PerformanceTest Version 7.0 Beta Release

[Michael (Passmark)]

A beta for the next major version of PerformanceTest is now available for download from;
32-bit: http://www.passmark.com/ftp/petst7beta.exe
64-bit: http://www.passmark.com/ftp/petst7beta_x64.exe

Update - 19 Feburary 2009
Beta 6 now available. Changes the way the Passmark rating is calculated in order to stop extremely high values on individual sub scores from having a disproportional affect on the overall score.

Update - 13 Feburary 2009
Beta 5 now available. Few more general bug fixes. Beta 4 also had a corrupted baseline format causing a bug in reading older PT7 baselines and causing submitted baselines to be corrupt. If you are using beta 4 please move to beta 5 (or at least don't upload any baselines from it), thanks.

Update - 11th February 2009
Moved to Beta 4, few more minor tweaks, new license key system in place. This new beta should give you another 30 days to play with, we should have the final release out before this expires.
There is now also a 64-bit Beta available.

Update - 4th December 2008
Moved to Beta 3, fixes CPU scaling issues in Beta 2. PT7 CPU scores are now more in line with PT6, however they are still not directly comparable due to changes in the underlying tests.

New Features for PerformanceTest 7 include:

• New DirectX 10 3D Test (Requires Windows Vista and a DirectX 10 compatible video card)
• Large improvements to the user interface, including near completely reworked main window.
• Can now view the system information from multiple baselines simultaneously.
• streamlined baseline selection dialogue
• Increased the default size of the main window to go along with the graphical improvements.
• Improved baseline search functionality.
• Removed almost all restrictions in the trial version, other than the 30 day limit.
• Increased the amount of system information gathered and displayed.
• Improved copy options for quickly getting system information from PT to the clipboard.
• /p option added to use the application directory instead of the user directory
• Added EXPORTBASELINE script option to export .pt file
• Overhauled most of the 2D test suite. Complex Vector tests and Image filtering tests now included.
• Changed fonts and text test to have text which is more representative of a real document.
• Increased intensity of "Windows Interface" test.
• Increased the resolution at which 2d and 3d graphics tests are run to place more load on the video card.
• Prime numbers test optimised.
• String sorting test intensified to use more memory.
• Image rotation test was moved from the CPU suite and added as part of the image filters in the 2D suite to show what is happening.
• Added CPU Physics simulation test
• Added Visualised Physics Test to advanced test suite. This is a customizable 3D representation of what the new physics CPU test is doing.
• Increased unchached memory test from 4MB to 64MB as many CPU's have 4MB caches now.
• Updated image export functionality.

We welcome any feedback you may have on the changes to PerformanceTest and the beta in general.



PerformanceTest will be a free upgrade for anyone who purchases V6 now, or during the 6 months before the V7 release.

This beta release is fully functional, except for the 30 day time limit.

[Jason Pearsall]

Like the interface, a question though; did V.7.0 create a new standard for computers to compensate for developments - In other words, my computer scored a 2374 in 6.0, then a 1581 in 7.0, a 2390 in 6.0 and then a 1579 in 6.0. Is this supposed to happen?

[Michael (Passmark)]

We tried to keep the individual sub-section scores roughly the same (we want to be able to include both V6 and V7 scores together in our CPU Charts) however there were some changes to the weighting of the overall Passmark Rating. Because CPU performance has increased a lot more over the years than the other components the CPU was starting to become a very dominating factor in the overall Passmark Rating, as such in V7 we reduced the weighting a bit to compensate for this. This will usually mean that systems with fast CPU's will get a lower overall score.

As the tests have changed (somewhat dramaticly in some cases) it's not advisable to do a direct comparison between V6 and V7 baselines where possible.

Out of interest, did any of your sub-section scores change significantly (CPU, 2D, etc). As I mentioned we have tried to keep these roughly equivelent and if there were any major changes in them there might be something else causing your change of overall score.

[Michael (Passmark)]

We have now moved to Beta 3. There were some problems with the CPU mark scaling in Beta 2 that have been addressed. PerfomanceTest 7 should now give CPU scores closer to PT6, however due to underlying changes in the cpu tests the two versions will still never be directly comparable.

[Windsurf48]

Is there a 64-bit version of V7 Beta? Also, are there V7 baselines available for download?

I have XP x32, XP x64, and Vista x64 in the three primary partitions on my hard drive. After changing to a x58 motherboard and Core i7 965, I ran Performance Test 6 x32 in XP x32 and the 64-bit version in Vista and XP x64. Predictably, both 64-bit operating systems and tests produced higher scores than the 32-bit version, with XP x64 200-600 points higher than Vista x64 at the same BIOS settings. I've only run V7 Beta in Vista x64, and obviously only the 32-bit Beta, getting a score of 4613, which at least sounds impressive compared to V6 scores, although it may not be once in the context of other V7 scores.

[Michael (Passmark)]

There is no 64-bit beta. The baseline search in PT7 does include PT7 baselines, however there is no easy way to find them amongst the PT6 ones at the moment. We plan on making changes to the baseline search to make it easier to separate v6 and v7 results before we get to the final release of PT7. For the time being if you're really interested you can search on filename. With PT7 we have started back at number 1 for the baseline file names, So if you search on "BL1." (the '.' is important) you should get only two results, one PT6 and one PT7, from there you should be able to tell which one is the PT7 baseline as it will be the more recently uploaded one. Note that this only applies to the search inside PT7, the web based baseline search has not been updated yet.

[Windsurf48]

Thanks. I noticed that the V7 baseline I uploaded had fairly low number (I think it is in the 500s and assumed that the numbers had been reset). I'll check the baselines when I'm back in my Vista partition.

[Richard Wolf]

Hi,

in Beta 4 I cannot save the results as a .csv file anymore. When I select this it creates a .csv.txt file. (PerfRes.csv.txt) and it is a simple textfile.

[Michael (Passmark)]

csv files are text files in effect. With PT7 we no longer output proper csv (comma separated value) files and instead use a semi-colon as the separator. We changed the extension to .txt because when excel is installed it tries to open csv files but imports them incorrectly due to them not being proper csv anymore. We will remove the restriction on pt forcing the file to be .txt for the final release of PT so that you can still csave them as such if you want.

[{SAS}TB]

wow, nice score increase with this, gone from 6300 odd to:

[Michael (Passmark)]

I downloaded and had a look at your baseline. Your score seems to have been greatly inflated by your Disk speeds (~2000 MB/s). Are you sure you are testing the same drive you used to? These speeds are far to fast for any standard drive, they look like they're from some sort of ram disk.

[{SAS}TB]

Yeas its a RAM drive, but the same drive i use for version 6.1; sample benchie:



Im not the only one seeing huge increases, a fellow (other) forum member is getting in excess of 22k.

After a cursory glance I suspect the difference between his results and mine are the GFX card. Following some monitoring of my rig during the bench ive noticed that my GFX card (4870x2) stays at 2D speeds for all but the final (full screen) 3D GFX test and only then hits 3D clocks with the second core not even hitting 50% utilisation.

Im on 9.1 Cat drivers, full rig spec below:

Enermax Infinity 720W PSU
Asus P5Q Deluxe Motherboard
Q6600 @ 3.825GHz
4x1024 OCZ ReaperHCP @ 1133
ATI HD4870x2 clocked to 800/100
2xWD Raptor 74GB in RAID0 - but benching from RAM disc
Dell Ultrasharp 2407 TFT
XP 64Bit

Hope this helps

[{SAS}TB]

Also, if it helps, comparisons between versions 6 and 7 (exactly the settings on the same PC - both 64Bit versions)

[{SAS}TB]

Well after some image mounting RAM disc antics I have a few more scores for you.

Basically if you mount a CD / DVD image on your RAM drive it hugely increases your score for that element. There is also a huge increase in the new 7 beta version over the old 6.1 using the same PC and settings, these two tests were run back to back straight after each other:

The 6.1 Version:



And the version 7 Beta:




Also after a bit of playing around with the individual tests and settings I tried the individual GFX tests (on the v7 beta) for the "Bouncing Balls" and first "Aeroplane" test, which are usually windowed compared to forcing them (as individual tests) as full screen.

The issue appeared that they were not actually forcing the GFX card into 3D mode (I have heard the same issue with nVidia card users as well).

By forcing the tests to run in Full screen it does then actually ramp up to 3D speeds as well as appearing to make full use of the two cores:

Bouncing balls test - windowed 1580 fps versus 5750 fps Full screen
1st Aeroplane test - windowed 499 fps versus 1164 fps Full Screen

As you can see, quite a difference and, I should imagine, would also make a huge difference to the overall test score if it utilised full screen, true 3D testing?

Is this something that could be addressed ?

[Michael (Passmark)]

Having a look at your scores it seems the increase in overall score between 6 and 7 is normal. The scoring system underwent a number of changes and although we tried to keep the overall score roughly equivalent for an average system your ram disk is anything but average.

One of the changes we made was to give the disk mark more weighting. The reason for this is that over the years disk drives haven't improved nearly as much as, for instance, CPU's. This meant that on an average system the disk was accounting for a far smaller portion of the final score than it was several years ago, as such it was given extra weighting in the new version so that the disk score had a reasonable effect on the overall rating.

In your case, the ram-disk is so fast the it is making up 80-90% of your final score and the change in weighting significantly changed the result.

In the end the overall rating is merely a quick, although not always accurate, way to get an idea of how systems compare. For fairly normal systems it is a somewhat effective comparison but as you can see when you start adding things like ram-drives it becomes less useful. For more useful comparisons you need to look at the individual test suites, or in some cases individual tests, and compare them based on what you want the system to do.

We are also aware that people can mount ram-drives for the CD test to get some absurd overall scores. Ultimately there's not much we can do to stop people doing this (and there are cases where people might actually want to test this). We may consider changing this before final release so that individual components scoring highly will not drastically affect overall scores.

As for the 3d issues you mentioned, we will have a look into this. Although the performance of windowed 3d applications is probably still a relevant test to perform even if the video card can go faster in full screen. Also we are aware that multi-GPU cards (and multi-card setups) do not really utilise their second core, we did look into fixing this but didn't have any luck. As such the 3d tests are only really suited at comparing single card, single-gpu setups.