My Titan Review - 59 games tested!

[BFG10K]

BFG, I've wondered why you don't overclock. I can understand why you wouldn't overclock for 10% gains but leaving a 2500k at stock seems criminal to me. What's your reasoning?

100% stable and correct operation is very important to me, and overclocking adds another possible point of failure. After all the tinkering/testing is done, I just want to sit down and enjoy my games without hassles.

Also overclocking the i5 would make little difference in these tests as the vast majority were at 95+% GPU loads.

 

[lavaheadache]

100% stable and correct operation is very important to me, and overclocking adds another possible point of failure. After all the tinkering/testing is done, I just want to sit down and enjoy my games without hassles.

Also overclocking the i5 would make little difference in these tests as the vast majority were at 95+% GPU loads.

Fair enough. I'm a risk taker though and do it in the name of hobby

 

[f1sherman]

Crysis games, Metro and STALKER CS should have been tested with AA. 90% of everything else is so ancient, the Titan is completely wasted on those games. Throwing SSAA at 5-10 year old games won't fix their horrible graphics by today's standards. I would have preferred a list of modern games to be honest.

yeah well... we see that only in any other review on the internetz

You can't be serious comp-lain-ing about 59 games being tested with SSAA/TrSS, and few games tested with no-AA is obviously a conscious decision.
This is one helluva gaming suite and great effort from BFG. It has almost theoretical value

 

[njdevilsfan87]

Nice work! My only suggestion : post GTX boost clocks. Both the Titan and 680 have varying boost clocks per card. I've seen Titan boost clocks reported as low as 967mhz and as high as 1058mhz. Same deal on 680s.

 

[SirPauly]

Crysis games, Metro and STALKER CS should have been tested with AA. 90% of everything else is so ancient, the Titan is completely wasted on those games. Throwing SSAA at 5-10 year old games won't fix their horrible graphics by today's standards. I would have preferred a list of modern games to be honest.

There are countless reviews with modern titles and refreshing to see how older titles scale and perform on modern hardware -- also nice to see that older titles run well and not ignored by companies.

Kudos to BFG10K

You wanna now why one needs to investigate older titles as well:

http://alienbabeltech.com/main/gtx470-performance-test-part-1-windows-xp/

Outstanding article by BFG!

Gamers also have favorite titles -- a library of titles -- decide to play an older title they might of missed!

 

[tential]

Please read the first paragraph I've added to the OP and reconsider replaying them. In the vast majority of situations those games will run much faster. It's the toughest areas that I test because putting up 400FPS on a Titan and "only" 275FPS on a GTX680 isn't terribly useful given both present a solved problem.

I do have additional (typical performance) benchmarks for older games, so if you really want to see "extreme" FPS scores at more common settings (e.g. 4xMSAA), I can put up a few to set your mind at ease.

I love replaying old games and I'd hate to put off someone else from doing the same just because my extreme tests scared them off. :thumbsup:

Well, thanks a lot! Haha, good thing I checked back. I'll have to get some older games now and play them since I missed a lot of top games simply because I refuse to play single player games by myself haha.

Edit: Seriously surprised to see someone test 59 games and have people complain about the review.... I'm sure you can find out how Titan performs on current games in almost any review you found. Great work OP.

 

[socialdarwinist]

If only I could make myself pull the trigger on a Titan...had it in my cart so many different times and could never go through with it.

 

[24601]

100% stable and correct operation is very important to me, and overclocking adds another possible point of failure. After all the tinkering/testing is done, I just want to sit down and enjoy my games without hassles.

Also overclocking the i5 would make little difference in these tests as the vast majority were at 95+% GPU loads.

Heretic :O

If I had the extra money on hand to do so, I would be using an IPS 1440p 120hz panel which would make 70-90 minimum frame rate important.

I have always found that CPU is the most annoying bottleneck since GPU limits seem to be less dynamic range between minimum and average frame rates.

 

[BFG10K]

Nobody has found the non-FPS in the list yet? Come on people, it sticks out like a sore thumb.

Nice work! My only suggestion : post GTX boost clocks. Both the Titan and 680 have varying boost clocks per card. I've seen Titan boost clocks reported as low as 967mhz and as high as 1058mhz. Same deal on 680s.

967MHz for the Titan. I didn't check the GTX680 because I didn't realize people wanted to know this information, but I'll remember for next time. :thumbsup:

 

[Magic Carpet]

BFG10K, thanks for the hard work. Very nice indeed.

Nobody has found the non-FPS in the list yet? Come on people, it sticks out like a sore thumb.

I have hardly played a third of the games, you brought in the comparison. What is it ?

 

[Jaydip]

Nobody has found the non-FPS in the list yet? Come on people, it sticks out like a sore thumb.


967MHz for the Titan. I didn't check the GTX680 because I didn't realize people wanted to know this information, but I'll remember for next time. :thumbsup:

Fallout 3, already said that in post1.

 

[BFG10K]

Fallout 3, already said that in post1.

Sorry man, that's not it.

When you said "Fallout 3" in post#1, I thought you were pleased to see it in the results, LOL. :awe:

 

[24601]

Sorry man, that's not it.

When you said "Fallout 3" in post#1, I thought you were pleased to see it in the results, LOL. :awe:

Tomb Raider Underworld?

All of those games are shooters though, and some of them are third person, therefore not "FPS"

 

[BFG10K]

Tomb Raider Underworld?

A beer and a cookie for you, my friend. :awe:

 

[Jaydip]

Sorry man, that's not it.

When you said "Fallout 3" in post#1, I thought you were pleased to see it in the results, LOL. :awe:

I want that beer and cookie
http://en.wikipedia.org/wiki/Fallout_3

 

[Phynaz]

Wow. 59 games, awesome!

 

[BallaTheFeared]

I think Jaydip and Communism should have a fight to the death for the beer and cookies.

 

[Jaydip]

I think Jaydip and Communism should have a fight to the death for the beer and cookies.

Amen

 

[RussianSensation]

Did you see the scores? In the tough areas those games run too slow on a GTX680 to use MSAA. It's possible a Titan might be able to bump Crysis to 2xTrSS from FXAA, but I need to test it more.

Isn't that the point? Don't you want to see that you can push AA with the Titan in those games and 680 is a slideshow?

Heh. Is that why Serious Sam 2 and MoH-PA showed the biggest performance gains in actual gameplay, at the actual settings I play them at? Also preliminary testing shows Titan is fast enough to run MoHPA at 4xSSAA, so that's a significant IQ upgrade to what I had before, especially in areas with shader aliasing.

This is where you and I differ. Even 1 trillion AA samples of the highest quality cannot make old games look good for me. I also do no understand how going from 75 fps to 124 fps in SS2 or going 103 fps to 167 fps in MOH: PA is an upgrade from a playability point of view? And SSAA in those games won't hide the fact that they are very ugly. But as long as you are satisfied with the upgrade for your IQ standards, that's what matters. :thumbsup:

 

[BFG10K]

Isn't that the point? Don't you want to see that you can push AA with the Titan in those games and 680 is a slideshow?

No, the point is to attain the best balance of performance/IQ. If Titan has enough extra performance then the IQ can be raised, certainly. In other cases I just take the extra performance, like in the Stalker games which really need it in the heavy swamp/forest areas.

Both boards were @ 99% GPU load anyway; what's the point in loading them further just to make a slideshow?

I also do no understand how going from 75 fps to 124 fps in SS2 or going 103 fps to 167 fps in MOH: PA is an upgrade from a playability point of view?

In the case of Serious Sam 2, the new Titan setting is 4xTrSS @ 88.50FPS. Vegetation is much sharper and crisper and it still runs faster than the GTX680 did at the inferior settings.

In the case of MoHPA, the new Titan setting is 4xSSAA @ 91.05 FPS. Slightly lower performance than the GX680 but justified given SSAA does a much better job cleaning up everything compared to the GTX680's inferior settings.

And SSAA in those games won't hide the fact that they are very ugly.

For a nine year old game, Pacific Assault holds up impressively well:

 

[ArchAngel777]

A beer and a cookie for you, my friend. :awe:

GD YOU! Now I am hungry for a giant cookie and the day just started. It does not bode well for me...

 

[Zanovar]

Great stuff,tips hat.

 

[tidy]

.

 

[BFG10K]

You get 65fps at 2560x1600 in Crysis? May I ask where you took the reading? Not in the shipyard area I suspect, that takes a royal dump on the CPU and will surely ruin my 60fps plans for Crysis once I get my 'sufficient' GPU.

I've built a custom benchmark in the ice/alien area. It's far more demanding than the harbour area, or anywhere else in the game for that matter.

I also run custom game settings as maximum levels would slideshow even a Titan.

I suppose at 60Hz old UT2004's CPU dependency probably doesn't cause you any grief, but I fear I won't be having 120fps onslaught matches until 2020 at this rate. I apply the same logic as you when it comes to overclocking, but sometimes I think I should get another 2600k and overclock it with two cores disabled just for that damned game. Unreal II is a menace too, what with it's dynamic shadows. Everything was going great until I touched those sliders.

You can't have had many bots going in the UT test, that's another member of the cripple thread brigade. IIRC last time I tried it on the 2600k, 20 bots in facing worlds already destroyed my 120fps target. Not that the engine can tolerate 120fps without glitching anyway...

I just tried Facing Worlds in UT99 with 16 bots (game maximum) and it ran at a constant 120 FPS.

In UT2004 I loaded 20 bots and it never dropped below 90 FPS, which is perfectly fine given the map recommendation is 6-12 players. Sure, you could load even more bots, but that just becomes a synthetic CPU test.

As for Unreal 2, I use maximum shadow levels (except blur which I don't like because it makes the shadows harder to see), and the game mostly runs at a constant 120FPS at my settings. My stress area benchmark is the worst-case framerate I ever see.

EDIT: I have another question. I'm using 306.23 drivers on the 580, and every time I try a newer driver I end up rolling back because F.E.A.R (among others) start exhibiting some strange behavior. I can see the fraps jumping a frame above/below my V-Sync cap and feel hitching. When I get my Titan or GTX 780/s, a driver update will be inevitable. Did you notice any hitching of any sort in FEAR with the new drivers?

I don't use vsync so I can't comment there, but the Fear 1 works and looks great @ 4xSSAA on my system.

 

[tidy]

.