3d mark cpu score

[DominicNikon]

Hello, since amd is getting a lot better in the cpu space with Ryzen 3rd gen,  when do you think a amd cpu will be on or if ever the 3dmark hall of fame for like time spy in the top 10? I know there are some 3900xs and 3950xs within the top 80 but traditionally the top 10 has been either the xeon w-3175x or i9 9980XE. With intel's higher overclocking headroom and higher ipc will amd be able to beat intel in this area? 

[Jurrunio]

3990X? 64 core vs measly 28 core, Clocking higher won't help at that point, unless intel decides to compete with 8 socket monsters

[DominicNikon]

41 minutes ago, Jurrunio said:

3990X? 64 core vs measly 28 core, Clocking higher won't help at that point, unless intel decides to compete with 8 socket monsters

In 3d mark the 3990x performs worse then the 3950X, i'm assuming this is based on IPC and the 3950xs ability to clock higher. Both resaults are the number 1 spot for timespy with the cpu and the same duel RTX 2080tis (cards clocks are different) 

[Jurrunio]

10 minutes ago, DominicNikon said:

In 3d mark the 3990x performs worse then the 3950X, i'm assuming this is based on IPC and the 3950xs ability to clock higher. Both resaults are the number 1 spot for timespy with the cpu and the same duel RTX 2080tis (cards clocks are different) 

Are they both LN2 results? Maybe 3Dmark just doesnt scale to that level.of core.counts then

 

@ShrimpBrime thoughts?

[Guest]

25 minutes ago, Jurrunio said:

Are they both LN2 results? Maybe 3Dmark just doesnt scale to that level.of core.counts then

 

@ShrimpBrime thoughts?

Tons of thoughts, but there's tons of variables between the systems as well. 

 

So 16 threads vs 64. This means the benchmark thread scaling doesn't go beyond maybe 8 - 12 threads depending also how the system is handling the threads at the time.

If one rig has core parking enabled, it would suffer score from core unpark lag times. So having additional threads wouldn't really matter to a certain extent.

 

The benchmark is heavily IPC and clock speed dependent.  

 

Since I cannot see the entire overclock details on each system, I cannot say why one would be faster than another when I could tweak either system identical and come with a different score just using different GPU driver versions.

 

But from what we CAN see above:

 

The 3950X looks like it's an all core static (manual) overclock @ 4.5ghz. 

The 3990X looks to be auto boosting. Probably has windows in balanced mode, core parking ect. 

The 2080ti's look to have a favor in clock speed on the 3950X which is running performance mode.

 

Other than that, there could be a vast difference in Memory frequency, timings, Infinity Fabric and so forth and on ward.

 

There's probably more we could touch on, but that's a good start I think.

 

 

[Jurrunio]

3 minutes ago, ShrimpBrime said:

If one rig has core parking enabled, it would suffer score from core unpark lag times. So having additional threads wouldn't really matter to a certain extent.

Windows's thread scheduler could also throw a spanner at the works, there could be so many cores that it's shuffling tasks needlessly, wasting its time

[Guest]

Just now, Jurrunio said:

Windows's thread scheduler could also throw a spanner at the works, there could be so many cores that it's shuffling tasks needlessly, wasting its time

Yep. Notice also the 3950X rig had Simultaneous Multithreading Disabled. 16c16t. That run was made to achieve a high as possible score in my opinion. The other was just somebody with a regular gaming setup and submitted a benchmark. Very poor for comparisons really.