Getting ready for benches. Need a little advice.

[3DVagabond]

Hopefully these 2 cards being compared are also comparable in price? If one card is 50% more expensive it's a monumental waste of time.

The problem with running a couple of benches OC is that they could be benches that heavily favor one card over the other.

 

[MarcVenice]

I'd run as many games as you could, at 1920x1200. The difference between 1920x1200 and 1680x1050 is 'just' 30%. At 1680x1050 certain bottlenecks won't suddenly dissapear (the gtx 480 will most likely have massive memory bandwith). The difference between 1920x1200 and 2560x1600 is 80%. Personaly, I'm benching more games then I usually do *mix of dx9 and dx11* at 1920x1200 and 2560x1600, and just skipping 1680x1050 altogether.

With 2560x1600 not being an option, I'd just stick with 1920x1200.

 

[dguy6789]

1680x1050 is by far the most common resolution. I don't think omitting it is very wise. If one is worried about it not being demanding enough, just crank the AA up to 24x or use super sampling. It's not like everyone who buys a 5870 or GTX 480 has a 30 inch screen. There's a very good reason for not doing so. Someone who buys one of these cards and runs it on their existing 22 inch screen would be able to game with all settings maxed out for 2-3x as long as they would be able to on a 24 or 30 inch screen.

 

[MarcVenice]

1680x1050 is by far the most common resolution. I don't think omitting it is very wise. If one is worried about it not being demanding enough, just crank the AA up to 24x. It's not like everyone who buys a 5870 or GTX 480 has a 30 inch screen. There's a very good reason for not doing so. Someone who buys one of these cards and runs it on their existing 22 inch screen would be able to game with all settings maxed out for 2-3x as long as they would be able to on a 24 or 30 inch screen.

You can allready play 99% of the games maxed out at 1920x1200, so you sure as hell can on 1680x1050. But you might be right, I'd prolly have to bench the GTX 470 as well, and at 1680x1050, and can't leave out a GTX 480 then. For Keys it will take a long time though, for me it's just another day at work

 

[Keysplayr]

You can allready play 99% of the games maxed out at 1920x1200, so you sure as hell can on 1680x1050. But you might be right, I'd prolly have to bench the GTX 470 as well, and at 1680x1050, and can't leave out a GTX 480 then. For Keys it will take a long time though, for me it's just another day at work

Probably because I'll take the time to do it right.

I'm kidding Marc....

 

[tviceman]

Sounds excellent. Can you make some suggestions? Much appreciated.

As far as 3 resolutions go, keep in mind that the highest I could run right now is 19x12. I'm pretty sure that would accomodate most gamers.

So, I would run:

1920x1200
1680x1050
1440x900

I'd run out and get a 3007 Dell, but I'm not THAT wealthy.

If you are benchmarking the two top end single gpu cards from both competitors - 1440x900 is pointless IMO. I think it would make much more sense to do 1680x1050 and 1920x1200 - with no anti aliasing, 4X aliasing, and 8X aliasing. The different levels of AA is important, IMO, because it can show bottle necks in the GPU's and give everyone a more clear indication of the highest playable usable settings for each GPU. EDIT: Other people have already stated 1440x900 is a waste of time...

Once again, since both competitor's cards are "DX11" GPUs, it would make the most sense to include as many DX11 titles as possible.

 

[jvroig]

If I may comment about it, you are certainly showing a lot of restraint, Keysplayr. Doing the benches correctly will no doubt help in your "credibility" woes (if that is the right term; I am a non-native English speaker) as I often see people jumping you for being a "fermi fan" or some such nonsense.

And while I understand your concern or motivation to do this right, I hope you are also expecting that unless your bench results show the 5870 beating the GTX285 (or GTX470? what are you testing again?) all the time and with power to spare, I doubt that those who normally jump you will refrain from jumping at you. You can't please everyone, after all.

I am interested in your results, how long do you think this will take? 2-3 days maybe? Maybe longer, I assume you also have a real career outside of this hobby? (Unless it is your job to review hardware). I am still using a 4770, so I am currently in the position of "upgradeable", and I am of neutral color (I had an 8600GTS before my 4770, that card was hated by most but it served me well, anyway), and am in no hurry to upgrade.

 

[1h4x4s3x]

A question out of curiosity. I am new here so bear with me.
Who are you resp. for whom are you doing these tests? Are you working for anandtech or is it just a private thing?

 

[Keysplayr]

A question out of curiosity. I am new here so bear with me.
Who are you resp. for whom are you doing these tests? Are you working for anandtech or is it just a private thing?

Who am I resp.? Not sure what that means.
I am doing these tests, as I have in the past, for all of us. Folks seem to enjoy it, and I enjoy doing it.

I don't work for Anandtech. But I am a moderator as well as a member.

What magazine do you write for? Hey, if I'm being interviewed, I should at least know that!

 

[Madcatatlas]

haha! well played!

 

[cbn]

Battleforge is DX11.

Can you get a timed benchmark for that?

 

[1h4x4s3x]

What magazine do you write for? Hey, if I'm being interviewed, I should at least know that!

I'd like to say Playboy but in all honesty, it's just my mom, she really likes you. :awe:

 

[crazylegs]

I wouldn't waste your time running 1440x900 benches. As others hafve suggested run the max resolution you can, with more combinations of added eye candy.

Anyone who buys a top of the range GPX card and runs it at 1440x900 needs their head examining - i'm not denying that some people may do this - but they dont deserve you spending your time deciding which card will pull 300 FPS and which will pull 301 FPS...

Look forward to seeing your review - as well as all the others once Fermi arrives. In Battlefield Bad Company 2, I may have found a game good enough to tempt me into upgrading my HD4890

 

[Keysplayr]

I wouldn't waste your time running 1440x900 benches. As others hafve suggested run the max resolution you can, with more combinations of added eye candy.

Anyone who buys a top of the range GPX card and runs it at 1440x900 needs their head examining - i'm not denying that some people may do this - but they dont deserve you spending your time deciding which card will pull 300 FPS and which will pull 301 FPS...

Look forward to seeing your review - as well as all the others once Fermi arrives. In Battlefield Bad Company 2, I may have found a game good enough to tempt me into upgrading my HD4890

K. No 14x9. I agree.

 

[3DVagabond]

Can you do openGL testing? Possibly checking frame rates in 3D aps.? Not everyone can afford or has easy availability to pro cards. Most applications offer free 30day full functioning demos if you don't have access to full versions of them.

 

[WelshBloke]

I'd like to say Playboy but in all honesty, it's just my mom, she really likes you. :awe:

pics of mum?

 

[SolMiester]

I think while 16x10 & 19 x 12(10) are the most common resolutions surely with card this powerful, only 25 x 16 will separate the men from the boys.....even the 5850 is enough for the 19 x 12

 

[BTRY B 529th FA BN]

Can you include Unreal Tournament III & Avatar?

10.2 drivers will still crash with XP Pro (latest updates)

This is a very very good idea!

 

[v8envy]

Ok, show of hands. Who here is going to fork over $600-$1200 for a video hardware to run a sub-$100 monitor?

1920x1200 is the lowest meaningful resolution for testing of enthusiast cards. Some people *MIGHT* be interested in the frame rates these beasts can manage on the cheapest monitors available at wal-mart, but those aren't the people buying this hardware.

High res and multiple monitor benchies, please. As many games and 3d apps as you can get your hands on. Also, I believe you can set a higher resolution than your monitor can display and render to that if you run the games in windowed mode. You'll have to pan around the desktop to see the whole thing. Check "display modes this monitor can't support." -- you may be able to run benchies rendering to the 30" dell resolution even on your 24".

 

[dguy6789]

Ok, show of hands. Who here is going to fork over $600-$1200 for a video hardware to run a sub-$100 monitor?.

$600-1200? We're talking about a 5870 and maybe a GTX 480, not one to two 5970s. Completely different ballgame. Some people are perfectly happy with a 22 inch screen. Maybe they just want more fps. They do make 22 inch IPS panels in case you guys didn't know.

 

[v8envy]

Do you really think one of the 5000-odd GTX480s will be below $600 on release? We already know 5870 performance at 16x10 -- it's sufficient even for Crysis.

 

[dguy6789]

If the GTX 480 is anywhere near $600, it is a complete and utter failure so no, I don't think it will be at or above $600. I am expecting the GTX 480 price to be very close to the 5870 price. I don't think anybody expects the GTX 480 to be as fast as a 5970, so releasing it at the same pricing level as the 5970 would be suicide.

 

[v8envy]

You are assuming NV plans to compete with a very, very limited edition card based on price/performance. ATI's marketing puts a $40 premium on DX11 + eyefinity compared to their last generation when talking 5830 vs 4890. Would DX11, PhysX, 3D and single GPU not be worth at least $300 more than a 5870 or $100 more than 295 to the right buyer? High end gaming gear is rarely about value or good bang for buck.

There's only a few of those bad boys being made. There's ALWAYS a market for a few items at the extreme right edge of the price/demand curve.

 

[dguy6789]

I don't think anyone cares about "single gpu" or physX more than they care about performance. Especially not at this level. At this level, they are spending all of this money for performance at any cost. If the GTX 480 is slower than the 5970 but is the same price as the 5970, I don't think anyone at all would buy a GTX 480.

 

[v8envy]

If the criteria is highest performance at any cost you've already disqualified the GTX480. The $1200 (estimated) Mars or even the $1200 Ares card from Asus should handily out-bench it.

The market for this card is people with a 30" monitor and desire to use all possible gaming settings on that monitor. For that, the same price or higher price than a faster ATI card which doesn't do PhysX is eminently reasonable. Which is why I said low res benches make no sense and 1900x1200 is the absolute minimum to test with.