Quad Core Thread: Overclocking, Cooling, Motherboards, Troubleshooting & General Info

[mroz]

And in which everyday scenario are you going to be encoding video at such high resolutions? gaming and benchmarking is one thing but using 1080p as an excuse for overclocking ? no chance of convincing there lol.

I'm not against it but using such a thing as an excuse like that is totally off because it's not an everyday scenario that 90% of the people here are going to be doing (if at all).

Lots of people (well, some ) archive tv programs & hd will play a bigger part in this as time moves on - there are already quite a few legit sources.

Video editing, processing & encoding is the main reason I've just bought a Q6600. You mention xvid; aside from transcoding for portable devices with low grunt or standalones, why still use xvid?

I doubt 90% of PC users do a lot of taxing video processing, but I imagine the proportion of users with high end machines who do is significant. I'm less sure about typical OC'ers, as I get the impression they're mostly gamers.

[mrk]

XVID was used as an example as it's the most single common and high quality compressed format for video people download and playback on pcs or upscaling divx dvd players etc.- since nobody in the UK can record or archive any HD source (currently SKY etc) without it being an illegal activity I don't see how this plays part here - if someone has a HD camcorder then they encode to a compressed hd format on the device itself ready for playback or transcode to DVD etc.

I just did a DIVX transcode of a LOST ep using the HD1080 profile and managed 24fps using virtualdub via divx6/AC3. I think this is perfectly fine especially for the output you're getting.

[Agent]

And in which everyday scenario are you going to be encoding video at such high resolutions? gaming and benchmarking is one thing but using 1080p as an excuse for overclocking ? no chance of convincing there lol.

Hold your horses
You use your PC for gaming and benchmarking, don't mean everyone else does

I do loads of 3D rendering, where the *only* thing that matters for output is CPU speed (providing the disks can keep up).

Last week I did a little benchmarking of my own. By overclocking, I shaved off around 1.5 seconds on a render. These renders were taking about 5 seconds each.
Now, these were due to be output at 60 frames per second (physics simulation), so that's 90 seconds saving, for each second of realtime footage. Take into consideration that the simulation was about 4 minutes long....
90s saving = 1 second realtime footage
5400s saving = 1 minute realtime footage
21600s saving = 4 minutes realtime footage, or 6 hours of rendering time to be saved.

That brings it down to 14 hours total rendering time, from 20 hours. For me, thats a big difference

I'm not against it but using such a thing as an excuse like that is totally off because it's not an everyday scenario that 90% of the people here are going to be doing (if at all).

"90%" of people on here? This is an enthusiast forum though, it stands a good chance that people want the extra power for a reason?
Thats a very bold statement

[Agent]

.....since nobody in the UK can record or archive any HD source (currently SKY etc) without it being an illegal activity I don't see how this plays part here .....

Ermmm.... what about people who render content from programs such as Maya, 3D Max, Blender ect.....

They can be higher than standard HD resolutions (720/1080) and are certainly not illegal.

[mroz]

XVID was used as an example as it's the most single common and high quality compressed format for video people download and playback on pcs or upscaling divx dvd players etc.

If you're encoding for pc playback & that hardware isn't ancient it'd be silly to encode with xvid, fine mpeg4 sp encoder that it is; h.264 is much nicer.

- since nobody in the UK can record or archive any HD source (currently SKY etc) without it being an illegal activity I don't see how this plays part here

I wouldn't know tbh as I don't have access to a source but I thought there was a bbc test transmission? Is the legality of recording hd different to any other broadcast?

- if someone has a HD camcorder then they encode to a compressed hd format on the device itself ready for playback or transcode to DVD etc.

I just did a DIVX transcode of a LOST ep using the HD1080 profile and managed 24fps using virtualdub via divx6/AC3. I think this is perfectly fine especially for the output you're getting.

It's probably a corollary of Parkinson's law or something similar - I tend to allow my activities to expand in complexity to make full use of the computing power available to me. I doubt I'll ever run out of ideas to explore which require more power than I have.

[Fraz]

I just find it a bit bizarre that people want to overclock these things like crazy when these chips rip apart everything to begin with! my philosophy on the matter is to have the quietest system possible but powerful and the stock cooler seems to do the trick and if in a years time Quad Core can do with a bit of an OC as the next gen gfx cards really push for it then this possibility is there! but only if it's needed of course.

I do a lot of grid computing, and overclocking my Q6600 CPU by 25% means 25% more work units processed in a day. And for the price of a huge CPU cooler (£35), my system is running quieter and cooler than it would with the stock intel one. Frankly, for a investment of £35, it seems stupid NOT to overclock your CPU. I've effectively turned my Q6600 valued at £170 into a Q6850 valued at £680, all for the cost of a £35 uber CPU cooler. That's the best £35 I've ever spent frankly.

[D-BEAR]

Apart from gaining higher benchmark scores does anyone actually benefit from overclocking even a Q6600? I got mine this week and it EATS everything - Mediacoder transcodes a full video in under 15 minutes flat and the average encode rate is 130fps when encoding to XViD @ 800kbps. Games are similar - I'm pushing 200fps average in Source alone on an 8800GTS.

Bioshock runs 50-70fps in DX10 mode (I don't play in dx9 mode )

I'm just wondering, it runs cool anyway at around 43 degrees on the stock cooler with the fan set to 50% power and rises to 45-50~ per core under full load. I also get 10,300 in 3dmark06 with a small OC on the gfx card.

I just find it a bit bizarre that people want to overclock these things like crazy when these chips rip apart everything to begin with! my philosophy on the matter is to have the quietest system possible but powerful and the stock cooler seems to do the trick and if in a years time Quad Core can do with a bit of an OC as the next gen gfx cards really push for it then this possibility is there! but only if it's needed of course.

A P.C. is not just for gaming and encoding.....As i put my idle pc to good use running WCG to benefit the world...with an extra 900mhz with a little increase in vcore, and a better cooler....it not only runs as cool as yours....but i get a lot more work done with my big free speed increase!....a bit silly not to really.

[mroz]

Okay, Deep Twit is finally up & running - stable at stock for about 24 hours, so now I'm venturing into overclocking. Vid's 1.2875. At stock, core temps were around 24 idle, 42 under load, with a 5 degree variation between cores.

Testing 8x400.

Tried Vcore of 1.2875 in bios, giving 1.248 idle dropping to 1.232 under load. Failed P95 within 10 minutes.

Upped it two notches to 1.3 (1.264 idle, 1.248 under load) & so far stable after three hours. Will leave it until the morning & if it passes, drop Vcore one notch & re test, before trying something higher, though tbh I can see me ending up running it at 8x400.

Core temps presently 47,44,44,47 all oscillating by about 1 degree up/down, so between 43 & 48 atm.

Ambient's been 23 to 24 all night.

A few questions:

My hsf is lapped. How many people would ever consider lapping their cpu ihs? My 4 to 5 degree variation between cores is bugging me.

What proportion of people found they could reduce this by twiddling the hsf mounting screws? I had to run it with the fan removed to do this - interesting to see the temps then gradually rise over half an hour to roughly stabilise around 60 under load. Anyhow, I couldn't close the gap. Tried both before & after adding foam under the hsf mounting bracket (which did stop the hsf twisting, thankfully).

Does anyone know how the four cores are physically laid out? Specifically, which core is where?

Coretemp: Is it known for reporting the wrong cpu speed? It's insisting mine's running at 450x8, when it's actually 400x8. This is with XP Pro SP2, 32 bit.

8x400 or 9x356? Any reason to go one way or the other, other than a fsb of 400 suiting PC6400?

Any info appreciated

[Harkin]

Cpu-z version 1.40.5 would be better to use for checking Cpu speed etc. Coretemp is good for temps

About adjusting the mounting screws, I've read Clunks prvious posts and he has had success doing it. It's better if you have P95 running while you are adjusting it, but it can be hard depending on what cooler you use.

(http://www.arcticsilver.com/pdf/appi..._quad_wcap.pdf) is where you can see where the cores are, taken from the 1st pag on Clunks quad thread

[mroz]

Cpu-z version 1.40.5 would be better to use for checking Cpu speed etc. Coretemp is good for temps

Cpu-z is one of the other reasons I thought Coretemp was wrong, rather than the more usual suspect, me.

About adjusting the mounting screws, I've read Clunks prvious posts and he has had success doing it. It's better if you have P95 running while you are adjusting it, but it can be hard depending on what cooler you use.

That is what I did, but like I said, I had to run it with the fan off to get at two of the screws. Interesting, but I couldn't reduce the temp gap. I imagine it only helps if the hsf is able to rock slightly. My hsf is flat so I'm guessing the ihs on the cpu is slightly concave, which would stop this technique working.

(http://www.arcticsilver.com/pdf/appi..._quad_wcap.pdf) is where you can see where the cores are, taken from the 1st pag on Clunks quad thread

Thanks, but I've got that. I want to know where the cores are located inside those chips & which core is which. Just curious

[Harkin]

Does sound like your ihs is mucking things up a bit. Lapping works but usually doesn't improve things that much( few degrees at most), and plus it voids your Warranty so its up to you. It would flatten the ihs and make contact better though, personally I wouldn't do it, but thats just cos Im a scaredy cat

[mroz]

Me too, but on the plus side I'm stupid

[fat jez]

Lapping works but usually doesn't improve things that much( few degrees at most),

Lapping my IHS got me about an 8°C improvement. The exact figures are earlier in the thread. I also have a discrepancy between core temps and no amount of twiddling the nuts on my heatsink mounting has evened them out.

Cheers,
Stephen

[mroz]

Lapping my IHS got me about an 8°C improvement. The exact figures are earlier in the thread. I also have a discrepancy between core temps and no amount of twiddling the nuts on my heatsink mounting has evened them out.

Cheers for the feedback. Did the IHS lapping kill the core temp differences?

[mroz]

Sorry for being lazy - found your post.

So, you saved up to 12C but temps were /very/ high before lapping, compared to other reports. Also, still a 4C variation after lapping between cores. Maybe in my case it's not worth it then. Thanks again.

[mroz]

Anyone care to comment on the 8x400 vs 9x356, if only to say it was a stupid question?

Deep Twit seems to be stable at 8x400 with Vcore of 1.3; now testing 1.29375

Another question. Does XP or Nvidia apply some sort of gfx card underclocking when the display blanks? I notice, from the graphs in Speedfan, when I come back to check on the machine & unblank the display, temps had been cycling in a stable pattern but then immediately begin to rise a degree or two before assuming the same pattern at this slightly elevated level.

I figure it must be gfx related so either the gpu has a nap or it's just down to the work the gpu is doing, however surely just displaying a mostly static XP desktop doesn't stress the card enough to put up the case temp 1 to 2 degrees?