Benchmark Lists

[crashtech]

Hi,

I am having a lot of fun testing my system with Performance Test, and comparing different CPUs in the Benchmark list. So thanks for providing these fascinating looks at comparative system performance.

I hope it would not be out of line to make a suggestion. I'd very much like to be able to view an Overclocked Single Threaded CPU list, and it would be even more fantastic if the clock speed at which the score was achieved was listed. Right now the overclocked list gives only the stock speed of the CPU.

Thanks again for the interesting opportunity to compare so many different pieces of hardware so easily!

[David (PassMark)]

There is already an Overclocked CPU chart and a single threaded chart. So an estimate could be of overclocked single threaded performance from this data. I think it might be a bit niche to have a new chart dedicated to this however.

[crashtech]

There is already an Overclocked CPU chart and a single threaded chart. So an estimate could be of overclocked single threaded performance from this data. I think it might be a bit niche to have a new chart dedicated to this however.

You are absolutely right. Is there any way to discover at what speed the overclocked score was achieved?

[David (PassMark)]

Not easily as the results on the web sites are averages from 1000s of machines and because there are different overclocking options. e.g. base overclock clock or turbo overclock.

[crashtech]

Not easily as the results on the web sites are averages from 1000s of machines and because there are different overclocking options. e.g. base overclock clock or turbo overclock.

Of course, I don't know why I hadn't put that together before now. The Overclocked score is then a bit of a catchall of every different achievable clock speed, and the Turbo multiplier just serves to throw another wrench in the gears.

I would like to be able to infer performance relative to clock speed within a given CPU type, but scores don't seem to scale linearly with clock speed, even with very similar CPUs, at least not without some other factor(s) that may be unique to each µarchitecture.

Thanks for taking the time to answer my questions!

[crashtech]

I noticed today that the E3-12xx V3 series CPUs are not listed in the "CPU Mark by Socket Type : LGA1150." It would be nice to see those in there if possible.

[Richard (PassMark)]

I noticed today that the E3-12xx V3 series CPUs are not listed in the "CPU Mark by Socket Type : LGA1150." It would be nice to see those in there if possible.

Thanks for bringing this to our attention. I have updated the socket type for the E3-12xx V3 series and they should be appearing on the socket LGA1150 chart.

[crashtech]

Wow, thanks!

[crashtech]

Forgive me for troubling you again, but it appears that at least two of the new Haswell processors (i3-4130, i3-4330) have appeared in the multithreaded CPU test rankings, yet they are not properly identified as LGA1150.

[Richard (PassMark)]

We try to gather as much details from the CPU themselves from within PerformanceTest, however, some of the details, i.e. supported sockets or TDP, are gather by scripts that scrape the pages on Intel's website. The scraper does a reasonably good job on gathering this information, however sometimes we do have to manually enter in values.

In this case, it seems that Intel has changed the formatting on their CPU pages, causing the scraper to fail to locate the socket types. I have updated the scraper and it should now correctly pick up the socket type.

[crashtech]

We try to gather as much details from the CPU themselves from within PerformanceTest, however, some of the details, i.e. supported sockets or TDP, are gather by scripts that scrape the pages on Intel's website. The scraper does a reasonably good job on gathering this information, however sometimes we do have to manually enter in values.

In this case, it seems that Intel has changed the formatting on their CPU pages, causing the scraper to fail to locate the socket types. I have updated the scraper and it should now correctly pick up the socket type.

Interesting. I just assumed such info was manually entered as CPUs became available. An automated method is a labor saver for sure. I did notice that some of the Ark pages look a bit different now.

[crashtech]

I hope it is OK for me to bring this old thread to the top (is necro forbidden here?) but I have noticed an anomalous entry, an "i5-3170K," a fake CPU name, occupying a spot almost at the top of the single-threaded performance chart! Somebody is playing around!

[David (PassMark)]

While not impossible, it is fairly hard to fake the CPU name.
So it might well be a real CPU. Over the years we have seen a number of unofficial CPUs and engineering prototypes pop up in the charts. The problem however is that the result that we have for this CPU are massively overclocked. But because it isn't an official CPU (at the moment) we don't have official specifications for it. Not having any official specs makes it hard to determine what the factory turbo speed is (before overclocking). This makes it hard to know if it has been overclocked as no ones knows what the base turbo clock is.

For the moment we have put in a dummy turbo value for the i5-3170K and this will force the result into the overclocked charts and out of the single threaded charts.

If it turns out this is a fake, then we'll just delete it. The benchmark numbers look believable however.

While looking at this, it seems we have a related problem with the new Pentium G3258 CPU. It seems people are able to overclock the turbo speed on it, when in theory it shouldn't have any turbo function at all. So we'll need to investigate that as well.

[crashtech]

Hi David,

The G3258 is the new 20th Anniversary Pentium, which features an unlocked multiplier. I think the results for it are probably legitimate, but it most likely belongs in the overclocked chart. It has pretty good ST performance when overclocked.

There's no mention of a 3170K on the Intel site at all. Being a bit of an aficionado, I think I would have heard of what would be an unlocked Ivy Bridge of some sort other than the common 3570K.

[David (PassMark)]

As best we can tell the G3258 is badly setup by Intel. Internally the CPU reports that it has turbo available, even though the marketing says it doesn't.

This isn't the only CPU that we have results for that isn't listed on the Intel site.

For example we have the, Intel E2697V, Xeon W5647, Intel 000 & Intel 0000. The last two are common names for Intel engineering samples.

Plus a few rare AMD units like the "AMD Eng Sample Mobile 2185_A1" and the "Phenom X2 Dual-Core GE-6060"