Benchmarking my very, very imbalanced build

[AtenRa]

1440p supports VSR only at 3200x1800, according to AMD. Is this not true? http://www.amd.com/en-us/innovations/software-technologies/technologies-gaming/vsr

Or are you suggesting to run 4K VSR from 1080p on a native 1440p panel?

I believe both we were talking about VSR of 3200x1800 on 1440p

 

[moonbogg]

Sorry for my complete disappearance from this thread for a while now, but summer tends to be a very busy time for me (being a student in Norway pretty much means working your ass off through the summer to survive). I managed to sell off the old GPU far faster than anticipated (although for next to no money, but that's more or less what it's worth anyhow), but I ran through a few benchmarks before then (not all, sadly). Thanks to a new Hyper 212 EVO I've also managed to squeeze my CPU up to 3.5GHz, and if I get the time I'll try to get it higher. OCCT is throwing out some errors after a few minutes, but I've yet to see an OS crash, so that's stable enough for me.

I'm at work right now, so I can't post any benchmarks from here (plus I'd prefer to make some graphs of some kind, rather than just post a bunch of screenshots). Some general observations, though: pushing the CPU seems to be helping minimum framerates in some scenarios (Metro: LL stands out, Bioshock Infinite is noticeable too), but maximums and averages are mostly the same. 3DMark suffers from low physics scores, but OCing amends that some too. Still, my scores are among the lowest I've found for the Fury X. On the other hand, not the worst. Also, the improvement in graphics performance is as expected: staggering. Going from maximum framerates in the low 40s at medium/high settings@2560x1440 in Tomb Raider, to maxing out every setting (including TressFX) and still having minimums above 60fps is pretty amazing. Being able to run Metro: LL at all at 2560x1440 is a whole new world of perfomance.

I'll post what I can give you at least: 3DMark scores.

My highest Fire Strike with the 6950: link.
Fire Strike with the Fury X, CPU @ 3,2GHz, 3,4GHz and 3,5GHz: link, link, and link.

Sky Diver with the 6950: link.
Sky Diver with the Fury X, same clocks as above: link, link, and link.

The 3,5GHz runs are in W10, but I'll try W7 too to see if there's any difference.

Hopefully I'll have time to post some more game benchmark results before I leave for a short holiday Sunday. Otherwise, they'll be up some time the week after.

NICE! So it worked. Excellent. Now all you need is a new platform and you'll be flying even faster.

 

[tential]

1440p supports VSR only at 3200x1800, according to AMD. Is this not true? http://www.amd.com/en-us/innovations/software-technologies/technologies-gaming/vsr

Or are you suggesting to run 4K VSR from 1080p on a native 1440p panel?

The limitations of VSR are so stupid it kills me. Why does VSR have these ridiculous restrictions? Especially when we know some are made up and the cards are clearly capable of this. We need an official answer from AMD.

 

[tential]

Just Now:

I believe both we were talking about VSR of 3200x1800 on 1440p

Page 2:

yea +1

For 1440p with VSR 4K and you are set.

The C2Q 9450 @ 3.2GHz + Fury X at 1440p (even without VSR 4K) will be faster than a Core i5 Haswell @ 4.5GHz + HD6950 in almost every 2013 onwards game, especially the latest 2015 games.

---

 

[lehtv]

I believe both we were talking about VSR of 3200x1800 on 1440p

Well, that makes no sense. 4K is not 3200x1800, and you were talking about 4K.

1800p is is 1.25 x 1440p, and results in a blurry image. Not a good idea IMO

 

[AtenRa]

Clearly I and I believe Silverforce11 forgotten we cannot use 4K VSR at 1440p, so obviously we were talking about 3200x1800.

And I dont think you will get a blurry image.

 

[Valantar]

I've started putting together a few graphs now, but it looks like I'll have to re-run a few benchmarks at slightly different settings if I'm to compare them to numbers from reviews of the card. Oh well. Shouldn't take too long, but I won't be able to do this until I return from holiday. I'll have the results ready some time the week after next, in other words. Who knows, maybe I'll be able to push my CPU even further, to see what kind of difference that makes. Sorry to keep you waiting for so long, though

 

[amenx]

Better than I expected, well done! 1440p should alleviate much of the bottleneck and in graphically intensive games should do well.

 

[DustinBrowder]

The CPU will definitely bottleneck the card, especially at lower resolutions.

I mean windows10 might alleviate some of the bottleneck, but it won't do magic.

I'd say if you game on anything bellow 1440p you'll be bottlenecked significantly by the CPU, less so if you use 1440p and higher resolutions.

 

[lehtv]

And I dont think you will get a blurry image.

1.25x DSR on 1080p results in a very blurry image. Maybe on 1440p it will be less strikingly obvious but it'll still be almost certainly noticeable and it will look worse than plain 1440p.

 

[SimianR]

Would love to see if you could get Ark: Survival Evolved going on that system. I have an old E8400 system that I wouldn't mind grabbing a cheap Q6600 and a somewhat new GPU and getting a second system going.

 

[NesuD]

Sorry for my complete disappearance from this thread for a while now, but summer tends to be a very busy time for me (being a student in Norway pretty much means working your ass off through the summer to survive). I managed to sell off the old GPU far faster than anticipated (although for next to no money, but that's more or less what it's worth anyhow), but I ran through a few benchmarks before then (not all, sadly). Thanks to a new Hyper 212 EVO I've also managed to squeeze my CPU up to 3.5GHz, and if I get the time I'll try to get it higher. OCCT is throwing out some errors after a few minutes, but I've yet to see an OS crash, so that's stable enough for me.

I'm at work right now, so I can't post any benchmarks from here (plus I'd prefer to make some graphs of some kind, rather than just post a bunch of screenshots). Some general observations, though: pushing the CPU seems to be helping minimum framerates in some scenarios (Metro: LL stands out, Bioshock Infinite is noticeable too), but maximums and averages are mostly the same. 3DMark suffers from low physics scores, but OCing amends that some too. Still, my scores are among the lowest I've found for the Fury X. On the other hand, not the worst. Also, the improvement in graphics performance is as expected: staggering. Going from maximum framerates in the low 40s at medium/high settings@2560x1440 in Tomb Raider, to maxing out every setting (including TressFX) and still having minimums above 60fps is pretty amazing. Being able to run Metro: LL at all at 2560x1440 is a whole new world of perfomance.

I'll post what I can give you at least: 3DMark scores.

My highest Fire Strike with the 6950: link.
Fire Strike with the Fury X, CPU @ 3,2GHz, 3,4GHz and 3,5GHz: link, link, and link.

Sky Diver with the 6950: link.
Sky Diver with the Fury X, same clocks as above: link, link, and link.

The 3,5GHz runs are in W10, but I'll try W7 too to see if there's any difference.

Hopefully I'll have time to post some more game benchmark results before I leave for a short holiday Sunday. Otherwise, they'll be up some time the week after.

I'm running that card with a i7-3770k at 4.2Ghz and in that same test i got 13098. That should give you an idea of the potential of the card with a current cpu. 3770 is a couple generations old and still show a very large improvement over your q9450. I'd estimate you should easily realize close to a 50% improvement in performance when you are able to move to a current cpu.

 

[sm625]

A locked haswell i5 plus a GTX970 scores the same as your Fury X with core 2 quad at 3.2GHz, so the results are right in line with what I was expecting. The good news is that you are getting near perfect cpu scaling, so the higher OC, the better.

 

[Yuriman]

My experience of moving my wife from a stock Q6600 to a Haswell i3 4130 was that FPS in most of the games we play went up almost 70%. Accounting for clockspeed differences, the i3 was still ~35% faster with half of the cores.

 

[AtenRa]

My experience of moving my wife from a stock Q6600 to a Haswell i3 4130 was that FPS in most of the games we play went up almost 70%. Accounting for clockspeed differences, the i3 was still ~35% faster with half of the cores.

What games are you playing ??

 

[Yuriman]

At the time, our main games were Guild Wars 2, Skyrim, and Civ5. Civ saw the smallest increase, with only 50% improved FPS. I was surprised by this though, since AFAIK, Civ is well threaded.

Guild Wars 2 improved from an unplayable ~25fps to 43fps, Skyrim from 26 to 44 in the test areas I chose.

 

[sm625]

Guild Wars 2 improved from an unplayable ~26fps to 44fps, GW2 from 25 to 43.

Arent these the same game?

 

[Yuriman]

Sorry, edited my post, was distracted when I replied.

 

[TheELF]

Civ saw the smallest increase, with only 50% improved FPS. I was surprised by this though, since AFAIK, Civ is well threaded.

No it's still basically a single threaded game,even beyond earth is not much better

Spoiler

 

[AtenRa]

No it's still basically a single threaded game,even beyond earth is not much better

Dont think so,

 

[AtenRa]

At the time, our main games were Guild Wars 2, Skyrim, and Civ5. Civ saw the smallest increase, with only 50% improved FPS. I was surprised by this though, since AFAIK, Civ is well threaded.

Guild Wars 2 improved from an unplayable ~25fps to 43fps, Skyrim from 26 to 44 in the test areas I chose.

Ahh yes, GW2 and Skyrim will benefit tremendously from a modern CPU like Haswell.

 

[Hitman928]

Windows 7
Core 2 Quad 9300 @ 3.65 GHz
8 GB DDR 2 RAM @ 974 MHz
Cat 15.7 (except Hitman which was done on 15.6) @ 1080p

All game tests taken from in game play and repeated for accuracy. BF3 played on 64 player server on Caspian Border map. PVZ was sharkbite shores or whatever the map is. PVZ and BF3 repeated 6+ times and the average of all tests was taken as the average frame rate. The lowest min seen was taken as the minimum. Settings for each game are what I deemed as playable for the 6870 so only Batman AC was nearly maxed out (2nd to highest tess, no MSAA). 6870 had a minor clock bump to 950 MHz, 290 stock reference clocks.

 

[Hitman928]

More graphs coming when I have time.

 

[SimianR]

Nice graphs Hitman, goes to show that a new GPU can breath some life into an old system. Can always use the GPU in a new system when and if you decide to build a completely new system as well.

 

[AtenRa]

A couple of graphs from my Core 2 Quad 9450@ 3.2GHz

Core 2 Quad 9450 @ 3.2GHz
2x 2GB DDR-2 1066MHz

Core i7 3770K @ 4.4GHz
2x 4GB DDR-3 2133MHz

HD6950 @ 810MHz
HD7950 @ 1GHz

Win 8.1 64bit
Cat 15.7

----------------------