clocks on air cooling and still retain better thermals and power usage than fermi.

On the other hand the Fermi is already clocked close to its maximum on air cooling.

The most powerful video cards are sold to enthusiasts. Enthusiasts mostly care about what a card is capable of rather than what it does at stock speeds. This includes fundamentals such as performance per watt, TDP, Power usage and performance scaling which all lead to overclocking and performance.

That is the most important thing to consider. If you can buy a 5850 and overclock it so that its faster than a 480GTX which seems to be the case while it still is running cooler and using less power then why would an enthusiast want a fermi?

[MonsterSound]

Originally Posted by sideeffect

What I was trying to highlight is that the architecture of the 5800 series can be clocked a lot higher than its current default clocks on air cooling and still retain better thermals and power usage than fermi.

On the other hand the Fermi is already clocked close to its maximum on air cooling. ....

Yes, I got it. Not news anymore.

Originally Posted by sideeffect

... Enthusiasts mostly care about what a card is capable of rather than what it does at stock speeds.

A lot of assumptions there.

Originally Posted by sideeffect

This includes ....
That is the most important thing to consider. If you can buy a 5850 and overclock it so that its faster than a 480GTX which seems to be the case while it still is running cooler and using less power then why would an enthusiast want a fermi?

OC'd 5850 faster than the 480? Good luck with that. I'd like to see your results, especially with DX11 (man tessellation looks great).
P.s. I'm very happy with my 5850 which plays great and overclocks well. Glad I got it early and at a good price. I'd also trade it for a 470.
Sorry if I'm sounding snarky, but the comments about heat, noise ( "with a crazy loud fan " ), and power seem overblown to me as well as the "mostly" and "most" generalizations/assumptions: and it's late.
peace

[belgolas]

Did anybody take the time to read the review? SKY said the fan is not that loud especially compaired to the 5970.

Also the 5850 can not come near the 480. MAYBE if it was under liquid nitrogen. But overclocking goes both ways. A good water cooled GTX 480 might be faster than the 5970 but then the 5970 could also be overclocked as well.

Who gives a hoot about power consumtion for the fastest card? Of coarse it will consume more power. Do you think a F1 car will be more fuel effiecent than a Honda civic?

Also your arguement about tesselation and nvidia's card not knowing if it could do that plus high graphics is a dumb arguement. Look at metro 2033 one of the BEST looking games in VISUALS and has tesselation and nvidia DOMINATES it. Look at the unigen benchmark and that has some of the best visuals in it aswell as tesselation. As far as the better DX11 card nvidia appears to be the better card. Maybe it was a good thing that nvidia was 6 months late so that when real DX11 games came out that they could test and iron out their cards. When ATI released their cards there were no DX11 games till much later so they had no idea what it would be like in DX11.

Also you can sit and argue about what would have happened if nvidia released first but this is how it is and it is utterly pointless to argue about what ifs.

[sideeffect]

GeForce GTX 470 & 480 review

Dirt 2 benchmark 480Gtx is shown at around 18% faster than a 5850 in Directx 11 at 1920x 1200 and 2560 x 1600.

My results at 5300 x 1050 Maximum settings AA off. Both tests performed using a replay of the same race exactly the same viewing angle.

725Mhz core 1000Mhz memory (stock)
Time (ms), Min, Max, Avg
60000, 31, 41, 37.033

1000Mhz core 1250Mhz memory (overclock) Air cooled 73 degrees maximum.
Time (ms), Min, Max, Avg
60000, 39, 51, 46.467

Difference 25% faster from overclock.

[burebista]

Originally Posted by belgolas

Look at metro 2033 one of the BEST looking games in VISUALS and has tesselation and nvidia DOMINATES it.

Hope you're joking right? 6 FPS's more means dominate? You can call it playable at 1920 without PhysX, without AA at 25 min and 32 average?
And what if you turn on 4AA and PhysX? You'll have better framerates?
Nevertheless Fermi looks promising for DX11 games but probably with next revisions.

[sideeffect]

Tessellation is in all the DirectX 11 games I have seen but the performance hit from enabling them on ATI cards is not that much. The difference in Metro 2033 results are not due to tessellation.

The game was released after the last ATI driver and has had no driver performance updates or profiles yet. The game was also tested and optimised for Nvidia cards under the TWIMTBP program. I doupt ATI even got a look at the game before before it was released. On the other hand the other Directx 11 games have all been out a while and Nvidia has had access to them.

[SKYMTL]

Originally Posted by burebista

Absolutely agree.
But what games we have with DX11 and Tessellation? I show them here. Here in UE GTX 480 is ~50% pricey than HD 5870. It worths now for gaming?

There is a massive pricing disparity from region to region whenever there is a new card released. Personally, I could never see myself spending more than $500 before tax of on GTX 480.

Originally Posted by burebista

BTW SKY if you still have both Fermis's around I have something to ask you. Can you made a benchmark in Metro 2033 1920x1200 with 4AA/16AF, DoF, Tessellation, PhysX and everything on max? I did't see those settings benchmarked anywhere (understable because ATI cannot render PhysX) but I want to see how's doing Fermi in their marketing game.
Remember that Optimum settings recommended for Metro are Nehalem, Fermi, SSD and 8 GB RAM.

Sure, but it may take a while since my system is currently being used to bench cards in prep for HD 5770 and HD 5850 roundups.

Originally Posted by sideeffect

Interesting review. Just some points to consider.

Minimum FPS is important but showing just 1 number as a minimum does not highlight the real situation. A Graphics card might dip to a low number for a single second in a benchmark run for a number of reasons but for the rest of the time may actually have a higher minimum FPS. The proper way to display the minimum FPS is using a graph.

Our minimums are a based on an average of two runs. That's as good as it's going to get for the time being unfortunately.

Originally Posted by sideeffect

The 5800 on the other hand was very conservatively clocked on release in order to meet a good TDP.

See, I don't agree with this. IMO, the HD 5870 was clocked quite high as is evidenced by how hard a time board partners are having when it comes to releasing pre-overclocked versions. Even with chip binning and extreme testing, companies like ASUS, Gigabyte, MSI, Sapphire, etc have all failed to release cards in excess of 950Mhz. We are told even 925Mhz has become an issue with the latest batches of chips. Supposedly, the HD 5850 clocks are even worse which has forced some board partners to cancel certain SKUs.

As for performance in Directx 11 games the results in most reviews have shown that both series are very close to each other. Dirt 2, AvsP, Bad Company 2, Stalker all show a very small difference in performance. Metro 2033 is very new and Nvidia had more access to the game before launch so performance might even out in that game also.

The don't show a small difference. From what I see from the results, most reviewers failed to enable DX11 properly in both Dirt 2 and BF: BC2. Their results line up almost perfectly with some preliminary DX9 runs I did at the beginning of the benchmarking process.

Originally Posted by traitoR

not trying to butt heads but i read more than once your suggestion that others should wait on an ATi purchase for Fermi and alluded to it looking to be very strong in comparison. I at the time thought you must have some inside info but in retrospect it looks like you did indeed have high expectations that carried over somewhat over enthusiastically.

I don't see where the issue is with that. The GTX 470 in particular provides better performance per dollar than the HD 5850. That to me is worth waiting for when it comes to certain people's cases.

crazy loud fan and have better cooling ?

I suggest you check TPU and Hardware.fr which did tests with a dB meter. The GTX 480 is no louder than a HD 5970.

You have to wonder if as per the norm these are cherry picked parts that run cooler and clock well, for the initial seeding of reviewers, just what will the mainstream cards a guy can buy........

I don't think NVIDIA did this since there just wasn't time. They literally turned around their shipment from Asia within 12 hours and sent the cards to reviewers.

The last time I heard about cherry picked cards was ATI's seeding of the HD 5830 to reviewers where they sent out cards which were completely at odds with what was available at retail.

Originally Posted by sideeffect

Tessellation is in all the DirectX 11 games I have seen but the performance hit from enabling them on ATI cards is not that much. The difference in Metro 2033 results are not due to tessellation.

The game was released after the last ATI driver and has had no driver performance updates or profiles yet. The game was also tested and optimised for Nvidia cards under the TWIMTBP program. I doupt ATI even got a look at the game before before it was released. On the other hand the other Directx 11 games have all been out a while and Nvidia has had access to them.

Don't agree. Metro 2033 had plenty of ATI dev help as well which is evidenced by their working Crossfire profiles when the game was released.

[sideeffect]

Originally Posted by SKYMTL

See, I don't agree with this. IMO, the HD 5870 was clocked quite high as is evidenced by how hard a time board partners are having when it comes to releasing pre-overclocked versions. Even with chip binning and extreme testing, companies like ASUS, Gigabyte, MSI, Sapphire, etc have all failed to release cards in excess of 950Mhz. We are told even 925Mhz has become an issue with the latest batches of chips. Supposedly, the HD 5850 clocks are even worse which has forced some board partners to cancel certain SKUs.

I havn't seen anyone mention anywhere that they have had a problem clocking a 5870 at 925Mhz seems hard to believe from the general results reported at only a modest voltage bump. Perhaps ASUS, Gigabyte, MSI, Sapphire have more stringent TDP guidelines than Nvidia used for the fermi.

Originally Posted by SKYMTL

The don't show a small difference. From what I see from the results, most reviewers failed to enable DX11 properly in both Dirt 2 and BF: BC2. Their results line up almost perfectly with some preliminary DX9 runs I did at the beginning of the benchmarking process.

Failed to enable it on ATI or Nvidia? It's well known that the Nvidia card was running in DirectX 9 mode on the Dirt 2 demo but the ATI cards havn't had any problem working in Directx 11 mode on both the demo and the full game. Guru3D used the full game to test and had results very different from this site.

Originally Posted by SKYMTL

I don't see where the issue is with that. The GTX 470 in particular provides better performance per dollar than the HD 5850. That to me is worth waiting for when it comes to certain people's cases.

Possibly if it was selling for the same price and only then if you don't include electricity bills. As it stands now the 5850 wins at both performance per dollar and performance per watt according to pretty much every other review in the world.

[thorn]

Originally Posted by belgolas

Who gives a hoot about power consumtion for the fastest card? Of coarse it will consume more power. Do you think a F1 car will be more fuel effiecent than a Honda civic?

You know what, at the rate the F1 is going that might not be so far fetched.

Originally Posted by belgolas

Also your arguement about tesselation and nvidia's card not knowing if it could do that plus high graphics is a dumb arguement. Look at metro 2033 one of the BEST looking games in VISUALS and has tesselation and nvidia DOMINATES it.

Well I guess it depends on what numbers you are looking at. In percentages sure you can state that in Metro 2033 at 1920x1200 that the GTX 480 provides a 27.26% increase in performance but that big number represents an increase of 6.82 fps on average. And I honestly doubt even the GTX 480 with those settings is playable (at the end of the day the whole point is to be able to play the game is it not?). And are you going to pay $75 for ~7 extra fps?

To me throwing around percentages alone is meaningless they need to be put in context, you can make stats say whatever you want them to say. I think reviewers should start adding standard deviations along with min, max, avg (maybe even mode) especially in the cases where there is no fps graph.

It is clear, for the majority of cases anyways, that the GTX 4xx series outperforms the HD 58xx series. I don't think anyone would contest that. However based on price differences you can't say there is a clear winner. It really comes down to what games you play (or folding, whatever you chose to do with your hardware) and the resolution, you can't use blanket statements.

[SKYMTL]

Originally Posted by sideeffect

Guru3D used the full game to test and had results very different from this site

No, they aren't. Where we used the same settings, their results line up nearly perfectly with ours.