Request to Anandtech staff after the GTX580 review.

[Ph0b0s]

Put this in the review comments but they are 12 pages long, so not sure if it will be noticed...

Now that it is getting to nearly two years since I brought my graphics cards, a pair of GTX285's, it is starting to get towards the time I am thinking about maybe upgrading to DX11. So I am starting to take a lot of interest in the reviews about AMD and Nvidia's end of year card releases. The problem is that I cannot tell how much faster this new crop of cards are than my current setup, as for the games I am playing my cards are not listed in the benches. This is a trend that seems to be happening everywhere.

I understand the reason is that my games are DX11 and my card is DX10, but my card still runs them and I would like to know how they compare even if one is running DX10 and the other running DX11. I am aware that it not a 100% fair comparison, but that is how people in my position work out if the upgrade is worth it or not....

I can understand if this is not done as part of a review, but even Anandtech's chart system gives no measure for my cards in these games . Please sort this out. Just because a card does not run the latest version of directx does not mean it should be forgotten. Especially since the people most likely to be looking at upgrading are those with my generation of hardware rather than people with DX 11 hardware. Cards should only be forgotten when they can no longer run a games at whatever version of directx at above 30FPS at the standard review resolutions. This not the case with my current setup yet. Else the upgrade decision would be a no-brainer.

Anyway that's my 1 cents worth...

 

[happy medium]

I would wait a little and buy 2 gtx570's, unless you have the psu,cpu and money for 2 gtx 580's?
I would not make a MAJOR video card puchase for the next 4 to 6 weeks.

Whats the rest of your system like? and at what resolution?

your gtx285's in sli are about as fast as 2 5850's in crossfire.
Mabe a tad slower 10/15%

 

[Arkadrel]

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_580/14.html

TPU includes a 285 in their review. 1920x1200:

Hawk: 285= 80.4 fps___vs____580=172.9 fps
Farcry2: 285= 58.1 fps___vs____580=119.4 fps


not all games are in dx9/10 anymore... anyways look at the review on their site.

if you can use crossfire: 6850x2 is faster than 1x580, 2x460 is about same as 1x580.

 

[Nintendesert]

http://www.anandtech.com/bench/GPU/


Once the 580 gets added to this you should be able to compare the cards pretty decently.

 

[PlasmaBomb]

Put this in the review comments but they are 12 pages long, so not sure if it will be noticed...

Now that it is getting to nearly two years since I brought my graphics cards, a pair of GTX285's, it is starting to get towards the time I am thinking about maybe upgrading to DX11. So I am starting to take a lot of interest in the reviews about AMD and Nvidia's end of year card releases. The problem is that I cannot tell how much faster this new crop of cards are than my current setup, as for the games I am playing my cards are not listed in the benches. This is a trend that seems to be happening everywhere.

I understand the reason is that my games are DX11 and my card is DX10, but my card still runs them and I would like to know how they compare even if one is running DX10 and the other running DX11. I am aware that it not a 100% fair comparison, but that is how people in my position work out if the upgrade is worth it or not....

I can understand if this is not done as part of a review, but even Anandtech's chart system gives no measure for my cards in these games . Please sort this out. Just because a card does not run the latest version of directx does not mean it should be forgotten. Especially since the people most likely to be looking at upgrading are those with my generation of hardware rather than people with DX 11 hardware. Cards should only be forgotten when they can no longer run a games at whatever version of directx at above 30FPS at the standard review resolutions. This not the case with my current setup yet. Else the upgrade decision would be a no-brainer.

Anyway that's my 1 cents worth...

Granted the 580GTX isn't there at the minute... but GTX 285 SLI vs GTX
480

http://www.anandtech.com/bench/Product/167?vs=158

You can also fiddle about with the Browse GPU Benchmarks

http://www.anandtech.com/bench/GPU/88

Unless you are going SLI again, there isn't much improvement going from SLI 285's to a 480 (and possibly a 580).

Minimum frame rates look better and power consumption would drop if that is a concern for you...

 

[Ph0b0s]

Thanks for the responses. My main point was that for the games I am playing at the moment BFBC2, stalker and Dirt2, all DX11 games and part of Anandtech's bench suite. I cannot tell how my SLI or even one of my cards stacks up against all these great options, as it is not listed, even in Anandtech's new bench database for these games. I know there are DX9 games there, but I would rather base my desicion on the games I will be playing.

Obviously I will take no final decisions until ATI launch their response, but just wanted to ask the review makers at Anandtech to please make my comparing a bit easier by having my card and others from the DX10 generation on the benchies for the games I am using in their test suite. Even if it is in a second article or in the bench system. Just want a graph for BFBC2, etc with the recent card FPS and my cards FPS even if one is running DX10 and the more recent ones DX11. Hopefully not too much to ask...

 

[Ryan Smith]

Hi Ph0b0s;

I can certainly appreciate your dilemma. I am going to have to decline your request, but please give me a moment to explain why.

We approach benchmarking from what's fundamentally the scientific approach. That is to say it's an apples-to-apples comparison where the only thing we change are the video card and the drivers. This presents the clearest data to our readers (you guys) as they can be confident the difference in performance is solely due to the card in question.

What you're requesting would compromise that kind of testing method, as the older cards would not be running with the same settings. This is actually more problematic than it sounds too, as most DX11 settings end up making a game slower - tessellation, contact hardening shadows, etc all require extra GPU time to setup and render. In fact the only feature that makes things faster is the threading changes, and Civ5 is the only game we test that even tries to make aggressive use of that feature. The end result would be that your GTX 285 running these games in DX10 mode would actually look pretty fast compared to all the other cards running in DX11 mode.

This is why our current benchmarking suite is a compromise, so that we don't end up with only DX11 games. More than half of the games we benchmark are still DX9/10/OpenGL, and while this won't tell you GTX 285 performance for the specific games you're interested in, I would not discount it as a general guideline of how your card compares to newer cards.

If this isn't the kind of answer you want I can certainly understand, but at the end of the day it's our goal to uphold the scientific accuracy of our results first and foremost, and that means we can't include DX10 cards in DX11 games.

-Thanks
Ryan Smith

 

[Scali]

In fact the only feature that makes things faster is the threading changes, and Civ5 is the only game we test that even tries to make aggressive use of that feature.

The multithreading in DX11 is also supported on downlevel hardware (DX9 and DX10), so older cards will benefit from this as well in DX11 games.

 

[Ph0b0s]

Ryan, thank you for responding. I can certainly understand your position. It is the reviewing acumen that keeps me coming back to this site for reviews and articles.

But I have to disagree with the only apples to apples comparison. In the lab this makes sense and keeps things as you say scientific. But outside the lab environment users are making apples to pears comparisons every time they purchase a new graphics with a new version for DirectX and look at their 'new' frame-rates.

Users should be informed that with a change of DX the newer cards may not seem as fast as you would expect. This has been the case since I first started buy graphics accelerators back when they arrives with the 3DFX cards. I almost automatically take this into account, but newer users should be briefed. This should not be a hurdle.

The other thing I take issue about is that in the GTX580 article, out of the 10 FPS benchmarks only 4 of them allowed any comparison between DX10 generation cards and the latest DX11 ones. And out of the 4, 3 were DX9. For some reason for the Battleforge DX10 bench only the 4870 appeared? To keep it somewhat brief I won't go into why only having DX9 as a comparison is not much use and why Crysis as the only DX10 titles is not so helpful as it is a law unto itself as a benchmark.

Now I understand why you would want to keep as small an amount of cards to test with as possible since you need to hit deadlines etc. But after the launch of a series of new cards, as is happening now, has died down. I would not have though it would be problem to do an article with older cards considered for the 6 titles they are not currently, with the provisos above about DX differences. Or just put them in the bench system with a note saying they were running DX 10 for the FPS reading. It would be interesting to see now that we have more DX11 titles that some argue do now show much of a visual improvement, how much of a performance hit DX11 gives. If it is a lot maybe some will smell a rat.

At the end of the day a site like this is here to allow users to easily workout if an upgrade is going to be worth it or not. But for users of Nvida 2xx and AMD 4xxx cards this is now not the case and I still believe that these are the users who at this point are more likely to be thinking about upgrading.

 

[Gryz]

But for users of Nvida 2xx and AMD 4xxx cards this is now not the case and I still believe that these are the users who at this point are more likely to be thinking about upgrading.

I agree. I upgrade individual components of my system when I think the upgrade would be good price/performance wise. For videocards that means I wait until I can get a card that gives me double the framerates of my current card. Price and heat/power/noise are a factor too. That's why I haven't bought an upgrade for my GTX260 yet. GTX480/5970 were too loud, warm and expensive. GTX460/5850 didn't give enough of a framerate improvement.

The GTX580 looks better. Still expensive, but at least my framerates would go up 100-150%. That will make it worth the money. I will wait to see performance of the 6950/6970.

Am I the only one in this situation ? Is it normal for GTX200 series owners to have waited upgrading their videocard until now ? I think so. I think the people who replace videocards for 20-30% improvements are the minority. Even on this forum. Even amongst enthousiasts.

But the GTX260 is slowly dropping off the charts. So when I upgrade it, I don't know how much an improvement it will be exactly. And that is part of the fun of upgrading. And part of the fun of reading Anandtech. The fun of knowing exactly what you did, knowing you made a smart choice. Knowing waiting 2 years was worthwhile. If I can't see exact numbers, what's the point of having detailed benchmarks ? So I can compare 10 cards I don't own ?

I understand Ryan's point. But maybe there's a way to get numbers that are still relevant for people like the OP and me.

 

[vshin]

My feeling is that these comparisons are not done because it will show very little difference and that older cards will perform well enough to make a case against upgrading.

Everytime a new generation of cards come out, I'm tempted to upgrade but once the dust settles, I notice that it won't give me a tangible increase in performance over my old card. So I keep waiting until the next generation of cards as well as games that will actually provide something worthy to upgrade for. Unfortunately I don't see this happening until the next generation of consoles come out (probably in another 2-3 years).

Right now I use a single HD4870 and play games at 1680x1050 in DX10 vsynch on and AA off. Even with brand new games like Black Ops, I can achieve a buttery smooth framerate without lowering graphics settings below default. So my guess is that if Anand reviews cards for gamers like us, with our graphic settings, the chart is going to show every card running the game at 60fps with no discernable difference between an old card like my 4870 and the brand new 580.

 

[darckhart]

at work, people ask me to test things for them too, and i tell them that if their part can perform at my testing conditions, i don't mind throwing it in the batch becaues it'll be basically no work on my end. but if their part cannot perform at my settings, i have too much work already to make exceptions. in an official review here, i would expect the same.

the best thing is to do what you've done. come to the forums and ask if someone else had a config similar to yours and already upgraded. they can share info. the other is to just do a rough estimate. for example, while i have the "current" cards, the benches in the reviews rarely ever use the settings i use so i already have to guess at where my cards fall, and where the performance that i want needs to be.

 

[Rhonen]

I would point out to the OP, that testing the newer cards against older cards is difficult because the CPU platform could also be a factor. A newer card may be bottlenecked by an older CPU, making the GPU seem less of a worthwhile upgrade than if it were paired with a newer CPU.

So if a site benchmarked the newer cards with older ones, they'd most likely do so on the newest CPU. If you had an older CPU, you would not see the same results.

 

[Seero]

Ph0b0s

I understand exactly what you mean as I'm on a similar situation where I want to upgrade but not sure if I will end up side-grading it. There is a disjoin between Dx10 gear vs Dx11 gear. Reviews stated that 460 has similar performance compare to 285, but after all those upgrade, I am not sure if it is a side-grade. If you assume that 460 is roughly 285+Dx11, then 580 is roughly 2x the performance of a single 285. Also, Fermi scales better, that means 580 > 2x285 in terms of raw performance.

Again, the disjoin is still here and I don't have solid evidence to support the claims above. What I can say for sure is Fermi supports Dx11, 285 doesn't. If you really want to upgrade now, go for 460SLI as you can't beat that price. Note that 460SLI scale better than 285SLI. If you can wait, 570/560SLI will probably have better power consumption and a bit faster. Of course, grabbing a 580 and pair it up later is also an option, but I will wait a week for AMD first and see if there will be a deal or not.

One of the problem about review is they don't bring back those old guns and test it with new games, so those who are using old guns don't really know where they are at. While I can see the difficulties finding old guns to benchmarks, I agree that it will generate more hits as this is what really viewers wanted to see.

 

[bryanW1995]

I agree. I upgrade individual components of my system when I think the upgrade would be good price/performance wise. For videocards that means I wait until I can get a card that gives me double the framerates of my current card. Price and heat/power/noise are a factor too. That's why I haven't bought an upgrade for my GTX260 yet. GTX480/5970 were too loud, warm and expensive. GTX460/5850 didn't give enough of a framerate improvement.

The GTX580 looks better. Still expensive, but at least my framerates would go up 100-150%. That will make it worth the money. I will wait to see performance of the 6950/6970.

Am I the only one in this situation ? Is it normal for GTX200 series owners to have waited upgrading their videocard until now ? I think so. I think the people who replace videocards for 20-30% improvements are the minority. Even on this forum. Even amongst enthousiasts.

But the GTX260 is slowly dropping off the charts. So when I upgrade it, I don't know how much an improvement it will be exactly. And that is part of the fun of upgrading. And part of the fun of reading Anandtech. The fun of knowing exactly what you did, knowing you made a smart choice. Knowing waiting 2 years was worthwhile. If I can't see exact numbers, what's the point of having detailed benchmarks ? So I can compare 10 cards I don't own ?

I understand Ryan's point. But maybe there's a way to get numbers that are still relevant for people like the OP and me.

the key is to know where your card stands in relation to other cards that are being benched. for example, I have a couple of gtx 260's lying around in some of my spare computers. when i upgraded to a gtx 460 768 I did a ton of research and discovered that I was only getting ~ 25 % upgrade. however, I didn't care so I did the upgrade. anandtech bench did actually have both of my cards in there, however. if they didn't then I knew that my card was faster than 5770, about the same as 5830, and slower than 4890/gtx 275.

of course, all of that crap went out the window when civ 5 came out and actually implemented some previously unused dx 11 features that favored dx11 cards greatly. Civ is my primary game now and will probably remain in my stable for years to come. I guess you could say that I chose the wooden cup made by a carpenter's son, though only because I liked the filigree around the edges

edit: @ seero, gtx 460 is definitely not an upgrade for 280/285. well, unless you bought a 285 to run at 1680x1050. at higher res the older cards tend to pull away. In fact, bfg10k wrote a very informative series of articles about this after fermi came out, iirc he was getting about the same fps with a gtx 285 as he was with a gtx 470 because he puts so much stress on the cards. nvidia's dumbed-down amd clones in the midrange can't handle high end graphics requirements any better than amd's midrange cards.

 

[RussianSensation]

Ph0b0s, to put it simply, if you are not gaming at 2560x1600 with 8AA, from a performance perspective, there is no single videocard upgrade worth spending your $$ on from GTX285 SLI setup. You will get better visuals with DX11 though (whether that's worth the upgrade is up to you). However, if you seek a worthwhile performance upgrade, HD6870 CF, GTX580 SLI are right up your alley.

A single GTX285 will around GTX465/460 performance +/- depending on the game. So if you scale that by even 1.6x for dual cards, it's obvious that GTX580 is not much of an upgrade for dual 285s.

 

[Vdubchaos]

I don't see a point of putting up benchmarks to older titles.

What's the difference if one card Averages 60fps and other does 70fps......they are both playable.