Disk scores in Performance Test 7

[jimbonbon]

Hi everyone!

Currently doing some benchmarking using PerformanceTest 7, but have one key wondering about the disk testing...

Looking at the top ever result as an example (although there are others similar), the disk scores are INSANELY high, showing read/write speeds of a few thousand MB/s. Is this a glitch or can anyone tell me what kind of storage device can show scores like this? The disk shows as 'Removable'.

Any input appreciated!

James

[wonderwrench]

Are you looking at the hard drive bench mark scores on line here or the overall disk score in Performance test 7 it self. Either way the scores are not in MB/s but relative performance. Only the disk sub scores are in MB/s. The fastest drives are "Solid State Drives" SSD's for short. SSD's are many times faster than magnetic drives and are a worth while upgrade to most any system. The only downside is cost per gig. SSD's are so much faster that an older PC with a SSD will feel faster than a brand new top of the line PC without one.

[jimbonbon]

Are you looking at the hard drive bench mark scores on line here or the overall disk score in Performance test 7 it self. Either way the scores are not in MB/s but relative performance. Only the disk sub scores are in MB/s. The fastest drives are "Solid State Drives" SSD's for short. SSD's are many times faster than magnetic drives and are a worth while upgrade to most any system. The only downside is cost per gig. SSD's are so much faster that an older PC with a SSD will feel faster than a brand new top of the line PC without one.

Hi there. and thanks for the response...

I currently run SSD's in my gaming machine, but am on laptop at the moment - below is a screenshot of what i mean. Obviously the overall disk score is a 'weighted' number based on the tests, but if you look at the read/write and cache scores these are listed as MB/s. For my laptop and a couple of the other reference computers I have randomly selected, the scores make sense and are the kind of speeds you would expect. But you can probably see the two I am referring to that seem odd




J

[David (PassMark)]

These number are too high for a SSD connected via SATA.

They are what you would get from a pure RAM drive. If you look at the system information for these baseline files I would think that the drive doesn't have any model number and it fairly small. Maybe only 1 -4 GB in size.

[wonderwrench]

I see, just for fun see what is reported in the system tab. Drive size model etc? I would guess people have figured out a way of creating a ram drive that can be tested as a physical hard drive. This has been done with optical drives for a long time. Another possibility is some type of very high speed external storage. I would guess these submissions are filtered out of the online results so no big deal really but it would be cool to know how it was done. Its to bad the baselines contain no info in their notes.

Edit: you beat me to it!

[jimbonbon]

These number are too high for a SSD connected via SATA.

They are what you would get from a pure RAM drive. If you look at the system information for these baseline files I would think that the drive doesn't have any model number and it fairly small. Maybe only 1 -4 GB in size.

See thats what I thought - it smacks of either LOADS of SSD's (see this from last year http://www.tomshardware.co.uk/samsun...ews-30563.html) or potentially a RAM drive.

Insane either way

I see, just for fun see what is reported in the system tab. Drive size model etc? I would guess people have figured out a way of creating a ram drive that can be tested as a physical hard drive. This has been done with optical drives for a long time. Another possibility is some type of very high speed external storage. I would guess these submissions are filtered out of the online results so no big deal really but it would be cool to know how it was done. Its to bad the baselines contain no info in their notes.

My thoughts were the same - RAM drive was the only thing I could think of that could potentially get to figures like that, but never seen such stuff in reality. The lack of info left me wondering, and I was trying to come up with a rational explanation with no success.

Will install on my desktop with SSD and run to show the vast difference, plus see what other information is there for that SR2 machine.


J

[jimbonbon]

So... below is a screenshot showing a number of tests run with my Intel SSD (some with OC's to see how CPU/RAM dependant the tests are) versus the SR2 machine.

EDIT. Changed to link as this forum doesn't seem to shrink the image preview:
http://img440.imageshack.us/img440/5422/whattheb.jpg

The disk just shows as 'Removable' and is apparently about 60GB. There is interestingly a rogue RAM module in slot 10 though which makes no sense in memory configuration terms...


J

[wonderwrench]

So... below is a screenshot showing a number of tests run with my Intel SSD (some with OC's to see how CPU/RAM dependant the tests are) versus the SR2 machine.

EDIT. Changed to link as this forum doesn't seem to shrink the image preview:
http://img440.imageshack.us/img440/5422/whattheb.jpg

The disk just shows as 'Removable' and is apparently about 60GB. There is interestingly a rogue RAM module in slot 10 though which makes no sense in memory configuration terms...


J

I just noticed we have the same SSD's. Your HD scores seem way low. I would guess its a sector offset problem or your not running your SSD on the Intel SATA 2 controller. See this thread for more info. http://www.passmark.com/forum/showthread.php?t=2748
Start at post #4

[David (PassMark)]

There is interestingly a rogue RAM module in slot 10

At the moment we collect the RAM stick information from SMBIOS. But these values are often wrong or incomplete because the BIOS sets the values wrongly (and doesn't have much incentive to get it right).

For the next major release we are going to do direct access on the SMBUS to get the SPD data from the sticks themselves. Which will be more accurate, but a bit more risky in terms of compatibility.

[David (PassMark)]

There are lots of RAM disks on the market. Try Googleing,
Dataram RAMdisk
Superspeed RamDisk
QSoft RAMDisk
Everstrike DiskBoost
Microsoft KB Article 257405

There are also some benchmarks here,
http://www.raymond.cc/blog/archives/2009/12/08/12-ram-disk-software-benchmarked-for-fastest-read-and-write-speed/

[David (PassMark)]

Just adding to this, we now also have our own free solution for creating a RAM disk.

PassMark OSFMount will allow the creation of RAM disks,
http://www.osforensics.com/tools/mount-disk-images.html

It isn't the main purpose of the tool, but it is an interesting additional function of the tool.

Here are my results for a 600MB RAM disk and the PerformanceTest disk tests.

PassMark(TM) PerformanceTest 7.0 (http://www.passmark.com)
Results generated on: Wednesday, 20 October 2010


Benchmark Results

Test Name: This Computer
Disk - Sequential Read: 2521.3 MB/sec
Disk - Sequential Write: 1307.0 MB/sec
Disk - Random Seek + RW: 1334.5 MB/sec
Disk Mark: 18671.2

System information: This Computer
CPU Manufacturer: GenuineIntel
Number of CPU: 1
Cores per CPU: 4
CPU Type: Intel Core2 Extreme X9650 @ 3.00GHz
CPU Speed: 3007.6 MHz
Cache size: 6144KB
O/S: Windows Vista (64-bit)
Total RAM: 4093.6 MB.
Available RAM: 1699.6 MB.
Video settings: 1600x1200x32
Video driver:
DESCRIPTION: ATI Radeon HD 3670
MANUFACTURER: ATI Technologies Inc.
BIOS: 113-AB42400-100
DATE: 8-13-2009
Drive Letter: E
Total Disk Space: 0.6 GBytes
Cluster Size: 4.0 KBytes
File system: NTFS

As expected the numbers are insanely high, and the random access result is equal to the write result. So there is no performance penalty for non sequential access.

It is just shame that any data on the RAM drive is lost when the machine is powered down.