3Dmark Time Spy - Post your scores!

[Frick]

I'm pleased to see my scores holding up nicely. Very nicely.

 

[Random Murderer]

I'm pleased to see my scores holding up nicely. Very nicely.

I don't see anybody touching those scores in quite a while...

 

[EarthDog]

I thought about leaving those scores off Frick... something is not right about them...

 

[Kabiel]

Here is mine:

Intel i7 6850k @ 4.25 Ghz | Asus Strix Gtx 1080 @ 2123mhz | 32 GB G.Skills DDR4 @ 3000 mhz | Asus Rampage V Edition 10 | Corsair 115i CPU AIO Cooler | Samsung 950 Pro M.2 SSD | Strix Air cooled GPU, and Air cooled Thermaltake Core x5 Riing Case | EVGA 850w PS

You may notice the validation failure on the screenshot, if you zoom in you will see this is due to the graphics driver. Which I just updated today, and it seems that has not filtered to 3dmark yet, but it is current newest build NVidia Driver

 

[P4-630]

My score: 0
I don't do windows 10 and DX12 (yet)....

Thanks for testing windows 10!

 

[Vellinious]

15140 - Vellinious / i7 6950X 4.0 / 2 x 1080 @ 2164 / 5499

http://www.3dmark.com/3dm/15733297

 

[BewareBoogeyMan]

Why use this... low very low program ?
You are not have eyes?
Looks like dx9 all test thats not beauthy textures.
And Why scores??? Only reason reason/import/meaning is The minimum fps if under 30 (sometimes 25) when you will not enjoy game
but thats is the point what kind game 3dmark???

 

[EarthDog]

Clearly something you don't understand...its a BENCHMARK. It has DX12 testing elements in it and doesn't look remotely like anything DX9.

Anyway, back to reality here... updated anyone that was close to using the right format (thanks Velli!)

 

[cadaveca]

Anyway, back to reality here... updated anyone that was close to using the right format (thanks Velli!)

You should have some more updates, since I find that latest NVidia benchmark improves scores somehow with 1070. Currently running benchmarks, seen 1000 point difference (93xx to 103xx), could have been getting low scores due to potential hardware problems in test rig though.

Seen anything like this lately?

 

[EarthDog]

@cadaveca

Decided to post it here but leave some things out...


I just received a new 1070 from a vendor a day or so ago... its clocks come in lower than their 1070 I tested a couple months ago... 1924 MHz boost vs 1987 MHz and the new card beats it out across the board except for in Rise of the Tomb Raider. In 3DM Time Spy, the 1924 MHz card scored 5,982 on 375.70 (latest WHQL), while the 1987 MHz card on 368.39 scored 5892. I don't have the original card to test, but its 60Mhz less and scoring as much or MORE than the other one, as well as across the gaming/benchmarking suite (again besides ROTR).

 

[cadaveca]

I am using same cards as a month ago. I have one Samsung-based card and two Micron. I benched the cards separately, and they scored differently, sicn ethey all boost to different clocks, and have different voltages. I have screenshots and saved results on the ORB.

Today I got a memory sample in from Team Group, and started benching before reverting to review config, and noticed the point increase.

Here are some links:

old:

http://www.3dmark.com/spy/614292


new:

http://www.3dmark.com/spy/664806

Now, I do understand that this is with different boards, and different memory, but that's part of why I started to ask...


Does CPU and memory now account for such a large difference? Note the GPU clock speeds are within margin of error for successive benchmark runs...

another benchmark since there are differences in CPU speeds reported:

http://www.3dmark.com/spy/614314

note that before, CPU accounting for 100-250 points was normal.


Thoughts? This is interesting because if I am working on a memory review, maybe there is a need to choose platform used for reviews wisely?

MS-B903 board model is MSI Aegis Ti PC, FYI. will swap cards back into MSI rig and retest in an hour or so, still handing out candy to kids and trying to finish memory benching at the same time, posting from my Surface. Also looking at checking 980's since that's what I use for memory reviews... if benchmark is at fault, I have some serious re-benching to do.

 

[EarthDog]

The only anomaly I have seen is between 2 1070's. The only difference in the system was the GPU and the driver...

... well, the clocks, but as I said the card running the latest driver runs 60Mhz less than the other 1070 and is faster... sometimes by a few % (not FPS, %) with less clocks (memory clocks are the same).

 

[cadaveca]

Can you run a single bench and see if there is any difference... look at the dates, etc... I am totally at a loss for what is going on. I'll bench at same speeds, see if we see some differences....

I have tested all the benchmarks in 3DMark, 980 seems unaffected.

 

[EarthDog]

Just confirm.... single bench to compare the drivers, yes? Once you confirm, I can do that...

EDIT: Wow, am I tired...

So, you want me to run this bench with both drivers (or any single bench with this card and the two drivers) right? LOL

 

[cadaveca]

Just confirm.... single bench to compare the drivers, yes? Once you confirm, I can do that...

Yep.

Normal CPU difference, single GTX 1070:


Stock (old card Samsung) driver 21.21.13.7306:

http://www.3dmark.com/spy/615231

CPU OC (old card, Samsung) driver 21.21.13.7557:

http://www.3dmark.com/spy/632671

New card OC (Micron) driver 21.21.13.7290:

http://www.3dmark.com/spy/535975



To add to your thing about old vs new cards... driver seems to paly a role as well, but not usually this much? I'd very much like to see your numbers to see how different they are from mine. I can pop in any of the cards I have to test, too. All are MSI GAMING 1070, Samsung is GAMING, Micron cards are GAMING X (but tend to clock slightly lower)

 

[EarthDog]

Old card was MSI 1070 Gaming X, new card... also a 1070(micron), but shall remain nameless until a slightly later date.

testing now.. will update and 'ping' you when its done.

 

[Vellinious]

Memory speed, command rate and timings all make a difference in the CPU score. I didn't notice any difference in GPU scores, though.

 

[Vellinious]

I've got an update, btw....went ahead and put some clock on the CPU.

15458 - Vellinious / i7 6950X 4.5 / 2 x 1080 @ 2202 / 5500

http://www.3dmark.com/spy/661446

 

[cadaveca]

Memory speed, command rate and timings all make a difference in the CPU score. I didn't notice any difference in GPU scores, though.

I did, and hence me bringing it up. AS I mentioned, 980 does not seem to be affected at all, but 1070 does in my testing today. Could be I had a problem earlier, but given that my scores seem in line with Earthdog's scores, I'm left a bit confused.

Like, is the difference then something only for TimeSpy? Or are some other apps affected as well? Free performance boost?

 

[Vellinious]

In the GPU test? Ya know, before today, and running Firestrike, trying my best to catch the guy sitting at #1 for 2 x 1080s, I would have said, "no, the CPU / memory makes very little difference in the graphics score". I no longer think that's the case.

That said, in Timespy, I've never noticed a difference in the graphics scores with clock / memory clock variations. In fact, my best graphics score to date on TimeSpy was with neither the memory or the CPU overclocked at all.

 

[cadaveca]

In the GPU test? Ya know, before today, and running Firestrike, trying my best to catch the guy sitting at #1 for 2 x 1080s, I would have said, "no, the CPU / memory makes very little difference in the graphics score". I no longer think that's the case.

That said, in Timespy, I've never noticed a difference in the graphics scores with clock / memory clock variations. In fact, my best graphics score to date on TimeSpy was with neither the memory or the CPU overclocked at all.

Take a look at the scores I posted above in post #161.

 

[Vellinious]

I saw, and noticed that each run had a different driver version. That, in and of itself, can make a world of difference in the graphics score....and considering all 3 runs are within a reasonable margin of error, I'd guess that's where the difference is.

Try the newest drivers, btw....they help a bit more, and can hold higher clocks than the earlier versions.

 

[cadaveca]

I saw, and noticed that each run had a different driver version. That, in and of itself, can make a world of difference in the graphics score....and considering all 3 runs are within a reasonable margin of error, I'd guess that's where the difference is.

Try the newest drivers, btw....they help a bit more, and can hold higher clocks than the earlier versions.

no, higher post up, the SLI scores. 1000 points is not within the margin of error. Do also keep in mind that STEAM says I have 158 hours of running this bench, so I am very familiar with it, since that doesn't account for all the hours spent before I added it to STEAM, or running it on lots of my review rigs.

So yeah, driver is likely the cause, but such a big jump is something I'd like someone else to confirm, because I may have VGA with *ahem* ...issues. I am just about to swap back to MSI Aegis Ti and re-test to see if scores pass to other rig, of if it was something on the ASUS board (which is now scoring much higher)

 

[EarthDog]

Scores are the same, 60 MHz difference in core speeds.

In games I am getting multiple % increases.

 

[Vellinious]

no, higher post up, the SLI scores. 1000 points is not within the margin of error. Do also keep in mind that STEAM says I have 158 hours of running this bench, so I am very familiar with it, since that doesn't account for all the hours spent before I added it to STEAM, or running it on lots of my review rigs.

So yeah, driver is likely the cause, but such a big jump is something I'd like someone else to confirm, because I may have VGA with *ahem* ...issues. I am just about to swap back to MSI Aegis Ti and re-test to see if scores pass to other rig, of if it was something on the ASUS board (which is now scoring much higher)

I saw a 850 point increase with the latest drivers with 1080s in SLI. /shrug

**In Timespy** Didn't do anything thus far in my testing for Firestrike.....