@HouseofMac

I've concluded the exact same thing reading through all the info that's come out since the SB release.

i7-2600K chip for the more advanced HD 3000 graphics, on an H67 board with no discrete GPU... I don't do any gaming so the integrated graphics should be enough for me.

This was an interesting article - Lucid are developing software to enable desktops to power down the GPU and thus enable the use of QuickSync on the IGP, for systems that have GPU's. (note - still on H67 boards only, not P67)

http://techreport.com/discussions.x/20217

Once that's available, you'd be able to install a GPU if you felt you needed more horsepower.

Time to go shopping.

@Philip,
Thanks for that will keep an eye out for Lucid. Looks like a way ahead if the Intel graphics are not what is reported.

No thanks to intel though for making what looks like a great CPU then locking it all up and have one of their subs (Lucid) make money out of unlocking it.

PS I'd go shopping today except the "K" version is not out in Australia yet -

Hi guys - same question as HouseofMac.... where do I spend my $ for best performance boost? Do mainly H.264 HD work, not a lot of special effects. Its seems from my surfing that - please correct me or confirm that my current video card is the first area to hit, but beyond that... help!
1- upgrade to a couple of Xeons & dual MB?
2- Add an OCZ Revodrive X2?
3- A Tesla card won't help much?
Got $5k to play with
Thanks for any help i7-940, 12Gig DDR3-1333, Asus P6TDelux, SAS 15k HDD, Radeon HD4850, Vista64, PD9 ultra64

I've got my i7 2600K bits on order. I decided to go with 16G ram, one SSD (program storage), and one HDD (data storage). I got plenty money, however, I suspect the bottlenecks are transcoding (hence the 2600K) and program access (hence the SSD). As I understand it, > 12G DRAM and you really don't need a swap file, would reduce wear on the SSD, and because even SSDs are *much* slower than DRAM, speed things up - even more so with HDD.

*edit* Apparently this whole page file/swap file SSD thing is quite contraversial. According to Microsoft http://blogs.msdn.com/b/e7/archive/2009/05/05/support-and-q-a-for-solid-state-drives-and.aspx you should put the page file on the SSD, but there is some discussion as to the amount of page file relative to the amount of RAM. So, I'll follow Microsoft's advice but maybe set a minimum page file of a couple gigabytes and see if it grows ...

I didn't see anything suggesting PD9 benefits from Tesla, but I don't know for sure.

There are really two issues with editing: editing, which is mostly 'meat' limited on any decent system, and transcoding which is hugely CPU intensive but can be done without human intervention. Its better to wait 60 minutes for a movie to transcode than 16 hours, and 30 minutes is better than 60, however, even 60 minutes is not so bad.

I wonder - do we have some sort of standard test file we could use to test our system performance? I'd be keen to find out what the difference are!

Intel HD Graphics 3000 is directly integrated in the processor core of the Sandy-Bridge processors. Alike its predecessor it does not have a dedicated graphics memory, but, can access the Level-3 cache of the CPU (which is now called 'Last Level Cache' - LLC) of up to 8 MB, depending on the CPU. Further memory requirements are covered by accessing the RAM of the computer.

Production on a LAPTOP? That is a re-occurring nightmare I have, and Sandy-bridge isn't going to make it go away. It only allows an additional 8MB of memory to be used by the GPU. NOT ENOUGH, especially for editing/rendering HD.

Quote: There are really two issues with editing: editing, which is mostly 'meat' limited on any decent system, and transcoding which is hugely CPU intensive but can be done without human intervention. Its better to wait 60 minutes for a movie to transcode than 16 hours, and 30 minutes is better than 60, however, even 60 minutes is not so bad.

Watching HD...no problem! But, if you intend on rendering HD content you will need a strong GPU with at LEAST 512MEG dedicated memory! Levels below that will be an experience at best. Just ask all the folks here who are having problems on graphic cards with low memory.

DISCLAIMER: OK, I realize that finding the posts that show which problems relate to low GPU dedicated memory may be difficult, in light of the fact that PD9 is having trouble with many graphic cards right now. Cyberlink ran ahead into the 64bit world, and not even the GPU manufacturers had their drivers ready!!! Nvidia is testing a new release to address some of these issues (I am sure ATI and others are trying to get them fixed, too), as new programs come online that use the 64bit technology.

FredB

It will be interesting to see if the memory limit issue is an important factor with Sandy Bridge. On transcoding benchmarks, the CPU seems to do very well. The computing model appears to be somewhat different from a straight up use a GPU and Cuda (etc.) so memory may not be as much of an issue - after all, GPUs are designed for drawing polygons, not transcoding.

When my system gets put together, I'll run some 1080p/60 video and time the rendering and post the input and output files and timing online. Perhaps others with different configs could run the same video and we could see what the differences are.

I am always interested in new tech...I just wanted to add that "cautionary note" for the visitors at this forum. They are not all willing to slide on the bleeding-edge (as you are) and optimized performance will surely follow the release of the new Intel Architecture, not precede it.

I look forward to your review!!

Understood - but my *old* desktop is about 8 years old (ish) and my main computer is a Netbook, so 'bleeding edge I ain't.

However, money I got, and about 10 hours of 1080p/60 video I also got, patience not so much!

Now if only Newegg would ship my stuff ...

Well then, welcome to the Sandy Bridge "edge". I will be glad to test with you when you get the new machine.

I have Intel(R) Core i7 CPU -920 @ 2.67GHz (8 CPUs)
Memory: 9216MB RAM
NVIDIA GeForce GTX 465
running on Windows7 64BIT.

Quote: Old card the CPU usage was about 30% when GPU rendering with GPU at 90%
New card CPU usage is 80% when GPU rendering with GPU at 40%

I'm not sure but I'll offer the following. I've been playing around with Nvidia GeForce 8800GTS 512, a GTX470, and a GTX580 video cards with various hard disks with and without multiple disks in a RAID(0) array and it appears that the load put on the CPU deals a fair amount with I/O capability. As you get a great GPU doing the encoding, I/O demands increase putting a load on the CPU and disk subsystems. Somewhat surprises me as the video I/O throughput in reality is still small compared to a single disk I/O performance. Since I currently don't have multiple CPU's to throw in my test box I was simulating CPU capability by adjusting affinity. I see trends very similar to what you describe and can control the split to some degree based on how good of CPU and I/O subsystem I have for a given capability GPU. The GTX580 with a slow 2.5", 5400rpm laptop drive mounted in my desktop essentially unloads the GTX580 GPU and loads up the CPU's.

Jeff

There was a significant change in the load handled by both my old card( GTX 260) and my new card (GTX 465) when using current drivers. This changed even more (greater % use of GPU) when I installed the latest BETA Nvidia drivers (version 266.35). GPU now runs around 90% of the load and CPU is dropping off to about 15%

Thanks for the comments Fred, I see similar very high CPU usage across all '8' cores, RAM usage is usually <30% of my physical, hence I was wondering what doubling to twin Xeons would do, and moving off SAS to the revo2 card to ensure I can feed both CPUs fast enough.
As you comment I would expect the GPU to then become the bottleneck, but I have failed to get any sort if handle on how a Tesla 1060 compares with the newer cards like your GTX580, or Quadros or any of the dozens of options out there... has anyone got any experience? thanks i7-940, 12Gig DDR3-1333, Asus P6TDelux, SAS 15k HDD, Radeon HD4850, Vista64, PD9 ultra64

Quote: As you comment I would expect the GPU to then become the bottleneck, but I have failed to get any sort if handle on how a Tesla 1060 compares with the newer cards like your GTX580, or Quadros or any of the dozens of options out there... has anyone got any experience? thanks

I don't see the Tesla 1060 listed as being supported by PD9, so until it is, it may not be a good purchase.

CL does list Tesla in the knowldege base for PD and imply CUDA support in PD7, hopefully they didn't go backwards in PD9!
http://www.cyberlink.com/prog/support/cs/product-faq-content.do?id=6503&prodId=4&prodVerId=-1

How do I know if my NVIDIA card is compatible with the CUDA version of CyberLink PowerDirector?
CyberLink PowerDirector requires NVIDIA CUDA technology and will run on NVIDIA 8 series or higher GPUs.

The CUDA version of CyberLink PowerDirector 7 currently supports the following NVIDIA VGA series:
NVIDIA GeForce and GeForce Mobile
NVIDIA Quadro and Quardro Mobile
NVIDIA Tesla

Jeff

JL_JL

I'm new to this, but from the PD9 System Requirements (presumably with respect to GPU accelleration)

Supported GPGPU graphics card
NVIDIA:
•GeForce 8500GT and above
•GeForce 9800GT and above
•GeForce GT/GTS/GTX 200 series
•GeForce GT/GTS/GTX 400 series
•GeForce GT/GTS/GTX 500 series
AMD:
•Radeon HD2600 and above
•Radeon HD4600 series
•Radeon HD5600 series
•Radeon HD6800 series


I'm just saying ...

Quote: This changed even more (greater % use of GPU) when I installed the latest BETA Nvidia drivers (version 266.35).

I didn't redo all my matrix of testing but enough to see no changed in rendering time with any of my 3 cards under test with the released 266.58 compared to 260.99 drivers. As usual, somewhat hard to generically bench mark as all have varied experience.

Jeff

Mingy ... I am eagerly waiting for your report on your new PC and the Sandybridge CPU! Dell Studio 1558 (laptop) * I5-520 * AMD HD 5470 * 4GB RAM - 1066MHz DDR3 * 320GB 7200RPM Hard Drive * Windows 7 Pro 64 bit * BluRay Burner

Oh my goodness.........this is too much for me....I just want to be able to produce my videos.

What program do you use to monitor GPU load? Apparently GPU-Z does not support Nvidia, but perhaps RivaTuner is OK?

Hi,
Please stay focused, this is a PowerDirector forum and for matters relating to that software.

Thread closed.

Dafydd