9900k vs 7820x...Timespy Results

[TahoeDust]

If these results that popped up on 3dmark are real, pretty interesting. Both CPU's all 8 cores at 4.8GHz with very similar hardware..

https://www.3dmark.com/compare/spy/4374217/spy/2216374#

 

[LTC8K6]

Clock for clock we'd expect close scores. Both are basically Skylake, with cache differences.

800mhz overclock on the 7820X, 100mhz overclock on the 9900K.

So I think in practical use, the 9900K is going to have a healthy lead on the 7820X due to easily clocking much higher.

 

[csbin]

9900K@4.8Ghz vs 2700X@4.45Ghz...

 

[Shmee]

My understanding was that the 7820x was not great in gaming though, but I guess timespy is able to make better use of the Skylake X architecture? I thought clock for clock, the HEDT was generally considered not as fast? I guess we will see how the 9900k compares in actual games, but it will be a wait.

 

[DrMrLordX]

My understanding was that the 7820x was not great in gaming though, but I guess timespy is able to make better use of the Skylake X architecture? I thought clock for clock, the HEDT was generally considered not as fast? I guess we will see how the 9900k compares in actual games, but it will be a wait.

A lot of it has to do with mesh speed. Overclock the mesh speed on Skylake-X and it can (sometimes) perform better in games than stock benchmarks and even some OC benchmarks would indicate. Skylake-X still has a cache setup generally more targeted at non-gaming tasks.

 

[Shmee]

That was my understanding. Based on this, I would expect the 9900k to be quite a bit faster in games still, and thus timespy as an indication of game performance would be perhaps disingenuous.

 

[Hitman928]

That was my understanding. Based on this, I would expect the 9900k to be quite a bit faster in games still, and thus timespy as an indication of game performance would be perhaps disingenuous.

Timespy CPU tests are basically pure physics tests, so more compute oriented than a real gaming scenario.

There also becomes a point where the number of cores becomes too much for a ring bus architecture and the ring bus starts to bottleneck the cpu. I don't think that will happen at 8 cores, but that's the whole reason the mesh was created to begin with.

 

[DrMrLordX]

Timespy CPU tests are basically pure physics tests, so more compute oriented than a real gaming scenario.

There also becomes a point where the number of cores becomes too much for a ring bus architecture and the ring bus starts to bottleneck the cpu. I don't think that will happen at 8 cores, but that's the whole reason the mesh was created to begin with.

Might be interesting to see how the 9900k fares on core latency tests. Though we've seen Xeons and HEDT CPUs with more cores than that (6950X has 10 cores and is not a mesh CPU).