UNIGINE-Valley BENCHMARK scores

[z1tu]

First time doing Valley, i5 4690K @ 4.5Ghz, Palit GTX 780 SuperJetstream @ 1241Mhz Core, 1584Mhz Mem

 

[purecain]

R9 290x@1130/1395 still trying to find the sweet spot...



I am so tempted to buy a GTX980... ive installed my second 290x again and I ran valley... heres my initial run 290x xfire @ 1000/1300mhz

 

[The N]

@purecain

Xfire score isn\'t at its best. should be more around 140 fps and close to 5600 points. anyways sir do some more overclocking soemthing close to 1200mhz. you might beat 780 jetstream. score above yours.

 

[purecain]

yep Ive been tied up but im looking forward to having a proper session now im running xfire... shadow of mordor was smooth before but it feels much more fluid with xfire... everyone should get that game btw...had a lot of fun with it...

 

[fullinfusion]

R9 290x@1130/1395 still trying to find the sweet spot...



I am so tempted to buy a GTX980... ive installed my second 290x again and I ran valley... heres my initial run 290x xfire @ 1000/1300mhz

Just for shits and giggles try setting the clock I'm an 5 MHz increase like 925-930-935.... Try it because these dam cards will go south real fast if you don't find its multi. Mine is in "5" and any value increase above or below 5mhz bumps retard the card and numbers. And I find memory at let's say 1350 will suck if I run 1007 core and 1350 mem but if I use 1323 memory the numbers really shoot up nicely.

 

[Jetpil0t]

I find some of these scores a little random. I think CCC settings make more difference than many people are suggesting and not posting them makes this thread somewhat misleading for a lot of people.

My System:
i5 2500k@4.4Ghz
8GB 1866mhz DDR3
R9 290 @ 1100/1400
Samsung 850 Pro 256gb SSD

Posting 2800 @ the settings originally set out in the guidelines, with all the quality options enabled in Valley, using the default CCC settings. However if I simply disable tessellation, force down AF samples, enable "performance" mode for AF and set the Surface Format Optimization setting I can post 3200 and things more or less the same. But on top this, scores can fluctuate by more than 100 points between end to end benchmarks, so I don't know that Valley is even all that consistent in the first place.

Also there are a couple hundred points to be found running on an SSD vs HDD due to the skip or stutter you can get between scenes, which amounts to zero difference in actual game performance, so again just a small but misleading factor.


Valley is perfectly reasonable gaming benchmark, but unless people are exposing their CCC settings, it's kind of a pointless comparison. I came on here to compare my overclocked GPU performance against the same GPU on different CPUs and my performance against a stock 290x, but the scores here are all over the shop it's impossible to make a comparison.

I don't see how a single 7970 at any speed is faster than an overclocked 290, which according to this thread, multiple people seem capable of doing, on poorer CPUs in some cases, which makes no sense.

Unless of course, their CCC options are tweaked for daily use in games, which if you are using an older GPU you probably will tweak down some quality settings in CCC because you do get quite a bit more performance allowing you to push in game details a higher, which makes a lot of sense to do.

I mean if you really look at some of the scores posted on here for older Crossfire and SLi configurations in particular, you can tell things start to make no sense pretty fast.

No one is really calling any of this out either, which I think is kind of misleading, especially for newbies who might sit somewhat puzzled by their dismally low scores, wondering how these benchmarking wizards can post such substantial scores.

To be clear, I like Valley. It is great for testing overclock stability and cooling performance in games, as opposed to something like Furmark. However for benchmarking purposes, I think there are better alternatives.

 

[The N]

I don't see how a single 7970 at any speed is faster than an overclocked 290, which according to this thread, multiple people seem capable of doing, on poorer CPUs in some cases, which makes no sense.

Brother First of all 7970 is not clearer Equalizer of 290. but Only a very Heavily OCed 7970 like VaporX, can equals or very close even touches single 290 (stock) score/performance in valley benchmark. i did 7970 vaporX to 2700 points which is quite close to my rival on other Forum Called "PakGamers" my Origin.

secondly not all 7970 can go higher @clocks. like reference 925core can max go to 1150mhz. which will score way lower to 290 in that case. i did Own a SApphire 7970 vaporX back a month. which has huge overclocking potential. like 1050mhz stock core and i achieved 1260mhz core through which i able to touch 2700 point in valley benchmark.

My Friend on other Forum scored 2761 Points (i7 2600K + 290 TriX 1000mhz core)
while my top clocks 1260mhz Pass me to 2700 Milestone (i5 2500 + 7970 VaporX @1260/1625mhz)

so there is no trick CCC involve regarding scoring here. and another thing, at least not at my end. i dont know about others on this forum.
this benchmark don't need any haswell to score higher, a 2nd generation cpu would be more than enough due to the fact, its wholly depends on GPU (core and memory) The more you overclock, much better score will be.

 

[z1tu]

Also there are a couple hundred points to be found running on an SSD vs HDD due to the skip or stutter you can get between scenes, which amounts to zero difference in actual game performance, so again just a small but misleading factor.

That's just it, this isn't relevant to gaming at all, it's just E-peen status. ALL synthetic benchmarks are irrelevant, but that's not what these threads are about.

 

[purecain]

valley is gpu dependant and as such gives you a good idea on how a card performs. clocks become irrelevant when your cards memory is producing errors. or your gpu is having to throttle back the core.
@fullinfiusion- ive spent about an hour oc'in the cards and just saw your post... tricky to find that sweet spot... im only getting 5330 atm at 1040/1395... trying your method of 5mhz at a time...

 

[z1tu]

valley is gpu dependant and as such gives you a good idea on how a card performs. clocks become irrelevant when your cards memory is producing errors. or your gpu is having to throttle back the core.

Then why isn't everyone doing furmark and rating gpus like that? Sure, it's fun to see these tests and how smooth they run but they still usually perform differently than a game.

 

[Vayra86]

this benchmark don't need any haswell to score higher, a 2nd generation cpu would be more than enough due to the fact, its wholly depends on GPU (core and memory) The more you overclock, much better score will be.

Wrong. You can consistently see faster CPU's score higher in virtually every benchmark suite, be it 3dMark or Unigine benches. i7's almost always outpace i5's of the same gen, FX 8xxx are usually on par with the i5's. Even when they are a few multipliers apart. Also in game, an i7 will always perform better with the same GPU than an i5 would. It's only 5%-7%, but it is definitely noticeable.

Almost nothing is completely gpu limited, as stated elsewhere, and even an SSD will influence results, not to mention RAM config.

For real world usefulness any Valley benches in the higher score range that deviate by 100 points should be considered similar results. You could very well score anything within a 100 point range spread with exactly the same rig at the same settings. Also, real world gains within this 100 point margin are really negligible, we are talking about a 1-2 avg fps difference here which is simply nothing.

 

[Tatty_Two]

Strange then, when I ran the test with my old Bloomfield i7 930 @ 4.3gig I realised I had HT disabled (it was in the summer and hot), so all GPU settings remained and I ran it 10 minutes later with HT enabled, nothing else touched apart from a slight raise in VCore to accomodate the HT and I scored 140 points less. When I get home I am going to run it again with this 2 core chip and see how the scores compare.

 

[The N]

Wrong. You can consistently see faster CPU's score higher in virtually every benchmark suite, be it 3dMark or Unigine benches. i7's almost always outpace i5's of the same gen, FX 8xxx are usually on par with the i5's. Even when they are a few multipliers apart. Also in game, an i7 will always perform better with the same GPU than an i5 would. It's only 5%-7%, but it is definitely noticeable.

Bro CPU may be give you 20-40 points extra but higher CPU doesn't mean it will give you 100 or 200 point extra. i was comparing with i7 setup in my last post you can see the difference. 800mhz difference between i7 and i5 clock rate. still my setup manage to get close to 290 score. ALL 3DMARK benchmarks relies on every component of your rig. even SSD. but in valley it is 80-90% on GPU.

Almost nothing is completely gpu limited, as stated elsewhere, and even an SSD will influence results, not to mention RAM config.

Yes in 3DMARK benchmarks depends on bits of rigs. i never used SSD but i heard SSd have impact on scores too especially in 3DMARK.

For real world usefulness any Valley benches in the higher score range that deviate by 100 points should be considered similar results. You could very well score anything within a 100 point range spread with exactly the same rig at the same settings. Also, real world gains within this 100 point margin are really negligible, we are talking about a 1-2 avg fps difference here which is simply nothing.

bro we should know here, why are we benching here. its all about competition and comparison with others. and for competition i really likes to go hard on GPU to analyze my GPU's potential. In valley benchamrk its all about points, every 20mhz OC gives you 20 or 30 point extend. so even 10 points needs a little effort it won't comes in free. Bro benching is just FUN, i do becasue i would like to do, 80%benching and 20%gaming these days. Ocing is damn FUN, but one should have a courage to do, and needs Big HEART inside you. its not for KIDs.

In Real world gaming, GPU/CPU performs totally different, even 5FPS improvemnt with OC is unnoticeable. so OC in games is no beneficial.

 

[The N]

so guys i just bought MSI 7950 TF III, only becuase i haven't used it yet. so before going to maxwell or kepler 7xx series, i want to test predecessor of tahiti.

Core i5 2500K @4.5GHz
MSI 7950 3GB TwinFrozer III @1220/1500 MHz/Voltage @1.30v MAX
Gskill 4GB 1600/7.8.7.24 2T

i beated my own 7970 vaporx @1180/1600 MHz......................Hell YEah.!!!!!!!!!!

 

[purecain]

strong card... glad you've had some good luck The N...

im getting 5366points@ 1040/1395


again at 1050/1395

 

[purecain]

cpu @ 3Ghz 3.5Ghz and 4.4Ghz(what im running now) has almost no effect on score... you cant compare furmark to Valley as furmark was designed to see if gpu overclocks were stable, or at least that's what I used it for... until it had been around for 12 months when new gpus came out and started switching themselves off due to the thermal temperatures being off the table. cards were dying/psu's popping.... so I decided not to ever use it again. not even for a second on any of my cards... I would like the solder on my cards not to turn into liquid... lol

Valley is fine as it gives us a very good idea which cards will perform well with this years games....

im thinking of going the 980 route... but I don't want to buy the card and have a better version come out 2months down the road... what to do...

 

[z1tu]

cpu @ 3Ghz 3.5Ghz and 4.4Ghz(what im running now) has almost no effect on score... you cant compare furmark to Valley as furmark was designed to see if gpu overclocks were stable, or at least that's what I used it for... until it had been around for 12 months when new gpus came out and started switching themselves off due to the thermal temperatures being off the table. cards were dying/psu's popping.... so I decided not to ever use it again. not even for a second on any of my cards... I would like the solder on my cards not to turn into liquid... lol

Valley is fine as it gives us a very good idea which cards will perform well with this years games....

im thinking of going the 980 route... but I don't want to buy the card and have a better version come out 2months down the road... what to do...

You'll always be in that position if you think about it. These cards have already been out for what, a month now? You'll be waiting 2 months and probably nothing will come out and you'll be thinking "yeah, something's bound to come out anytime now, I've already waited 2 months, one more can't hurt". You already have a good setup so why are you thinking to upgrade, for the energy and temp savings? (good reason, I'm just asking)

 

[the54thvoid]

im thinking of going the 980 route... but I don't want to buy the card and have a better version come out 2months down the road... what to do...

FWIW the 980 will most likely overclock high enough to match my 780ti (which scored 105 in valley at crazy clocks). An overclocked 980 will beat a 290x by a decent margin alone (review sites talking 20%+). But, there will almost certainly be a larger Maxwell card to come, probably based on any HPC part. Unless the energy bill is noticeable, I wouldn't move away from 2 x 290X's. The gains in performance versus the price step up aren't worth it, especially with 'something' round the corner.

 

[Locksmith]

@purecain
Update..

single GTX 970 MSI Gaming 4G TwinFrozer V

Been Pushing it, since I broke her in..

OC
1521Mhz GPU
8000Mhz Vram

I noticed in the leader-board my CPU is wrong.. its a 4790K devil canyon .. u have 4770K
OC @ 4.7Ghz

New Score, should take to up to just below you.

Thanks Pal..

 

[The N]

Nice Overclock there on 970 @Locksmith But you marginally beat 290X of @purecain

 

[z1tu]

@purecain
Update..

single GTX 970 MSI Gaming 4G TwinFrozer V

Been Pushing it, since I broke her in..

OC
1521Mhz GPU
8000Mhz Vram

I noticed in the leader-board my CPU is wrong.. its a 4790K devil canyon .. u have 4770K
OC @ 4.7Ghz

New Score, should take to up to just below you.

Thanks Pal..

This is why I was saying these tests don't do justice to the cards, all that overclock and you can't beat my score, which you should. In games however, the story is different, your overclocked 970 will have higher fps than mine. I wonder if there are any new games out there that paint the picture differently and have these new cards perform worse than their older counterparts.

 

[The N]

well that level of ocing, 970 should beat 780. officially and with many reviews 970 better to 780 but real testing differenct story. @Locksmith may be you need some tweaking in testing environment. but still won't make any noticeable difference.

 

[Locksmith]

Nice Overclock there on 970 @Locksmith But you marginally beat 290X of @purecain

Thanks pal...

EDIT:
@purecain
new Score....
GPU 1534Mhz Ram 8032Mhz
View attachment 59774

 

[purecain]

@z1tu was out of it there for a few days.... it would be the power bill... I watched a few youtube videos of the 290 vs 780Ti vs 980 split screen timed demo through bf4, crisis 3 and metro2033. the most surprising thing is the 3 cards have near identical performance...

the amount of extra frames in game using a 980 was 3fps... ive heard in some scenarios the card pulls even further ahead but I think its the new AA method being implemented within the new architechture where it takes half the sample info it did before and then uses alternating positions & frames to produce the AA effect, which plays as bigger part as the more efficient shaders. (I like the new AA method btw as using less data to get the same result can only help with latency and give the card more power for other jobs)

if the 980 had 500more shaders I would buy one right now... although I hear AMD are in need of some revenue and their cards will be out in just a few months. for the sake of keeping the competition alive and prices reasonable I may support them.

a sub £400 40% more efficient 2816 shader gpu, that ran at 1400mhz(1800mhzOC) with 8gb of 7000mhz Gddr5 would put a smile on face.

that's 'if' the efficiency of the new 20nm node crosses over to AMD's latest iteration.

 

[kurosenpai]

Hey its my first time in this techpowerup forums so HI everyone.

Just bought the Zotac GTX 970 AMP Omega and I wanna share my Valley Score too.
Mind you Im using my old i5 3450 so I think I might get a bit bottleneck here. ( Correct me if Iam wrong )

1102/1762 boost 1441mhz

then i overclock it a bit

1302/1862 boost 1600mhz ish or something i dont remember