Importance of processor cache size
-
mspears
Importance of processor cache size
I'm getting ready to update the processor in my computer and was wondering just how important cache size should be in my decision. Would I see any difference, all other things being equal, between a 1 mb and 2 mb cache size on the processor? I'm asking because I've found a killer deal on a P4 560j that would be hard to pass up. 3.6 ghz and HT, but it's not 64bit capable and has 1 mb of cache. Older 90nm technology, but it's still fast at 3.6 ghz. Or would it be smarter to go with a 64bit capable processor looking to the future. Will MSP be modified to 64 bit in the future?
Last edited by mspears on Wed May 24, 2006 3:00 pm, edited 1 time in total.
I'd prefer double-core to HT. I don't think the cache size will make much difference for video (until it becomes a Gb or so!). Don't have a crystal ball re Ulead's future policy.
[b][i][color=red]Devil[/color][/i][/b]
[size=84]P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.[/size]
[size=84]P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.[/size]
Yeah, I would go for the 560 if you can get a great deal. Processor speed is still the biggest factor in video editing, although this will change eventually as new software starts to be written to take advantage of dual processors etc. The intel chip also has inbuilt instruction sets like SSE that speed up encoding. Just make sure you have decent cooling! I have a 650 and it was running at about 65 celsius during encoding. I got some extra fans. Now it stays around 50/55 which in my mind is still hot, but better.
I don't think 64 bit MSP is in the near future, seeing as none of the other companies have released anything, or even talked about releasing anything. (Except final cut?)
I don't think 64 bit MSP is in the near future, seeing as none of the other companies have released anything, or even talked about releasing anything. (Except final cut?)
What are you talking about? MSP does take advantage of HT, dualies and dual-core!troppo wrote:Processor speed is still the biggest factor in video editing, although this will change eventually as new software starts to be written to take advantage of dual processors etc.
[b][i][color=red]Devil[/color][/i][/b]
[size=84]P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.[/size]
[size=84]P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.[/size]
Sorry Devil, but on ALL the reviews I read the dual cores only have slight advantage and sometimes no advantage over their single core cousins during encoding. As for HT, the processor mspears was talking about already has it anyway.What are you talking about? MSP does take advantage of HT, dualies and dual-core!
The only software that specifically takes advantage of dual cores during encoding is Sony's vegas as it is written to specifically assign different parts of the encoding to a different processor, even on another computer.
Every piece of software 'takes advantage' of the fact there is a second processor, but very few are actually written to harness the full potential, MSP included.
Sorry, but must disagree.
MSP7 and 8 were written specifically to take advantage of them, when available.
Mark Dileo organised a test when MSP7 came out. This was done with a standard (very difficult) sequence of effects and the results are published at
http://www.hyperactivemusic.com/MSProFi ... esults.htm
The top few in the table (best overall performance) just happen all to be dualies, not as fast, either, as the singles/HT in the table. You will also note that there is a difference (usually by about 2 places) between HT on or off, in the cases where both were tested on the same computer.
IMHO, there is no doubt that dualies (and presumably dual core) do perform much better with MSP than singles of the same speed. I agree - and have said so for years - that HYPErthreading is well named. My own experience with HT on 2 computers with very different Intel chipsets is that on rendering I get ~3-5% improvement with it on and, with the other, 5-15%. Not stupendous, but not insignificant, either. OTOH, the wait time between clicking Instant Preview and the preview starting is cut by 30% with one and 45% with the other: this IS significant. Both CPUs are about the same speed (3.06 and 3.0 GHz) but the FSB is 533 and 800 respectively. It is possibly more important to have dual channel RAM to get the max effect out of HT and this is chipset-dependent.
So, please, correct your thinking in this matter.
MSP7 and 8 were written specifically to take advantage of them, when available.
Mark Dileo organised a test when MSP7 came out. This was done with a standard (very difficult) sequence of effects and the results are published at
http://www.hyperactivemusic.com/MSProFi ... esults.htm
The top few in the table (best overall performance) just happen all to be dualies, not as fast, either, as the singles/HT in the table. You will also note that there is a difference (usually by about 2 places) between HT on or off, in the cases where both were tested on the same computer.
IMHO, there is no doubt that dualies (and presumably dual core) do perform much better with MSP than singles of the same speed. I agree - and have said so for years - that HYPErthreading is well named. My own experience with HT on 2 computers with very different Intel chipsets is that on rendering I get ~3-5% improvement with it on and, with the other, 5-15%. Not stupendous, but not insignificant, either. OTOH, the wait time between clicking Instant Preview and the preview starting is cut by 30% with one and 45% with the other: this IS significant. Both CPUs are about the same speed (3.06 and 3.0 GHz) but the FSB is 533 and 800 respectively. It is possibly more important to have dual channel RAM to get the max effect out of HT and this is chipset-dependent.
So, please, correct your thinking in this matter.
[b][i][color=red]Devil[/color][/i][/b]
[size=84]P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.[/size]
[size=84]P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.[/size]
-
mspears
After clicking on the link and taking a look at the testing done with MSP7, it appears that the test utilized all single core processors, but the top performers were overclocked cpu's with DUAL channel RAM. One of the side effects of overclocking a cpu is a proportionally raised FSB (front side bus) and an increase in RAM mhz. This looks to be a key factor in MSP performance while encoding in addition to Devil's observations that HT and dual core processors outperform a single core during the editing process especially during instant preview which is heavily used (by me) along with a smaller, but not insignificant increase in encoding times. With this new knowledge, I've decided to buy a Pentium D 930, 3.0 ghz Dual Core which will be used on my 945 chipset motherboard that will allow me to mildly overclock it to 3.2 ghz thereby raising the FSB past 800 mhz and my dual channel RAM past 533 mhz. I bet it'll rock! Thanks guys! BTW, it also doesn't appear from the test that processor cache size made any significant contribution to performance.
Is overclocking wise for video work? Rendering not only heats the CPU, but also the N'bridge and RAM, like there is no tomorrow. The problem is that you are using the resources in a constant mode, unlike for e.g., gaming, so there is a greater risk of overheating. I know I can, on one of my 'puters, cause the CPU to reach critical levels with an Intel heatsink and fan, canalised from outside the case, with a long render. Even now, just typing, it's at 54°C (ambient temp in the room 31.8°C (very hot here, today - see http://www.cypenv.org/weather/), so a 22°C differential isn't bad. The way I feel is that I would rather allow a 2 h render take an extra 10 minutes, than to ##### it up after 1 h 50 min because the system glitched. Especially as the CPU goes into slow mode if it does overheat, so you possibly would gain negatively.
Just my feeling.
Just my feeling.
[b][i][color=red]Devil[/color][/i][/b]
[size=84]P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.[/size]
[size=84]P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.[/size]
-
mspears
Devil, you are correct. Unless you really know what you're doing, using an overclocked computer is not wise for video editing. You really have to stay on top of temperatures, provide sufficient cooling and be absolutley 100% POSITIVE the computer is stable. Not impossible, but this is probably beyond the capabilities of the normal video editor person. I stated in my previous post that I might 'mildly' overclock the Pentium D 3.0 ghz (that I just ordered), but after more thought, how much would I really gain? Probably not much, and then I have to worry about instability issues causing glitches in a computer that I use to make money. I decided against overclocking. It's not worth it. But thanks for helping in my decision about buying a processor. All your information was valuable.
