Bad Firestrick combined score

[TheDavLab]

Hi all!

Some months ago I do some OC and recorded some score on Firestrike.
Them I noticed that my single score (both graphics and physics) was some time better then others with my same hardware but my combined score was really bad.

It could be because I use 2x16GB of DDR4 double side instead of single side memory? Can I tweak something to achieve better result?

My rig:
CPU: Ryzen 7 1700
GPU: Vega 64
RAM: G.Skill Trident Z 3200 GHz 14-14-14-34
MoBo: Asus Prime X470-Pro
OS: Win10

Here my score:
https://www.3dmark.com/fs/16592472

Here single score below mine but better combined:
https://www.3dmark.com/fs/14529855

Thanks in advice for your reply

 

[EarthDog]

The first thing I see is different driver versions.

 

[Johan45]

I don't think it's a driver, combined score is affected by stability. Either the CPU or GPU being a bit unstable can really affect the combined score

 

[TheDavLab]

Thank you both!

I'll do some test, using also the same driver and try to have the whole system more stable.

 

[Bill_Bright]

I see many differences starting with a different user.

Beside different GPU speeds, the CPU speeds are different too. Windows has been updated. RAM has more than doubled and there is a different hard drive model.

These are not the same computer.

 

[EarthDog]

I see many differences starting with a different user.

Correct. That was never in question.

Beside different GPU speeds, the CPU speeds are different too. Windows has been updated. RAM has more than doubled and there is a different hard drive model.

And the vast majority of those doesn't really affect the GPU score.

* The GPU speed differences are negligble really. 30 Mhz on core (of 1700+) and 5 Mhz on RAM.
* System RAM doesn't matter much (and I think one of these is reading it wrong at 10000MB). 8GB isn't used so more quanitity isn't relevant. That said, speed can play a small role here.
* HDD makes zero difference. This isn't a system test.
* Windows version may have a bit to do with it...
* CPU speeds don't have much affect here in the graphics score - but I would still raise it to 4 GHz and see what happens. I also note one of these shows 8c while the other shows 16c. Typically this doesn't matter either (except on CPU score), but if that is reading correctly, he should run with all c/t.

 

[TheDavLab]

Bill_Bright, maybe I wrote something wrong in my first message, let me refrase it.

First score is my pc, second score is someone else with my same hardware (CPU and GPU), exept from the RAM as I write it was my main concerns beacuse of how Infinity Fabric afffect the overall performance.

In here I'm asking how can I get more combined score as they are pretty low compared to others score, also PCs that get less single scores than mine.
As you mentioned Windows version, are you aware if different build can get you different scores? Is there something to tweak?

Thanks

 

[EarthDog]

What memory speeds are you (both) running? One is 3200 Mhz... the other?

 

[Bill_Bright]

And the vast majority of those doesn't really affect the GPU score.

True - sorta. But they affect the performance of Windows itself and of the benchmarking program.

Bill_Bright, maybe I wrote something wrong in my first message, let me refrase it.

First score is my pc, second score is someone else with my same hardware (CPU and GPU), exept from the RAM as I write it was my main concerns beacuse of how Infinity Fabric afffect the overall performance.

So these are two different computers! You say same hardware, do they even have the exact same brand and model motherboard with the exact same revision numbers and exact same BIOS versions?

Do both power supplies output the exact same +12.00V, +5.00V and +3.30V? Or does one output +12.27V, +5.21V, and +3.42V while the other PSU outputs +11.61V, +4.87V, and 3.18V?

As EarthDog correctly points out, the different areas may play "a bit" here and "small role" there that, individually result in negligible differences. But it is common for the "additive effects" of very slight variances to add up to something significant, even when each variable is well within allowed tolerances.

These are two different computers with very similar results. That's a good thing!

 

[EarthDog]

True - sorta. But they affect the performance of Windows itself and of the benchmarking program.

They don't. Honest. THis isnt TPU...lol. OCF is a place where overclocking, benhmarking competitively and using LN2 is common and is not shunned. The knowledge base here on these type of things is exponentially higher (but that doesn't mean we can't learn or be corrected).

The fact both PSUs could output different values (which is a given) has nothing to do with performance. So long as they are within spec (which needs to be measured by a MM of some sort as you know) it won't matter. It does not matter that two PSUs are exactly the same voltage wise.

There is literally 2 things of your list which can affect scores... system RAM speed differences, and potentially windows version. The clockspeed difference can play a role in combined score, but not terribly much in GPU score.

 

[Bill_Bright]

The point, which you agreed with BTW, is slight differences can result in slightly different scores. That is does not mean something is wrong with one or the other.

 

[EarthDog]

Let's sharpen the point a bit...

These differences aren't slight though. 6.7K to 8.6K combined... is significant. It may be due to memory speeds or the CPU speed difference (though it should show in Physics score as that is CPU based), potentially Windows, but it has nothing to do with RAM capacity, PSU voltage, HDD/SSD, or slight GPU speed differences.


I was thinking it could be a Spectre/Meltdown thing (one PC is patched, one isn't). But that would show in Physics as well and this guy has both faster Physics and Graphics, but that combined is lower which is odd. Its either stability, or RAM speed if I had to put two guesses on it. THe user can lower the GPU clocks (both mem and core) and see if that helps. If not, do the same thing to to the CPU.. either lower clocks or add a bit of voltage to increase stability.

 

[Bill_Bright]

These differences aren't slight though. 6.7K to 8.6K combined... is significant.

Yes but due to what? We don't know. You say "it may be" due to this or it may be due to that. I say it "may be" due to the cumulative effects of this + that + that other thing too! Again, these are two different computers where we still don't know if they are using the same motherboard!

Plus when looking under the Run details, it is 20551 vs 19994 with both rated at 95% of all results. That's not a significant difference.

IMO, there are too many unknown variables to draw any conclusions at this point in time.

 

[EarthDog]

That is your opinion and you are entitled to it. But the point remains that RAM capacity, PSU voltage, HDD/SSD, or slight GPU speed differences do not play a role. At all.

RAM Capacity - The test doesn't use close to 8GB so having 16/32GB doesn't matter.
PSU Voltage - that they don't match isn't remotely relvant to performance.
HDD/SSD - No test here has anything to do with the HDD.
GPU Speed - the difference is so small that run variance would eat up 30 MHz (of 1700+) and 5 MHz on RAM. That would also show in the graphics score, but it is higher.
Motherboard - Ive tested dozens across all platforms and the differences between them at the same clock (to isolate the mobo only assuming like systems which is what I have) the difference is run variance.

The "signficiant difference" is in the COMBINED score (as I stated already - you went with overall)... what the OP specifically asked about.

Its something, I just know it isn't those 5 things and there is lower hanging fruit to attack . Just trying to help the OP focus a bit more on the parts that matter instead of a potential wild goose chase.




Anyway, waiting for the OP to test... I think we TPUd this place up enough.

Either the CPU or GPU being a bit unstable can really affect the combined score

THe user can lower the GPU clocks (both mem and core) and see if that helps. If not, do the same thing to to the CPU.. either lower clocks or add a bit of voltage to increase stability.