PerformanceTest V8 Feature List

[David (PassMark)]

We are hard at work on development of PerformanceTest 8 and have progressed far enough that we are able to publish a feature list and coding progress of the early Aug 2012.

• Add a new memory latency test. [Done]
• Re-write standard RAM tests in assembler to be less CPU dependent and more RAM dependent. [Done]
• Add blu-ray burning speed test & DVD burn on Win7/8 as a advanced test [Done]
• Allow multiple automatic test runs & take max result [Done]
• Improved advanced disk test [Done]
• Allow network test to run for much longer [Done]
• Improve system information. Include RAM stick model and timing info, CPU temp & more [Done]
• DirectCompute test as part of standard GPU/3D tests [Done]
• DirectCompute as an advanced test [Done]
• Re-balancing of instruction mix in the Integer maths test and prime number test in the CPU Suite test [Done]
• Single thread CPU test [Done]
• Faster startup on launch despite additional system information collection [Done]
• Better looking graphs / new charting engine in advanced tests [Done]
• Support for testing more than 64 CPUs / cores [Done]
• Better baseline search, both for basic and advanced search [Done]
• Combining of baseline management windows, adding bookmarks, etc.. [Done]
• User interface updates with 3D models [Done]
• Updated 2D tests with Direct2D API. [Done]
• New baseline file format [Done]
• Replace Tokamak physics engine with Bullet or Havok engine [Dropped]
• Better video card model # detection, especially dual cards and Intel HD GPU [Done]
• DirectX 11 3D test & dropping of one of the DirectX 9 tests. [Done]
• Better SLI & Crossfire support [Done, but not perfect, but it the best we can do without support from nVidia & AMD]
• IP6 support in network test [Done]
• Official Windows 8 support [Done]
• Official Windows Server 2012 beta support [Done]
• A new restructured & optimised database on the PassMark web site to support graphing of PT8 data [Done]
• Combining 32bit and 64bit releases into a single install package [Done]
• Better compression on install package to offset extra size of 3D textures, etc. [Done]
• Faster install after download by removal of some install dependences [Done]
• Polishing [Working on that now]
• And various other minor changes....

We expect to have some type of beta release available mid Aug 2012.

Update: Beta is available now. Any comments on the beta should be made in the new beta release thread.

- David

[godfather]

Looks great.
I'm especially looking forward to the crossfire and SLI support, as well as the single-threaded CPU test.

But I'm wondering how the single-threaded CPU test would work. Will it count as simply one of the tests in the overall CPU score? Or will it be published on its own? Thanks

[David (PassMark)]

Updated coding progress today in the first post.
We have been working on the 2D tests this week. Been spending some time looking at what DirectWrite and Direct2D can do.

Big problem with all this new fancy 2D stuff is that there is no support for it in Windows XP. But we don't want to drop support for XP just yet. So we'll need to work out a solution for that.

The single threaded CPU test will just be one of the existing tests and make up part of the V8 CPUMark score. So it will pull down (slightly) the results of CPUs that have poor single threading performance.

[godfather]

So it will pull down (slightly) the results of CPUs that have poor single threading performance.

Which will improve the usefulness of the scores. As it stands now, AMD's FX series is way overrated.

[David (PassMark)]

Should also allow us to do a "single threaded" graph.

[wonderwrench]

How about HD video encoding/decoding including actual video playback. On CPU, GPU etc. Scores based on speed, accuracy, dropped frames, CPU/GPU load etc.

I know I mentioned the above before but today's CPU's can handle all normal tasks at any price point in a matter of seconds, not counting gaming and video encoding. Video encoding can be a slow process on any hardware. What hardware is the fastest? Best cost performance ratio? As far as HD video decoding goes not all hardware is created equal. Sure for the most part any newer system should be able to decode HD video. The trouble is there are many decode features and modes that are not supported on some common hardware available today. Sandybridge can't output 24 fps. Lower end GPU's can't handle edge enhancement, noise reduction etc. It would sure be nice to see a database comparing HD video encoding/decoding. I doubt I'm the only one that wants to see it.

Bill

[Vovan]

Should also allow us to do a "single threaded" graph.

So there will be new, separated single-thread benchmark list?

The single threaded CPU test will just be one of the existing tests and make up part of the V8 CPUMark score. So it will pull down (slightly) the results of CPUs that have poor single threading performance.

Also it's good idea run not only single and max-thread CPU benchmarks, but also 2-4 threads, and show results on click as extended info. For instance I personally need the fast CPU for two threads, so composite 8-core & single-core performance wouldn't help so much.

[David (PassMark)]

We are planning on looking at 1 thread and max threads. Max threads on some systems is now 64+. It doesn't make sense to look a every value between 1 and max. As pretty much every CPU now has 2 or more cores the single threaded performance will give a very good indication to the 2 thread performance.

[Vovan]

Well, I can agree.

Anyway, will single thread result be available on the WEB-site either as individual list or extended info? (first is better)

[David (PassMark)]

Should also allow us to do a "single threaded" graph on the web site.

[tehcheese]

• Better baseline search [Done on client, 30% done in cloud]

could you further elaborate on this? the reason i ask is because along with the "Sort by" function in the "Advanced Search" feature, it would be good to have an option added to only show values of a specific range(greater than/less than). It would be convenient for someone like me who use G3D scores when upgrading my graphics card to shoot for a good/reasonable G3d mark to better plan what my new setup should be like and also get to look at the different setups i find under this range. Yes i can use the select page function but its a bit tedious to pin down a range i want, not to mention the majority of "The Connection to the Internet timed out" errors that i get when doing it this way.

[David (PassMark)]

Thanks for the suggestion.

The problem with including the ability to do a ranged search on G3D is that it is the thin edge of the wedge. It wouldn't make sense from an interface point of view to do it just for this one value. What about the CPUMark, 3Dmark, Diskmark, etc..
On the other hand doing it for every value would result in a much more complex data entry field. The form would be too large to fit on some monitors.

Given there is a work around of doing a sort and then just ignoring the unwanted results, I don't think the added complexity justifies the benefit.

More interesting (for us) is the 'time out' message. This should only happen if the search takes longer than 15 seconds. I did some searches here today but we couldn't get any search to take this long. More typical was search times around 4 sec. What was the search criteria you are using to generate this long searches? If there really are searches taking longer than 15 sec then we need to fix it. It could of course be some other issue, like a bad Wireless connection to the internet.

[tehcheese]

It wouldn't make sense from an interface point of view to do it just for this one value. What about the CPUMark, 3Dmark, Diskmark, etc..

Yes i know it would make sense to do it for the other scores, i was just using G3D as an example.

On the other hand doing it for every value would result in a much more complex data entry field. The form would be too large to fit on some monitors.

Well the way i was thinking it could be implememnted would be just one checkbox that reads, "Show values under this range", when checked this checkbox could enable a dropdown box with the options "lesser than" and "greater than", then beside this dropdown box would be a textbox for the desired range input. Also note that this checkbox could be disabled until the user chooses a sort option that sorts by some kind of numerical value(score) because this range search obviously wouldn't work for the other stuff like "CPU Type" and that sort.

More interesting (for us) is the 'time out' message. This should only happen if the search takes longer than 15 seconds. I did some searches here today but we couldn't get any search to take this long. More typical was search times around 4 sec. What was the search criteria you are using to generate this long searches? If there really are searches taking longer than 15 sec then we need to fix it. It could of course be some other issue, like a bad Wireless connection to the internet.

I'm on an ethernet connection. I usally get this error when I switch to the advanced tab, then under "Sort by", choose some score then hit search. The initial search is usually flawless but when i attempt to do a huge page change; like selecting the 1400th page on the page select box, that's when i get the error, smaller changes from one page to the page right after that usually works but that is somewhat slow and sometimes i still get the error. might just be my connection idk.

Lastly thanks for the reply, i understand it isn't very beneficial vs. the complexity. I just thought it would be good to have. I really like all of your software, every single software made by Passmark is made to perfection. Also looking forward to the beta release of Performance Test

[David (PassMark)]

Was able to reproduce the timeout here. So it isn't your connection.
Workaround for the moment is to reduce the scope of the search. e.g.
Select a date range, or a CPU type, etc..

The problem is that not all fields are indexed in the database. So some queries which depend on non indexed field results in linear search and then a time consuming sort. Indexed fields are pre-sorted so don't have this problem. But making all fields indexed would significantly increase the size of the DB and slow down other operations. Anyway we'll take a deeper look next week and work out the best solution.

[David (PassMark)]

Another development update on PT8.

We are just finishing testing an internal alpha release at the moment.
We should have a beta release available next week, Mid Aug 2012.

I have also update the post at the top of the thread with some new details.

Update: the beta of PT8 is now available to the public.
I will close this thread. Any comments on the PT8 beta should be made in the new thread above.