Dual CPU mb frequency not recognized

[Matsglobetrotter]

Hi, im playing around with some dual CPU mb. I noticed that when submitting a system with 2 overclocked cpus with cpuz validation the system reports it with 0 points as its reported under the 1 cpu section.

I have a higher single CPU score as its with a different MB where i can adjust BCLK. 

Meanwwhile it is ofcourse a very different complexity overclocking 2 CPU's in one system.

https://hwbot.org/submission/4613634_matsglobetrotter_cpu_frequency_2x_xeon_e5_2678_v3_3199.25_mhz/

I also note the same issue on  Pifast and Superpi.  Indeed these as single thread banechers but it should be consistent for same reason as above. Seemingly same also on Geekbench one core scores

 

Edited November 27, 2020 by Matsglobetrotter

[Antinomy]

1 hour ago, Matsglobetrotter said:

I also note the same issue on  Pifast and Superpi.  Indeed these as single thread banechers but it should be consistent for same reason as above. Seemingly same also on Geekbench one core scores

Just don't use dual CPU setups for single-thread tests. Simple as it sounds.

[Matsglobetrotter]

3 minutes ago, Antinomy said:

Just don't use dual CPU setups for single-thread tests. Simple as it sounds.

Thanks. Do note that scores have been posted for the same in the past by others. Even if i have higher scores now it does not get recognized.

I believe the challenge is to get any hardware we have to perform better thats why we oc :-) 

I guess this is also why we in some cases have dual ram slot competitions as its easier to OC a single ram than 2 or 4.

Thus i must beg to differ.

[Antinomy]

44 minutes ago, Matsglobetrotter said:

I believe the challenge is to get any hardware we have to perform better thats why we oc ?

Yes that's why, for CPU-Z, Super Pi, Pifast you can disable unnecessary cores to improve your result or put a Xeon in a desktop motherboard to push it higher than it would be in a dual setup.
Though I like the idea of special validation for multiCPU configurations but I think that wPrime/CB is enough for this.

[Matsglobetrotter]

1 hour ago, Antinomy said:

Yes that's why, for CPU-Z, Super Pi, Pifast you can disable unnecessary cores to improve your result or put a Xeon in a desktop motherboard to push it higher than it would be in a dual setup.
Though I like the idea of special validation for multiCPU configurations but I think that wPrime/CB is enough for this.

Thanks Antinomy. If that is the decision can then the multi core uploads already getting scores be removed? and can we make it clear that such is not applicable. I have significantly better scores here but they are not recognized.

Edited November 27, 2020 by Matsglobetrotter

[Antinomy]

6 hours ago, Matsglobetrotter said:

I have significantly better scores here but they are not recognized.

It would be better if you've put links to all the results but I got you. You mean you have better results in both x1 CPU and x2 CPU but since your x2 CPU is lower, it's hidden and in the x2 rankings a lower result is shown. This is an interesting bug and I believe that the right way is to show your x2 result with zero points awarded.
I'll try to report this bug but it's a long to-do list.

[Matsglobetrotter]

16 minutes ago, Antinomy said:

It would be better if you've put links to all the results but I got you. You mean you have better results in both x1 CPU and x2 CPU but since your x2 CPU is lower, it's hidden and in the x2 rankings a lower result is shown. This is an interesting bug and I believe that the right way is to show your x2 result with zero points awarded.
I'll try to report this bug but it's a long to-do list.

Thanks Antinomy. I fully understand the bug list is long. I am not going to get big scores on this one but this was also the reason people went crazy ocing the SR-2 Classified in the days (still want to play with those one day) that even today costs over 700 USD. Simply to see if it is possible to sync to separate CPUs in OC. So the resolution for me is not urgent. I flag what I see as i keep playing with hardware

 

Intel Xeon X5670 @ HWBOT

the links to my subs are below

This first one is 3199 Mhz with 2 OC'd Xeons thus dual socket card

Matsglobetrotter`s CPU Frequency score: 3199.25 mhz with a Xeon E5 2678 v3 (hwbot.org)

This on is one single Xeon in a different MB single socket mb

Matsglobetrotter`s CPU Frequency score: 3517.76 mhz with a Xeon E5 2678 v3 (hwbot.org)

 

pifast dual socket

Matsglobetrotter`s PiFast score: 20sec 550ms with a Xeon E5 2678 v3 (hwbot.org)

pifast single socket

Matsglobetrotter`s PiFast score: 18sec 750ms with a Xeon E5 2678 v3 (hwbot.org)

 

Superpi dual socket 1m

Matsglobetrotter`s SuperPi - 1M score: 11sec 264ms with a Xeon E5 2678 v3 (hwbot.org)

Superpi single socket 1m

Matsglobetrotter`s SuperPi - 1M score: 10sec 338ms with a Xeon E5 2678 v3 (hwbot.org)

superpi dual socket 32m

Matsglobetrotter`s SuperPi - 32M score: 10min 6sec 716ms with a Xeon E5 2678 v3 (hwbot.org)

superpi single socket 32 m

Matsglobetrotter`s SuperPi - 32M score: 8min 49sec 14ms with a Xeon E5 2678 v3 (hwbot.org)

Geekbench 3 dual socket

Matsglobetrotter`s Geekbench3 - Single Core score: 3143 points with a Xeon E5 2678 v3 (hwbot.org)

Geekbench 3 single socket

Matsglobetrotter`s Geekbench3 - Single Core score: 3599 points with a Xeon E5 2678 v3 (hwbot.org)

 

 

[Antinomy]

You kind of miss the big picture heree. Let's go back in the days of P3 and AMD Athlon. Why would anyone want to buy a dualCPU config instead of a single CPU? It costs around twice more and wasn't the best choice back in the days in most cases. In most but not all of them. Why people have bought dualCPU setups was multithread performance - rendering, video editing, scientific stuff and all of this. That's where and why those were used. Nobody brought them for single thread tasks.

So running Pifast on a dual socket setup may be curious but has no value at all. You have Cinebenches, wPrimes, x265, GPUPI and all sorts of stuff for that.
The reasons people were crazy about SR-2 was globals for multithread tests, don't make a mistake on this one too. Both some 2D and 3D gained from second socket.

Edited November 28, 2020 by Antinomy

[Mr.Scott]

I have benched many dual socket boards. What Alex says is correct. Don't even waste your time on anything single threaded.

[Matsglobetrotter]

15 hours ago, Mr.Scott said:

I have benched many dual socket boards. What Alex says is correct. Don't even waste your time on anything single threaded.

no worries Scott. Then the scores for those who have obtained such should be removed.

[Strunkenbold]

Dunno if we still talk about the same thing? Why removing those points?

There are long standing bugs, first one is not really a bug but could be made a little bit more elegant:

1. For single core benchmarks, it doesnt matter how many CPUs you have. So those "Single Socket", "Dual Socket tabs should be only visible for multi threaded benchmarks. Maybe we can group those benchmarks on the CPU overclocking records page differently? 

2. If you click the benchmark ranking of single threaded benchmarks, Number of cores should be "Any" by default.

[Strunkenbold]

On 11/27/2020 at 1:23 PM, Antinomy said:

It would be better if you've put links to all the results but I got you. You mean you have better results in both x1 CPU and x2 CPU but since your x2 CPU is lower, it's hidden and in the x2 rankings a lower result is shown. This is an interesting bug and I believe that the right way is to show your x2 result with zero points awarded.
I'll try to report this bug but it's a long to-do list.

Since the score of the best 2x result doesnt matter if there is already a better result with only one CPU and the ranking itself for 2x CPU is totally irrelevant for single threaded benchs, I wouldnt call that a bug. In the end its a unified ranking handling out points equally to all competitors.