UNIGINE-Valley BENCHMARK scores

[trog100]

i have done some more testing with valley.. there is something wrong with it..

heaven works as it should do.. with my two 980 ti cards.. it scales and responds to setting changes as it should do with one or two cards running..

valley with two cards running dosnt work properly.. i can go from ultra 8 x aa right down to low with aa off.. whatever settings i put in my fps stays the same.. low runs the same frame rates as ultra with valley.. something is wrong with it,, he he..

the other thing i notice is that both valley and heaven default to the ultra settings when the benchmark run is selected.. they dont change the selected aa setting but everything else defaults to ultra..

with valley and two cards running i see bugger all difference in frame rates with it just running in ulta.. low.. or whatever.. it seem to act as if its changed the settings but my frames rates stay the same..

my (apparent) poor sli scaling in valley seems to be down to the fact that with just the one card running the benchmark does score lower or higher depending on the aa setting.. with two cards running it scores exactly the same no matter where i put the aa setting.. it should not do this.. heaven dosnt..

playing with heaven (sli) i see nearly a doubling of frames rate going from ultra to low.. with valley i see bugger all difference.. none at all.. i conclude.. something is wrong with valley.. not because my cards aint scaling that well.. just because my frame rates stay the same no matter what setting i run it at.. its broke..

trog

 

[Vellinious]

i have done some more testing with valley.. there is something wrong with it..

heaven works as it should do.. with my two 980 ti cards.. it scales and responds to setting changes as it should do with one or two cards running..

valley with two cards running dosnt work properly.. i can go from ultra 8 x aa right down to low with aa off.. whatever settings i put in my fps stays the same.. low runs the same frame rates as ultra with valley.. something is wrong with it,, he he..

the other thing i notice is that both valley and heaven default to the ultra settings when the benchmark run is selected.. they dont change the selected aa setting but everything else defaults to ultra..

with valley and two cards running i see bugger all difference in frame rates with it just running in ulta.. low.. or whatever.. it seem to act as if its changed the settings but my frames rates stay the same..

my (apparent) poor sli scaling in valley seems to be down to the fact that with just the one card running the benchmark does score lower or higher depending on the aa setting.. with two cards running it scores exactly the same no matter where i put the aa setting.. it should not do this.. heaven dosnt..

playing with heaven (sli) i see nearly a doubling of frames rate going from ultra to low.. with valley i see bugger all difference.. none at all.. i conclude.. something is wrong with valley.. not because my cards aint scaling that well.. just because my frame rates stay the same no matter what setting i run it at.. its broke..

trog

If Valley was broke, others would have the same issue....and they're not.

Aren't you the same guy that started the thread claiming that NVIDIA and it's partners were fudging the claimed clock rates on the box? Just sayin.....you were wrong there too.

 

[Vellinious]

i have done some more testing with valley.. there is something wrong with it..

heaven works as it should do.. with my two 980 ti cards.. it scales and responds to setting changes as it should do with one or two cards running..

valley with two cards running dosnt work properly.. i can go from ultra 8 x aa right down to low with aa off.. whatever settings i put in my fps stays the same.. low runs the same frame rates as ultra with valley.. something is wrong with it,, he he..

the other thing i notice is that both valley and heaven default to the ultra settings when the benchmark run is selected.. they dont change the selected aa setting but everything else defaults to ultra..

with valley and two cards running i see bugger all difference in frame rates with it just running in ulta.. low.. or whatever.. it seem to act as if its changed the settings but my frames rates stay the same..

my (apparent) poor sli scaling in valley seems to be down to the fact that with just the one card running the benchmark does score lower or higher depending on the aa setting.. with two cards running it scores exactly the same no matter where i put the aa setting.. it should not do this.. heaven dosnt..

playing with heaven (sli) i see nearly a doubling of frames rate going from ultra to low.. with valley i see bugger all difference.. none at all.. i conclude.. something is wrong with valley.. not because my cards aint scaling that well.. just because my frame rates stay the same no matter what setting i run it at.. its broke..

trog

My apologies Trog...you may not be entirely wrong on this after all. Last night I was playing around with settings, and noticed on another forum that an SLI run with 0 x AA was just HORRIBLY low. So, I ran a single card 970 run through and scored 5089. Then, I enabled SLI, and...scored 5100 something. I tried it several different times, with different clocks, power settings, even edited the bios file and tried a different bios...no change. It just wasn't utilizing the GPUs to full potential. Usually hanging around about 60-70% usage. Someone in another thread suggested that maybe CPU core clock could be affecting the score with 0 x AA. My 5820k was at 4.4 (my daily clock) last night when I did these tests, so, I'm going to run the test runs again at stock CPU clocks, and then again at my benchmark setting of 4.7ghz and see what the difference is.

Considering this with 2 x 970s

Valley on Extreme HD: 5581
Valley on highest settings with 4 x AA: 6500
Valley on highest settings with 0 x AA: 5100ish

Yeah...something is wrong. I'm thinking it may be something like the Furmark "auto throttle" "FEATURE" that NVIDIA added to keep that stress test from cooking GPUs. I'll post more tonight when I get more information. I've also left a message for NVIDIA and EVGA tech support to see if they know of any "issues" that would cause this, or if it IS actually a "feature".

Doesn't really explain the 4 x AA issues you're having, but....they could be related somehow.

 

[Vellinious]

Ok...did some more testing with 0 x AA and 4 x AA. With 4 x AA, the SLI scores are normal, no problems. I ran a single card in 0 x AA, pushed the clocks up quite a bit to have some fun with it, and scored 5700 something. Then, enabled SLI, kept the same clocks, and.....5500 something.

Seems to work fine with 4 x AA and 8 x AA, but with no AA? Something is jacked up with either Valley, or...as I speculated in the previous post, NVIDIA has some funky "throttling" feature when AA is not enabled, and won't allow over a certain FPS / power / TDP / heat limit? The weirdness continues....

4 x AA won't fully load the GPU. Gotta be something NVIDIA has done. I asked a few others to do the same testing, and they all confirm that they get the exact same issue. With 0 x AA, the GPU is loading up to about 67% max, with 4 x AA, around 90 - 92%, and with 8 x AA up to 99%.

I noticed that with the 4 x AA setting, if you enable KBoost in Precision X, it helped quite a bit, but it's still not loading up all the way.

I asked a friend who's running Kepplers to run the same tests, and he doesn't have the same issues. So, pretty sure that's an NVIDIA / Maxwell thing.

 

[Gregory Hartley]

i7 4790k
GTX 970 Gigabyte G1
1519/3505

 

[trog100]

4 x AA won't fully load the GPU. Gotta be something NVIDIA has done. I asked a few others to do the same testing, and they all confirm that they get the exact same issue. With 0 x AA, the GPU is loading up to about 67% max, with 4 x AA, around 90 - 92%, and with 8 x AA up to 99%.

I noticed that with the 4 x AA setting, if you enable KBoost in Precision X, it helped quite a bit, but it's still not loading up all the way.

I asked a friend who's running Kepplers to run the same tests, and he doesn't have the same issues. So, pretty sure that's an NVIDIA / Maxwell thing.

what going on with my pair of 980 ti cards seems simple enough.. with two cards running the benchmark runs fully maxed out with 8 x AA no matter where i put the settings.. i think the score is correct for the settings the card is actually running at..

what skews things is this thread is based on only 4 x AA.. a single cards runs at 4 x AA.. two of them run fully maxed out at 8 x AA.. they just wont run at only 4 x AA even though its set that way.. in fact valley will only run my two cards fully maxed and score accordingly which is okay until you start making comparisons with a single card running on lesser settings..

why this is i havnt a clue.. i blame the valley benchmark.. heaven works as it should.. valley is broke..

i like using furmark it honestly throws everything at the gpu it can and its the built in gpu load or temp limits that throttle things down to what nvidia think the hardware can stand..



trog

 

[Vellinious]

what going on with my pair of 980 ti cards seems simple enough.. with two cards running the benchmark runs fully maxed out with 8 x AA no matter where i put the settings.. i think the score is correct for the settings the card is actually running at..

what skews things is this thread is based on only 4 x AA.. a single cards runs at 4 x AA.. two of them run fully maxed out at 8 x AA.. they just wont run at only 4 x AA even though its set that way.. in fact valley will only run my two cards fully maxed and score accordingly which is okay until you start making comparisons with a single card running on lesser settings..

why this is i havnt a clue.. i blame the valley benchmark.. heaven works as it should.. valley is broke..

i like using furmark it honestly throws everything at the gpu it can and its the built in gpu load or temp limits that throttle things down to what nvidia think the hardware can stand..



trog

Except that NVIDIA built in a "throttle" for furmark....it's not even a good stress test anymore. I'm wondering if there's not the same kind of thing with no AA, low AA settings that keep frame rates, power settings/consumption at certain levels? I dunno.

This is the only site I'm a part of that hasn't used just the Extreme HD preset for their "Valley leaderboards".

 

[Aquinus]

Ok...did some more testing with 0 x AA and 4 x AA. With 4 x AA, the SLI scores are normal, no problems. I ran a single card in 0 x AA, pushed the clocks up quite a bit to have some fun with it, and scored 5700 something. Then, enabled SLI, kept the same clocks, and.....5500 something.

Seems to work fine with 4 x AA and 8 x AA, but with no AA? Something is jacked up with either Valley, or...as I speculated in the previous post, NVIDIA has some funky "throttling" feature when AA is not enabled, and won't allow over a certain FPS / power / TDP / heat limit? The weirdness continues....

Keep an eye on CPU load. You could be encountering a CPU bottleneck. When you increase AA you're putting more load on the GPU, reducing frame rate, thus reducing CPU load.

You should be able to confirm this by running the benchmark with the CPU at stock with SLI enabled, then again with the CPU overclocked. The score and overall GPU load should increase under 0x AA if that's the case.

Another way to confirm is by running at a lower resolution and seeing if you're getting the same kinds of frame rates which could also indicate a CPU bottleneck.

 

[Vellinious]

The CPU clock doesn't make any difference there. I did several runs at my daily clock of 4.3ghz on my 5820k, then did several on the benchmark clock of 4.7ghz and got nearly the same results. Maybe a 20 point difference, but that could be accounted for in the wind.... lol

I'll run some more tests this weekend to see, but....I don't think that's the issue.

Keep an eye on CPU load. You could be encountering a CPU bottleneck. When you increase AA you're putting more load on the GPU, reducing frame rate, thus reducing CPU load.

You should be able to confirm this by running the benchmark with the CPU at stock with SLI enabled, then again with the CPU overclocked. The score and overall GPU load should increase under 0x AA if that's the case.

Another way to confirm is by running at a lower resolution and seeing if you're getting the same kinds of frame rates which could also indicate a CPU bottleneck.

I went ahead and played with it tonight. You're right. At no AA, the CPU clock speed made a huge difference in the outcome. At 4 x AA though, it didn't make a noticeable difference.

Good info! Thanks!

 

[...PACMAN...]

i5 4690k@4.5Ghz
PaliT GTX 960 Jetstream 4Gb
1428/3705

 

[Tatty_Two]

Finally got around to tinkering with my 290X, trying to find that sweetspot, actually broke the 3500 score and I feel there is still more to come, the VRM's despite the over voltage stayed cool, damn this is one fine cooler!

Sapphire R9 290X 4GB Vapour-X OC @ 1140mhz
i5 4670K @ 4.1Ghz (actually got better scores reducing CPU overclock?....strange)

 

[Ferrum Master]

Okay... CPU @4.7Ghz now. GPU 1455Mhz, VRAM 3800MHz

 

[trog100]

Except that NVIDIA built in a "throttle" for furmark....it's not even a good stress test anymore. I'm wondering if there's not the same kind of thing with no AA, low AA settings that keep frame rates, power settings/consumption at certain levels? I dunno.

This is the only site I'm a part of that hasn't used just the Extreme HD preset for their "Valley leaderboards".

furmark is the only thing i can get that takes my power limits over 100%.. with furmark if i set 110% i get 110%.. with heaven or valley my power usage never goes over the 100% usually it a bit less.. these cards have temp throttling.. power throttling and voltage throttling all built in with the default bios.. its all in the so called boost speed.. you get more or less depending on the three factors i have mentioned..

they are pretty much "user" proof..

one interesting thing with furmark is one card throttles down more than two cards.. i can go lower than 1000 mhz with one card but i get around 1140-ish with two cards..

valley and heaven only use around 20% cpu usage at high settings.. i can run wprime and valley at the same time and still see a similar score.. they are not in the slightest cpu bottlenecked.. they are a pure gpu test which i supposed is what they are meant to be..

trog

 

[Tatty_Two]

furmark is the only thing i can get that takes my power limits over 100%.. with furmark if i set 110% i get 110%.. with heaven or valley my power usage never goes over the 100% usually it a bit less.. these cards have temp throttling.. power throttling and voltage throttling all built in with the default bios.. its all in the so called boost speed.. you get more or less depending on the three factors i have mentioned..

they are pretty much "user" proof..

one interesting thing with furmark is one card throttles down more than two cards.. i can go lower than 1000 mhz with one card but i get around 1140-ish with two cards..

valley and heaven only use around 20% cpu usage at high settings.. i can run wprime and valley at the same time and still see a similar score.. they are not in the slightest cpu bottlenecked.. they are a pure gpu test which i supposed is what they are meant to be..

trog

Yes, as I said a couple of posts up, I actually get 100 points more from my CPU being at 4.1gig rather than 4.4gig, although I find that strange it's clear that predominantly at least it's a GPU orientated bench.

 

[trog100]

cpu usage 13.2 percent..

gpu usage 99 percent

power usage 98 percent..

core boost speed at 1442..

core tempt 66 C

my power limit is set at 105 percent and my temp limit is at 83 C..

this is just with one card running.. something is holding it back.. it could well be built in voltage limits.. i aint sure..

the cpu load is not much more than idle..

trog

 

[jboydgolfer]

now that i have my new(to Me) GPU installed, i thought id add my results here...Im running on the stock voltages for the 7970 ,it gets grumpy when i mess with them. i5 is running @ 4.1Ghz(could have gone as high as 4.7Ghz, but i was too lazy to go into my bios )
sorry about the tiny process explorer window Right Below Here, I needed to cover up all of the porn bookmarks

 

[Tatty_Two]

Your score should be higher than that with a 7970? With those clocks you should be expecting around 2400 I would think....... unless you have full detail still on in CCC?

 

[Vellinious]

cpu usage 13.2 percent..

gpu usage 99 percent

power usage 98 percent..

core boost speed at 1442..

core tempt 66 C

my power limit is set at 105 percent and my temp limit is at 83 C..

this is just with one card running.. something is holding it back.. it could well be built in voltage limits.. i aint sure..

the cpu load is not much more than idle..

trog

You should really use HW Monitor and monitor peak CPU usage. I did this the other day on the Extreme HD preset and found that with my 5820k at stock clocks, it was hitting 99% on one core, and during those times, the GPU utilization was dropping off of 100%. Cleared it up when I ran the CPU at 4.7, but..... This is the only benchmark I see that on though, I tested Heaven in 1080 at the highest settings, doesn't do it there, and tested Firestrike, Skydiver, etc. No bottleneck.

Even if that were the issue.....your second card should at least be running some of the time. And it doesn't appear to be doing anything.

 

[trog100]

i am not really saying i have a "bottleneck".. just that something limits the cards power usage below what i have it set at.. if i set the power limit at 105 with an overclocked card i would expect to see 105 not 99 or less..

one thing i am sure of.. heaven and valley would probably be just as happy running on an old dual core chip.. the firestike benchmarks show a higher score with a fast (more cores the merrier) chip but that is only down to the separate physics (cpu) score which does use all the cores and hyper threading that is available.. the graphics part dosnt need much cpu power..

i tried switching my hyper threading off the other day.. with things that really do use it (which isnt much) i saw a 10 c drop in cpu tempts and maybe a 25 percent drop in the physics part of the firestike benchmark.. with normal gaming i saw no difference..

i favour wprime for cpu testing and furmark for gpu testing.. the cpu burner built into furmark is pretty usefull as well.. it heats things up more than wprime but not as much as prime95 which i think is a silly tool to use..

furmark loads the cpu at around 23 percent.. valley at 12-ish percent.. both are pure gpu testers.. cpu speed (or the number of cores) makes next to no difference..

valley also gets my gpu boost speed well wrong.. it certainly aint doing the 1650 plus that shows as the core speed.. it seems to get the tempts and memory right but is way out with the boost speed..

i have just pointed out oddities i have noticed with valley.. maybe off topic maybe not.. he he..

i do have a pair of 980 ti cards installed.. for the gaming resolutions i use one card is more than enough.. i have sli disabled.. burning an extra 250 watts of heat for no noticeable gain dosnt make much sense.. even to me.. my second card is there if i need it..

comparing sli and none sli with valley did point out the odd way it behaves..

trog

 

[Vellinious]

i am not really saying i have a "bottleneck".. just that something limits the cards power usage below what i have it set at.. if i set the power limit at 105 with an overclocked card i would expect to see 105 not 99 or less..

one thing i am sure of.. heaven and valley would probably be just as happy running on an old dual core chip.. the firestike benchmarks show a higher score with a fast (more cores the merrier) chip but that is only down to the separate physics (cpu) score which does use all the cores and hyper threading that is available.. the graphics part dosnt need much cpu power..

i tried switching my hyper threading off the other day.. with things that really do use it (which isnt much) i saw a 10 c drop in cpu tempts and maybe a 25 percent drop in the physics part of the firestike benchmark.. with normal gaming i saw no difference..

i favour wprime for cpu testing and furmark for gpu testing.. the cpu burner built into furmark is pretty usefull as well.. it heats things up more than wprime but not as much as prime95 which i think is a silly tool to use..

furmark loads the cpu at around 23 percent.. valley at 12-ish percent.. both are pure gpu testers.. cpu speed (or the number of cores) makes next to no difference..

valley also gets my gpu boost speed well wrong.. it certainly aint doing the 1650 plus that shows as the core speed.. it seems to get the tempts and memory right but is way out with the boost speed..

i have just pointed out oddities i have noticed with valley.. maybe off topic maybe not.. he he..

i do have a pair of 980 ti cards installed.. for the gaming resolutions i use one card is more than enough.. i have sli disabled.. burning an extra 250 watts of heat for no noticeable gain dosnt make much sense.. even to me.. my second card is there if i need it..

comparing sli and none sli with valley did point out the odd way it behaves..

trog

You do realize that the power target slider you're setting in AB or PCX, whatever you're using is just a number in the bios, yes? You could set it to 1000 watts in the bios, and you'd still only be pulling so much power through the card at a certain clock. Setting the power limit to 105%, doesn't mean that the card is going to run at 105%. lol

On every benchmark run you ever do, you should have the power target slider at maximum, no matter what. It'll make sure you're not hitting a power limit perf cap. = )

 

[Mercennarius]

View attachment 69355 Finally got around to tinkering with my 290X, trying to find that sweetspot, actually broke the 3500 score and I feel there is still more to come, the VRM's despite the over voltage stayed cool, damn this is one fine cooler!

Sapphire R9 290X 4GB Vapour-X OC @ 1140mhz
i5 4670K @ 4.1Ghz (actually got better scores reducing CPU overclock?....strange)

Nice Score. What does your 290X score at stock clocks?

 

[jboydgolfer]

in regards as a seperate post in the hopes that someone might shed some light on this, I apparently am getting slower than optimal speeds, so im trying to figure out if i can do anything. Im running PCIE x8 V2.0 MY board has a issue where it CANNOT run in the top pcie slot, so please dont bother suggesting it, unles syou know of a way to get MY exact board to work with that slot and this card, but after a Large amount of trying , and bios updates, this is the best i could get.

if its a problem for me to ask these questions here, just ignore it, i dont want someone being grumpy that im trying to solve a problem in a bench thread. technically its not for this purpose. im just desperate, its so frustrating.

thanks

 

[ASOT]

 

[ASOT]

now that i have my new(to Me) GPU installed, i thought id add my results here...Im running on the stock voltages for the 7970 ,it gets grumpy when i mess with them. i5 is running @ 4.1Ghz(could have gone as high as 4.7Ghz, but i was too lazy to go into my bios )
sorry about the tiny process explorer window Right Below Here, I needed to cover up all of the porn bookmarks

I beat u ))) no OC!

 

[jboydgolfer]

good for you, even thoug hmy setting s are running @ x8, i dont think you should be boasting that, since i am close to your score running @ a fraction of my potential speed.