PCMark 7 To Arrive on Thursday

Hey, at least we're not the only ones postponing releases!

Originally planned for May 3 release, Futuremark delayed the long awaited system benchmark to this Thursday, May 12, 2011. According to Oliver "Ollie" Baltuch, president of Futuremark Corporation, the company decided to delay the launch due to last-minute fixes and tweaks in order to ensure stability of performance results.

Given that the last PCMark worked on Windows Vista alone (runs on Windows 7 too), don't be too surprised to hear that the only OS that can run PCMark 7 is - Microsoft Windows 7, of course. Futuremark decided for such move in order to ensure complete system optimization across the board - regardless of what hardware you run, Windows 7 should yield apples-to-apples results.

...


40

Belgium leeghoofd says:

let's hope a little less storage dependant output... the one with the most Acards, RAID controller and SSD's wins.........

Finland FM_Jarnis says:

Mirrors are already getting the file online (they were authorized to do so at 11AM GMT, 45min ago) and official pages should go live in about 25 minutes. So... happy benchmarking :)

Greece George_oc says:

First test run on my laptop :D Looks good so far, but it takes like 20 minutes to finish because each test is ran 3 times and then the final score test derrives from the average of these three... I downloaded it from techpowerup http://www.techpowerup.com/downloads/1998/mirrors.php ;) http://3dmark.com/pcm7/1975

United Kingdom El Gappo says:

Running 3 times will make it real easy to spot bugged runs. No more "oh my video encoding is bugged wot is dis power toy O.o"

Belgium HybridChiller says:

i'm running it right now to.

Done

Score 3288

subscores :


Greece Tiltevros says:

again only 4-6 cores are testing.... why we the pcmark7 doesnt use all cores in case of a 16 or 24 thread pc??

Belgium HybridChiller says:

you would expect that it uses more then 4-6 cores because it's the newest version.

Even 3Dmark Vantage uses all cores for cpu tests.

i'm surprised about it that it didn't uses all cores

+1 3Dmark Vantage
-1 Pcmark7

Greece Tiltevros says:

pcmark vantage used uptoo 24 cpus in gaming....

Belgium Massman says:

Interesting. Futuremark made the same call in the 3DMark11 benchmark: scaling goes up from 1 to ~ 6 cores but after that it doesn't scale that much anymore. I guess hardware or game developpers told them they won't be using more than 6 cores in the next couple of years.

United States Hondacity says:

thats good.... :D

Greece Tiltevros says:

Massman said: Interesting.

Futuremark made the same call in the 3DMark11 benchmark: scaling goes up from 1 to ~ 6 cores but after that it doesn't scale that much anymore. I guess hardware or game developpers told them they won't be using more than 6 cores in the next couple of years.




its a benchmark not a game mate...
this is really sad... pcmark7 is optimised for 2x00K whitch performes 25-35% better than the 980x-990x at full speed....

btw everyday usage is not encoding and decoding videos....

Belgium Massman says:

I know it's a benchmark, but this benchmark is designed to reflect real-world performance as much as possible. That's why FM usually sits together with hardware and game developpers to understand what the future will bring and how to set up their benchmarks so it would would not be outdated when a new generation of hardware arrives. Obviously after 3DMark Vantage, which had too much focus on the CPU part, FM changed their benchmark so it would correctly reflect the core and thread scaling. For daily usage, you don't need more than 6 cores. So it makes sense for benchmark developpers to take this into account and don't make the benchmarks scale over 6 cores.

Greece Tiltevros says:

most of the encoding and decoding video and especialy image manipulation needs cores... Video transcoding Video transcoding workloads transcode videos by changing the resolution, framerate and bitrate of the video. Video transcoding uses Fast Transcode feature found in Windows 7 Media Foundation. Video downscaling workload (found in Entertainment and PCMark tests) source video is H.264 (1080p, 30fps, 12Mbps). The destination is H.264 (320x240, 30fps, 500kbps). High quality video transcoding workload (found in Creativity test) source video is the same as with downscaling workload. The destination is H.264 (1080p, 24fps, 10Mbps). Image manipulation The workload uses the Windows Imaging Component (WIC). Source images are in HD Photo format. This workload simulates everyday image manipulations such as color correction, stretch, flip and rotate. now for gaming... R.U.S.E is using upto 32 threads ;)

Belgium Massman says:

Hmm. Well, it's just what I assume Futuremark did. Of course I don't know the real reason why the scaling is so bad. You make a valid point :)

Belgium HybridChiller says:

Pcmark7 is a little disapointment.

Newest benchmark and not even uses all cores..............

video encoding, i never do that pppfff.

Hope this doesn't get to hwbot

Canada Vinster says:

I find this a a disappointment. A benchmark should test a platform to it's maximum ability. if you have 12 cores, it should run 12 cores.. not 6.

If I wanted a test for daily usage I'd rely on the Windows Experience test.

Vin

Australia Dinos22 says:

surprised they didnt at least scale it to 8 cores

so who's done some tests and who will tell FM to stop being jerkoffs and include offline tests for basic license or reduce that $40 fee a bit sheesh

Belgium HybridChiller says:

Vinster said: I find this a a disappointment. A benchmark should test a platform to it's maximum ability. if you have 12 cores, it should run 12 cores.. not 6.

If I wanted a test for daily usage I'd rely on the Windows Experience test.

Vin


Agree

Greece Tiltevros says:

the bench is not even 64bit

Belgium HybridChiller says:

that's even more worse

Canada Vinster says:

Chiller said: that's even more worse


I agree...

So they are charging $40 for a weak bench and that is only 32bit... what are they doing? Benchers and clockers would be buying this, not the everyday joe... plus 3DMV is only $6 for a basic licence and imo would be better to test the CPU than this.

Vin

Denmark riska says:

I get 5823 points with my 980x @ 4912
and a small gtx550 ti 1gb gfx card and my new vertex 3 ssd

http://3dmark.com/pcm7/34682

dont now if it is a good score..

United States BenchZowner says:

It is a very good score Henning ;)

Denmark riska says:

thanks my freind

Greece Tiltevros says:

FM is not responding if there it will be any patch fix for 64bit and compatibility with multisocket pc's(using all cores)

Belgium Massman says:

Response on Facebook.

Finland FM_Jarnis says:

Tiltevros said: the bench is not even 64bit


This is absolutely not true.

It installs both 32 and 64bit binaries on 64bit systems. Look under x64 folder - all the benchmark workload bits are there in 64bit form and are used on 64bit systems.

The GUI ("PCMark7.exe") is managed code written using C# and it runs 32bit on 32bit systems and 64bit on 64bit systems. Just look at your task manager - there is no *32 behind PCMark7.exe now is there?

So PCMark7 is 64bit. Stop trolling.

Greece Tiltevros says:

it runs only in 32bit binaries..... open ur task manager before u run the application and u will see...

Finland FM_Jarnis says:

On the subject of "multi-socket compatibility". It is obviously compatible with multi-socket. Runs just fine. As for scaling under different CPU environments, you have to understand that PCMark recreates existing tasks you can do using Windows. A lot of the code that creates the load is operating system code. So your complaint can be roughly translated to "Microsoft Windows 7 OS routines do not scale to multi-socket, more-than-6-cores in common usage scenarios". Video Transcoding is bog standard Media Foundation Fast Transcoding H.264 job. It is true that Intel offers very good hardware acceleration for that on Sandy Bridge (and future) CPUs and this affects results. It could also be GPU accelerated if GPU vendors support the feature. The benchmark does not know what it runs under - it just does what a modern video transcoding app would do on Windows 7. More on Media Foundation here; http://en.wikipedia.org/wiki/Media_Foundation and Fast Transcoding here; http://msdn.microsoft.com/en-us/library/dd317903%28v=vs.85%29.aspx I've requested a more detailed technical description about the workloads and will provide it to you soon. We'll probably do also a revision of the whitepaper to provide more technical detail.

Belgium Massman says:

Thanks for the update Jarnis! I think the hardcore PCMark benchers would absolutely love that whitepaper :D

Finland FM_Jarnis says:

Tiltevros said: it runs only in 32bit binaries..... open ur task manager before u run the application and u will see...


Not true.

I just ran it on my Windows 7 64bit.

PCMark7.exe (no *32)
MFFastXcode.exe (no *32)
MFPlayback.exe (no *32)
etc..

The only 32bit processes it fires up are related to SystemInfo system scanning.

Finland FM_Jarnis says:

Massman said: Thanks for the update Jarnis! I think the hardcore PCMark benchers would absolutely love that whitepaper :D


PCMark 7 whitepaper is available here and contains already a lot of detailed information;

http://www.pcmark.com/wp-content/uploads/2011/05/PCMark7_Whitepaper.pdf

But again, we're looking into revising and extending this with more detail. Feel free to suggest what you think is missing.

Belgium HybridChiller says:

Tiltevros said: it runs only in 32bit binaries..... open ur task manager before u run the application and u will see...


indeed

Finland FM_Jarnis says:

Massman said: Interesting.

Futuremark made the same call in the 3DMark11 benchmark: scaling goes up from 1 to ~ 6 cores but after that it doesn't scale that much anymore. I guess hardware or game developpers told them they won't be using more than 6 cores in the next couple of years.




This is mostly a demonstration that, yes, most of these benchmark scenarios are heavily GPU limited. And it wouldn't be much of a 3D graphics benchmark any other way.

Does it mean that investing into massively multi-core multi-socket system for gaming is... sub-optimal..? Well, uh, yes. The vast majority of games right now are squarely GPU limited and tossing them more CPU power does roughly nothing. Benchmarks reflect this.

The main reason that 3DMark 11 overall score scales at all are the CPU-dependant parts. If you look just at the framerates from individual game tests, tossing more CPU cores does not help.

Finland FM_Jarnis says:

Chiller said: indeed


Could you provide some screenies of this? It runs 64bit binaries on my system.

This is starting to sound like some kind of an issue with some configurations...?

Belgium Massman says:

FM_Jarnis said: This is mostly a demonstration that, yes, most of these benchmark scenarios are heavily GPU limited. And it wouldn't be much of a 3D graphics benchmark any other way.

Does it mean that investing into massively multi-core multi-socket system for gaming is... sub-optimal..? Well, uh, yes. The vast majority of games right now are squarely GPU limited and tossing them more CPU power does roughly nothing. Benchmarks reflect this.

The main reason that 3DMark 11 overall score scales at all are the CPU-dependant parts. If you look just at the framerates from individual game tests, tossing more CPU cores does not help.


In Vantage, the game tests didn't scale with cores/threads either, but the CPU tests had a very big impact on the end result (which is what most people use to compare).

How did you do the evaluation of the weight of the CPU test? In Vantage, the CPU test kept on scaling with more cores and threads, whereas in 11 it seems that the scaling stops around 6 active threads. Was this purely based on game developper feedback (eg: not going to use more cores in games)?

I must say 3DMark11 is quite a bit different from Vantage benching wise. And not in the bad way ... :D

United States Movieman says:

FM_Jarnis said: This is mostly a demonstration that, yes, most of these benchmark scenarios are heavily GPU limited. And it wouldn't be much of a 3D graphics benchmark any other way.

Does it mean that investing into massively multi-core multi-socket system for gaming is... sub-optimal..? Well, uh, yes. The vast majority of games right now are squarely GPU limited and tossing them more CPU power does roughly nothing. Benchmarks reflect this.

The main reason that 3DMark 11 overall score scales at all are the CPU-dependant parts. If you look just at the framerates from individual game tests, tossing more CPU cores does not help.


BUT,BUT, some of us like the dualies! LOL
The EVGA SR2 board is the best thing since sliced bread IMHO.
Just kidding, best of luck with this.

Greece Tiltevros says:

FM_Jarnis said: This is mostly a demonstration that, yes, most of these benchmark scenarios are heavily GPU limited. And it wouldn't be much of a 3D graphics benchmark any other way.

Does it mean that investing into massively multi-core multi-socket system for gaming is... sub-optimal..? Well, uh, yes. The vast majority of games right now are squarely GPU limited and tossing them more CPU power does roughly nothing. Benchmarks reflect this.

The main reason that 3DMark 11 overall score scales at all are the CPU-dependant parts. If you look just at the framerates from individual game tests, tossing more CPU cores does not help.


those CPU cores helps to break 2600k in half with dual xeons
here is a screen that u asked

dual xeon E5520 scoring ~11000 points in transcoding

Belgium HybridChiller says:

let it see you're subscores, whant to compare with mine X5550 @ 4.1ghz and RevodriveX2 960gb, when i runned Pcmark7

and i didn't even had 11.00 transcoding but 5600

Please log in or register to comment.

Leave a Reply: (BBCODE allowed: [B], [QUOTE], [I], [URL], [IMG],...)