Low Allocate Small Block

[vexellent]

Hello there, I just got my computer brand new a few days ago from DELL. Out of curiosity I wanted to see how my specific desktop stacked up with 6 others having the same CPU and GPU as me. All the systems seem quite equal in terms of score, except for my allocate small block score. What causes this drop in my score? Is it faulty memory? Although in my comparison there are different types of ram for each of the systems, my ram is the only one with low allocate small block score. What can I do?

[passmark]

What was your benchmark result for this test?
What type of CPU & RAM do you have?
Are you running 32bit or 64bit O/S?
What version of PerformanceTest are you using, and is it the 32bit of 64bit release?

[vexellent]

My computer (AMD Phenom II X6 1055T; 8GB RAM @ 1333 MHz, 2 GB in 4 physical slots; windows 7 64 bit, performancetest is running in 64 bit)

****I used task manager to check that performancetest is running at 64 bit (it doesn't have the *32 after the process name)


Passmark rating: 1857.2
CPU mark: 5422.2
2D Graphics mark: 345.8
3D Graphics mark: 2500.6
Memory mark: 1250.5 --> here the most fluctuation in scores occurs

Allocate small block mark: 1888.3, the other 6 computers all have marks 4500+
Read cached: 1405.2, other computers have about the same
Read uncached: 1290.5, other computers have about the same
Memory write: 1440.8, other computers have about the same
Memory Large RAM: 3428.7, here my computer has one of the higher scores which all approximately lie in the 3400 area and I confirmed that the high scoring computers in this area all have 8 GB RAM; the low scoring computers have around 1500 and all have 4 GB RAM or less

Disk mark: 946.9
CD mark: 346.8

[passmark]

The results for the "Memory Large RAM" test are easy to explain. You get better results in this test if you have more RAM. And your 8GB is above the average.

The "Allocate small bloc" result is low. But harder to explain. As you say, other machines with similar specs do better. If you reboot, and wait for the machine to be idle, then run just this individual test from the menus half a dozen times what results do you get.

[wonderwrench]

Since you can use a Phenom II X6 in both AM2+ and AM3 motherboards memory performance would surely have more variance I would guess?

Bill

[vexellent]

I forgot to mention that i have using PerformanceTest 7.0 Evaluation version. I have done the test several times using short to very long test duration with the same score for memory allocate block. Rebooting didn't help.

In my comparison, two other Dells have the same motherboard model and BIOS version, and yet their allocate small block score is higher and the same with regards to each other.

[wonderwrench]

I forgot to mention that i have using PerformanceTest 7.0 Evaluation version. I have done the test several times using short to very long test duration with the same score for memory allocate block. Rebooting didn't help.

In my comparison, two other Dells have the same motherboard model and BIOS version, and yet their allocate small block score is higher and the same with regards to each other.

Are they using the exact same CPU you are?

[vexellent]

They are all using AMD Phenom II X6 1055T and ATI HD Radeon 5870.

[vexellent]

Is there no explanation for my case?

[brojo]

I have the same exact issue. Asus M4A89GTD PRO/USB3, AMD 1055T, 4GB Corsair CMX4GX3M2A1600C7. All results are on par with other ASUS/AMD 1055T combos except the Allocate Small Block results. Mine is 2003 compared to: 4414, 4481, and 4335.

[passmark]

I had a look at dozens of results for the 1055T. I found about 1 in 15 machines show low performance on this test. But I can also find a similar effect on some other CPUs. (e.g. 1 in 10 to 1 in 20, look low compared to similar results from other CPUs of the same type).

Didn't see any pattern however. Doesn't seem to be O/S related, nor motherboard related, nor related to amount of RAM.

So I don't have any real explaination.

Could be BIOS settings, RAM type, 3rd party software, or some other factor. Don't know.

[passmark]

Further speculation.

The allocate small block test measures the time require to allocate system RAM then set the values in that RAM to zero, and then de-allocate the RAM. This is done using the calloc & free API calls.
http://msdn.microsoft.com/en-us/library/3f8w183e(VS.71).aspx (http://msdn.microsoft.com/en-us/library/3f8w183e%28VS.71%29.aspx)

But it seems in some implementations of calloc, there might be the possibility of the API call be optimised to pick up pre-zeroed RAM from the operatng system.

There is a background task that runs at very low priority on Windows called the "zero-page thread". The zero-page thread is a system thread responsible for zeroing any free pages when there are no other threads that need to run.
http://msdn.microsoft.com/en-us/library/ms685100(VS.85).aspx

The use of zero pages is a security requirement to prevent processes being able to read other processes' data that might be still held in RAM after being used by another process. If the zero page list is exhausted, Windows will use the free list and zero the page on demand

So I am speculating that in most cases a call to calloc would pick up pre-zeroed RAM, and then recognising the RAM as being zeroed, not bother to zero it during the API call itself. This assumes you system is not short on free RAM and has been idle for a while.

So systems that don't have much free RAM and / or are not often idle might get lower scores.

The test uses 200 blocks of around 800 bytes (160KB total) during the text loop. So not much RAM is required for the this test.