Geekbench 4 result not possible

[Ace123]

The link result is not possible with a 16C CPU at the reported GHZ. I believe this is a bugged run. The result would put it in 3rd place for an 18C cpu with two more cores, 4 more threads. This isn't possible.

https://hwbot.org/submission/4209331_mllrkllr88_geekbench4___multi_core_core_i9_9960x_82217_points

[Leeghoofd]

and what clocks does that 18C cpu run at ? What memory settings,, was the 18C run an efficient run or just a run?

So BenchMate didn't detect the bugged run?

[_mat_]

The result was done with BenchMate 0.8, which didn't have child process tracking of timer calls yet. So there is a small chance for time skewing, which could go by undetected.

That said there are lots of Win 10/Zen 2 scores on the bot that are not using BenchMate, which is against the rules and even more prone to undetected skewing. So I would give this one the benefit of the doubt as long as nobody else achieves nearly 6G on a 3900X in Geekbench 4.

Edited September 12, 2019 by _mat_

[Noxinite]

Lemme just repaste from the report: "Comparing the scores via the GB3 links, show this score is faster across most sub-tests. There is not one (or other) sub-test that is massively out of line." So it is unlikely to have bugged. It would have had to bugged the timer pretty crazily to get such a big boost.

 

Edit: Although, I haven't looked at the other Intel GB4 scores, so maybe it is out of line.

Edited September 12, 2019 by Noxinite

[keeph8n]

I don’t know that is a bugged run. 

I haven’t done GB4 enough to know what 16 core scores are like. I know there is a 2950X run done with Linux that scores way above what it should at the speeds it is at. 

 

I can throw the 2950X back in after 2970WX testing to see how far up I can push the 16 core inn gb4 if you would like. 

[_mat_]

28 minutes ago, Noxinite said:

So it is unlikely to have bugged. It would have had to bugged the timer pretty crazily to get such a big boost.

The timer used in this score is HPET btw.

[bigblock990]

Nothing to see here. Both gb3/4 get a boost depending of win10 version. Same shakeup happened in gb3 18c ranking recently.

[Splave]

^ what he said

 

Attached the difference on 7980x of proper windows version boost ? I appreciate your openness to discuss results that you think are flawed though. That is the only way the system can work. 

[keeph8n]

Linux can also give a large boost

[yosarianilives]

Result looks in line to me, it's about 20k above a 7960x (pasted version of 9960x), but the 7960x score is at 5 ghz vs the 9960x at 5.9 ghz. Also the 9960x is at 1800 mhz mem with much tighter timings vs the 7960x is at 1700 mhz mem with fairly loose timings. I think on clocks alone you could make the difference even before OS effi.

If there's anything "wrong" with this sub it's that it's a benchmate score which iirc has it's own ranking, but I don't know if the older version of benchmate would mean the score should get moved or not.

[_mat_]

 

5 hours ago, Leeghoofd said:

So BenchMate didn't detect the bugged run?

I want to be very clear on this: The only thing that's in the way of proofing that this score is 100% valid, is the fact that it was made with BenchMate 0.8. Which is still about 1000% better than any other score made on Zen 2/Win 10, that still has its points enabled.

With BenchMate 0.9.x I can say without any doubt whatsoever, that every single timer call of the benchmark is tracked, emulated as HPET (if it doesn't harm performance) and additionally validates HPET against other time sources. BenchMate 0.8.x and before could do that in Geekbench.

I think we can agree here that this score is not out of line but uses the right combination of OS and CPU to achieve the score.

Last but not least, I hope that one of the bot devs can finally get me access to configure the benchmarks for the submission API. It's time to put BenchMate to good use.