Erronious results in PC Benchmark database ???

[theJaybo]

Ive got a pretty good system...

System - 5199
CPU - 11423
G2D - 972
G3D - 4640
MEM - 3159
SSD - 4207

And this is Overclocked CPU/VID/RAM.

This is on par with the top 20 systems (Im number 17) with the i7-3820 cpu I have.


But looking at the top systems over every processor, seeing CPU results in the 16,000-22000's and DISK results in the 60,000's even with RAID 0 am I to believe these results? What could account for some systems (albeit with processors that cost 5x as much as mine such as the i7-3930k) getting DOUBLE what other systems are getting? This holds true for the same processors within the list (as compared to others of the same model).


Are finding a way to fudge the results?

Thanks

[David (PassMark)]

The top systems are *really* overclocked. They are typically running at around 5Ghz. Plus many of the machines with the top CPU scores are dual Xeon CPUs with 8 cores in each CPU and hyperthreading (32 threads in total, compared to your relatively puny 8 threads).

The very highest 'disk' results are often solid state RAM drives, or 3rd party caching solutions that act more or less the same as RAM drives.

[hawks]

Wish they had there own catagory and would sure be nice if we had a top 10 list for amd or intel just for non overclockers

[David (PassMark)]

Sort of like the paralympics?

[Fitch]

Ive got a pretty good system...

System - 5199
CPU - 11423
G2D - 972
G3D - 4640
MEM - 3159
SSD - 4207

And this is Overclocked CPU/VID/RAM.

This is on par with the top 20 systems (Im number 17) with the i7-3820 cpu I have.


But looking at the top systems over every processor, seeing CPU results in the 16,000-22000's and DISK results in the 60,000's even with RAID 0 am I to believe these results? What could account for some systems (albeit with processors that cost 5x as much as mine such as the i7-3930k) getting DOUBLE what other systems are getting? This holds true for the same processors within the list (as compared to others of the same model).


Are finding a way to fudge the results?

Thanks

What "Top 20" list are you looking at? Can you post the link? I have higher scores than you and I am not in the top 20. Thanks, Cliff

[David (PassMark)]

Top 20 List is here.

As discussed in some older posts we do exclude some results from the Top 20. Mainly those where the hardware description doesn't match the hardware the tests were run on or where they are duplicates.

[Fitch]

Thanks David. That is the list I refer to as well. My system score is only 5224 so I am way down the list.

[theJaybo]

Thanks David. That is the list I refer to as well. My system score is only 5224 so I am way down the list.

Since they dont have a searchable list online, you need to do all your view in the Performance Test program.

You do a search in the "Managed Baselines" menu, under the Advanced search tab, I chose CPU - Core i7, keyword 3820, with CPU ratings above 9,830.

My system comes in at #19 with an updated score (I reinstalled my os) of 5201.

So of all the systems with the i7-3820 Im # 19.

Does that make sense?


Even as I write this, there are the top systems with this processor (a normal desktop processor) that have disk ratings in the 20,000's with regular OCZ Vertex 4 drives. I don't understand how a normal system will bench over 5x the speed of other "normal" systems.

Are you saying (passmark staff) they are are using custom software too? Or could I be that their system is in RAID 0, and the Drive model category just inst picking up on that like it is for others?

My disk rating that attribues to my Passmark score is 3948, thats within 1% of the Passmark benchmark score for my drive on the hard drive benchmark portion of passmark.com. The OCZ Vertex 4 that is achieving a 20,000+ score is only rated 3,716 in the list. Thats over 5 times the performance. Even in raid 0, I don't think a system could achieve that.

A single drive highest score recorded is a OCZ RevoDrive M @ 8,886.

Is it possible they are fudging the results somehow? If this is a real score, what can I do to get it?

[David (PassMark)]

Scores of 10,000+ are typically RAID. To get the extremely high numbers, 20,000+ they need to be using RAM drives of one sort of another. Either an additional later of caching, or a pure RAM drive. Note that we wrote the disk test to avoid measuring the standard window's operating disk cache. But there are 3rd party caching that can't be easily avoided. In reality these 3rd party solutions are typically no better than in the in built caching in Windows. It is just that we can instruct Windows not to cache during the disk test, so you don't see the standard caching impact in the results.

Where the drive listed in the baseline obviously doesn't match up with the result, we exclude the result from the Top 20 list on the web site, but don't typically delete the baseline file. In the case of duplicates we typically deleted them as we come across them.

[nabardwell]

most of them use ram disk llc it turns a user determined amount of your memory into a hdd. with 2133 you can get disk scores of 40,000 or more