Making a Quad go 100%

DJPadre wrote on 1/11/2008, 2:08 AM
Ok, well after an enitre 20hours of recofngirugin, backing up and finaly removing my old systems Mobo, CPU and Ram, ive succssfully installed my new Quad without the need to reinstall or fekk with anythign.
With the help of one member its been one of those gambles that if i didnt take, id still be creaming at the latency and lag times for renders. Im glad i can at least render 4x faster than realtime..
BUT...

After running tests, Ive noticed that the highest the CPU has gone is about 45% with all four cores churning out.
When changing the affinity to 2, its only a VERY marginal difference in speed. (a 26 secnd clip using tonnes of colour and noise reduciton took 10 secono render on 4 cores, and 12 seconds on 2)

Now, considering the throughput of the drives (im using an ide for read and a seperate ide for write (yes its a trippy board) i woul dhave thought that the throughput itself wouldnt be an issue.
Im not intersted in Raid at this time as ive had some bad BAD experiences with it but is there ANY way i can crank the CPU to go higher than this?

I mean its goot that i cna render and do other shit while it renders, but if i cna crank the CPU to its fullest potential, I'd like to....

thoughts? Suggestions?

I havent overclocked this yet.. .

Comments

megabit wrote on 1/11/2008, 2:21 AM
DJ, when and why a QUAD is used 100% by Vegas and when / why not has been a question I've kept asking myself for a long time. With Vegas 7, if I remember correctly it was always 100% (or near to it), so changing affinity to 2 cores would make render times 2x longer.

But since the 8.0, I ONLY am getting 100% CPU usage with 8bit video (and not always either). With 32bit, it seldom goes higher than 70% - and yet it can be as fast if not faster than with 8bit (see my post about the rendertest-hdv.veg done in 2 mins with 8bit, and 1m:22secs with 32bit setting).

I know this only adds to the confusion, but I really don't know the answers. What's even more strange, the very same projects/settings are always rendering at 100% under Vista x64...

Go and figure :)

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

JJKizak wrote on 1/11/2008, 6:16 AM
With my AMD 4600 X2 on a standard 4 x 3 DV project the processor shows 90-100% during render and HDV 16 x 9 about 55% during render which leads me to believe that it is showing more "speed" than "CPU Usage". I think there is some kind of defective parameter monitoring software faults built into the task manager since the actual CPU usage stays the same but the frame speeds follow how fast each frame is being processed.
JJK
DJPadre wrote on 1/11/2008, 6:56 AM
interesting...

i ran the render test and lo and behold, the CPU cranked all 4 cores to 100 percent and it didnt skip a beat.

Flatlined at 100%

Did a render using Magic Bullet, and knowing that MB uses 1 core + GPU i was expecting it to only show the one core in use.
Instead i noticed the first and 3rd core were in use.
So i believe that either Vegas consideres each primary core as an individual entity as opppsoed to working in parallel with each other, else we would have seen only the priamry core lus seconday parallel unit in action whiel teh other 2 remain dead.

Im thinking maybe Vegas isnt as fast as the CPU???
I mean i was only running 4:3 footage with a PNG letterbox overlay using levles and a sharpness filter.
I am unsure if the sharpness filter utilises multiple cores though, but i ran a similar test with sharpness removed and it pretty muhc ran in the same fashion.

Now, i dont know if its a HDD bottleneck, but i seriously doubt it. Im rendering from 25mbps to 25mbps DV AVI so in theory the CPU shoudl be cranking up to 100% with the onlly lag (if any) coming from the HDD which would only be sporadic in any case considering these IDE's have 16mb buffers..

Lets hope Sony are working on an update to manage this issue.
In any case, its remained stable through all my torture tests today, and considering this is a build off an existing windows install, time will tell.



goshep wrote on 1/11/2008, 7:36 AM
As you continue testing, remember the length of your test renders will also effect any difference in performance you notice. All things being equal (which they never are) a 20% increase in performance is only going to be 2 seconds of a 10 second render. As you work with longer and more CPU intensive renders you should notice more significant differences in renders.

I had the same concerns when I first started using my quad. I really noticed the difference with effect heavy, longer renders. I still occasionally notice the CPUs running at 40-50% of max but that is the exception rather than the norm. There are effects/processes within Vegas that are not optimized for multi-threading and that may have some effect. I'm no software engineer nor am I a Vegas Power User. I just know that my quad slices and dices my old single-core 3.6!
rmack350 wrote on 1/11/2008, 9:06 AM
Intel quads are two dual cores on a single assembly so maybe it's not so wierd that magic bullet uses processor 1&3. These would be the first core of each physical CPU. Seems to me like even on processes that can't use dual cores the audio still goes to the second core.

Seems like SCS keeps tweeking the multicore performance of Vegas with every version, so maybe there are things in VP8 that throttle Vegas down on 4 cores.

Rob Mack
DJPadre wrote on 1/11/2008, 3:51 PM
i havent installed 8 on this machine yet due to stability. Im testing on my Dual core laptop first and seing how it fares Once im 100% satisfied only then will i add v8 on this machine. Ive had issues in the past with uninstalls so i dont want to fall into the same same issue.

once i do get it n this system ill compare and see the difference