Radeon HD 7950 - 1080P Performance in Crysis 3, Tomb Raider & Metro LL

[cmdrdredd]

Very High SMAA 2x 1080p SLI
Frames: 5964 - Time: 83891ms - Avg: 71.092 - Min: 46 - Max: 103

It's interesting how the fps can vary so much with different settings and even turning differently from run to run.

 

[BallaTheFeared]

Nice - I'm trying to grow more back hair.. I ran the welcome to the jungle test at very high as well so that should do

http://www.youtube.com/watch?v=cemiLAxrUvk&feature=player_detailpage#t=24s

 

[Face2Face]

Very High SMAA 2x 1080p SLI
Frames: 5964 - Time: 83891ms - Avg: 71.092 - Min: 46 - Max: 103

It's interesting how the fps can vary so much with different settings and even turning differently from run to run.

Yes, they can vary. You have to keep the same path every run. I have tested and tested back to back runs and I am seeing very little variance. Maybe at the most a couple hundredths of a second.

http://www.youtube.com/watch?v=cemiLAxrUvk&feature=player_detailpage#t=24s

LOL! :biggrin:

 

[Face2Face]

Added a couple runs from Welcome to the Jungle with the same settings as above. I also added the Very High system spec as well.

A demo of the run can be seen HERE. I set FRAPS to record for 55 seconds. I stand at the last ledge of the stairs and walk per the video and then jump on top of the car and the benchmark ends.



This benchmark is showing a different result vs. The Post Human benchmark. From 800/1250 to 1250/1750 it's only yielding a 25% increase in performance. That's an increase of 56% for the core and 40% for the memory frequency. Pretty odd results, but I tested them multiple times to make sure they were correct.





This run was set with the system spec set to Very High and everything else remained the same. Were seeing a 45% increase in performance from 800/1250 to 1250/1750. Pretty nice when you can see 55 FPS with Very High enabled. Still visually, there is very little difference from the High system spec.

 

[BallaTheFeared]

Nice, tried to replicate, but probably failed.

Used my daily driver clocks, 4.8/4.8 on the i5, 2400 on the ram, 1050/1525 for the cards (ULPS enabled OC//zero core win) - no voltage control on the second card, no extended limit on clocks. This is the best my cards can do with the default voltage of the 8+6 pin card which is 1.18v core and 1.55v vram. When I say that I mean tested in over 50 hours of gameplay, this is not a suicide run for the voltage I'm using, completely artifact free and fully stable

(Very High)

Single @ 1050/1525 Out of the Jungle

Frames, Time (ms), Min, Max, Avg
2624, 55000, 35, 56, 47.709



CF @ 1050/1525 Out of the Jungle

Frames, Time (ms), Min, Max, Avg
3999, 55000, 56, 90, 72.709

 

[MrK6]

MrK6, do you think you could do a comparison? You have a 1300mhz 7970. This would be great to see how Crysis 3 scales with the extra shaders and textures of the 7970.

I would like to but I don't think it would be valid as I would have to very closely match Face2Face's run. I'll try if I have time but I doubt it as I have a tough schedule for the next few weeks. I think the 7970 should probably be 5% faster at most, and certainly not mine as I think my vRAM tops out in games at ~1825MHz, maybe less now that the card is over 18 months old.

 

[Face2Face]

Nice, tried to replicate, but probably failed.

Used my daily driver clocks, 4.8/4.8 on the i5, 2400 on the ram, 1050/1525 for the cards (ULPS enabled OC//zero core win) - no voltage control on the second card, no extended limit on clocks. This is the best my cards can do with the default voltage of the 8+6 pin card which is 1.18v core and 1.55v vram. When I say that I mean tested in over 50 hours of gameplay, this is not a suicide run for the voltage I'm using, completely artifact free and fully stable

(Very High)

Single @ 1050/1525 Out of the Jungle

Frames, Time (ms), Min, Max, Avg
2624, 55000, 35, 56, 47.709



CF @ 1050/1525 Out of the Jungle

Frames, Time (ms), Min, Max, Avg
3999, 55000, 56, 90, 72.709

Looking very nice You have a really nice setup - Your avg for a single card is about the same as mine that is clocked a little higher. I have my single 7950 Crysis 3 stable @ 1250/1750 @ 1.185v and 1.6v - Temps around 60-63C. I however cannot game for an extended amount of time with 1300/1850 - It just get's too hot, no artifacting, just too hot :$

BTW - I added some benchmarks for Tomb Raider as well.

 

[BallaTheFeared]

It's just run variance I believe, oddly I still beat you it seems in TR with equal clocks though

I wish I could control voltage with ULPS enabled, but you can't so it is what it is.

Difference between ULPS enabled vs disabled with at idle or browsing is around 20w, which isn't a lot but when you look at your kill-a-watt seeing 50s is nicer than seeing 70s :|

Also because I'm at 1080p, when I play a easy to run game like Dota 2 if I run it in borderless windowed mode CF won't kick in, and the second card will stay off. The impact on power usage here is a lot more, because the cards use a lot more power fully clocked up than they do when turned off, or even in low power states.

Kind of a pain, but it's nice when you can get it to do what you want it to.

 

[cmdrdredd]

For comparison in Tomb Raider a single GTX 670 gets me Min38 Avg57 Max80

Quite a bit slower than a 7950 with a good overclock. I've always said that AMD cards make more sense from a performance perspective if you are looking at a single card when compared to the GTX 600 series.

 

[Face2Face]

Updated with Metro Last Light benchmarks.

 

[bgt]

Nice OC! Mine goes quit high too.Could not get my mem to 1600.

 

[el etro]

Hey OP, do you have scores with your processor @stock and your GPU @boost(925Mhz GPU) on Post-Human and Root of all Evil at VHigH/VHigh 8xMSAA 1440x900?? I would thank you if you have.

- At those settings, mine 670 does 45FPS on Post Human and 29FPS on Root of all Evil.

 

[Face2Face]

Nice OC! Mine goes quit high too.Could not get my mem to 1600.

Very Nice :thumbsup:

Hey OP, do you have scores with your processor @stock and your GPU @boost(925Mhz GPU) on Post-Human and Root of all Evil at VHigH/VHigh 8xMSAA 1440x900?? I would thank you if you have.

- At those settings, mine 670 does 45FPS on Post Human and 29FPS on Root of all Evil.

I don't, but maybe I can do a run of Post Human for you. FYI - From what I can tell you will not see an improvement going from a 3570K @ Stock to 4.8Ghz in your average frame-rates, only in your minimums (Frame-dips). I already tested this out in Post Human and was shocked at the outcome.

 

[el etro]

Post Human is very much GPU-bound, use porly the CPU power.

Before I get my i5 I was running a 7770 on a G620 and still getting get good results in this first level. Once it comes to the jungle the vegetation destroyed my Pentium's rig FPS.
Third level stress better the two.


Thats why i asking you the Root of all Evil results.
Anyway by now thanks for the answers.

(english is not native language here )

 

[Face2Face]

Post Human is very much GPU-bound, use porly the CPU power.

Before I get my i5 I was running a 7770 on a G620 and still getting get good results in this first level. Once it comes to the jungle the vegetation destroyed my Pentium's rig FPS.
Third level stress better the two.


Thats why i asking you the Root of all Evil results.
Anyway by now thanks for the answers.

(english is not native language here )

Ok good to know. I will give it a try - I lost all of my saves, so I have to play the game to get to the third level again, so It may be a later on. Also, playing @ 1440X900 which is equivalent to 1280X1024, you will have more CPU usage regardless.

 

[Jacky60]

The cooling makes a huge difference, my AC extreme coolers have always enabled a substantially improved overclock, I can't fit two three slot coolers in my rig but core overclocks jumped 50Mhz-80Mhz above reference at least( a few generations ago). The memory overclock is hugely impressive. If I didn't have my 6990's I'd have gone 7950 tri-fire a long time ago.