c2d - different speed on different cores?

[TehQuick]

I've had e6600 for a couple of weeks now. I've been playing around with oc'ing and got another e6600 for comparison. Now, I've never had any troubles with the first one, but the second one, while booting off at lower voltage (1.2 or so vs 1.325 for the first one), behaves kinda strange. When overclocking, it does not error in orthos, it either reboots the computer or crashes orthos in windows...

OK, getting to the point now. I'm running the second e6600 at stock speed and noticed after about 3 of hours, that the first core lags behind the second one in orthos, eg, second core is at test 6, round 10, while first one is at test 3, round 10... At that, first core is running hotter than the second one (constant delta in TAT of 2-3c between the two, and I did try reseating HSF a few times).

So I wonder, why is that? I mean, is first core unclocking slightly or something due to higher running temp? I'm using stock cooler, btw. Also, is it possible that I got a bad CPU (2nd)? I mean, it does not error in Orthos no matter what I do, but rather reboots/crashes, and also this weird difference in computational speed b/n cores (can't recall if the first e6600 did that)

 

[TheRapture]

2c-3c diff in temps between cores is normal.

 

[TehQuick]

2c-3c diff in temps between cores is normal.

Yeah, i know that. My question was more about speed differential. It's kinda weird because I did disable all kinds of cpu temp enhancements and such in the bios, but it seems, cpu is still doing something... At this point (4 hours) the difference b/n cores is almost five tests (i.e. core 2 is doing test 9, while core 1 is still doing test 4 of the same 14th round)

 

[luihed]

I dont think its something you should worry about, mine does the same.... just remember to use the fast core when benching single threaded benchmarks.....

 

[TehQuick]

Interesting development. I've installed watercooling and found out that with old HSF, the thermal compound didn't cover all of the IHS on the cpu. It was making good contact but apparently not good enough. So with the new watercooling setup, the load temps are identical for both cores (idle still shows a couple of degrees celcius difference) and their speed is also identical - 8 hours into orthos at 3.6ghz, and both cores are neck in neck. Apparently, if one of the cores is overheating, it's clocking down somewhat... thus I had the speed discrepancy with the old mount...

 

[boshuter]

All dual core cpus are going to have one core thats faster than the other to some extent. Thats why you always want to find out which one is fastest and use that core to bench stuff like SuperPi, on some cpu's the difference can be significant, on others it's negligible.

 

[blade52x]

Mine both are almost at the same point after 10 hours...

If I start using Firefox, listening to music, or just using the computer in general, one of my cores starts to lag behind the other.

 

[TehQuick]

All dual core cpus are going to have one core thats faster than the other to some extent. Thats why you always want to find out which one is fastest and use that core to bench stuff like SuperPi, on some cpu's the difference can be significant, on others it's negligible.

Well, I'm fairly confident that this could be attributed to how well your hsf is making contact with the cpu... the cores that don't get quite enough cooling are getting slowed down This is what happened to me anyways In both cases I wasn't using the computer at all during the orthos runs

 

[ziddey]

are you using the pc at all while it's load testing? it could be detracting clocks from one of the cores, causing what you see. or whatever windows has in the background.

 

[TehQuick]

are you using the pc at all while it's load testing? it could be detracting clocks from one of the cores, causing what you see. or whatever windows has in the background.

I repeat myself: no, i did not use the computer, was just checking it once in a while by moving mouse to wake up the display.

 

[boshuter]

Well, I'm fairly confident that this could be attributed to how well your hsf is making contact with the cpu... the cores that don't get quite enough cooling are getting slowed down This is what happened to me anyways In both cases I wasn't using the computer at all during the orthos runs

LOL.. well I'm pretty confident that it wasn't lack of contact causing it. I benched my 950es Presler last week in 1M and 32M pi and setting affinity for the faster core gained me significant time; I was benching 1M at 5.9ghz on dry ice at -63c, 32M was at 5.4ghz, I really don't think cooling was a factor in one core being faster than another. I've had similar results on every dual core I've benched, My E6700es was the same, not setting afinity on it was keeping me at 11.1** sec in 1M, setting affinity got me in the 10's at 10.984 at 4.7ghz.

 

[m0deth]

I've also noticed that XP distributes load quite differently than say 64bit vista or linux 64....that may be a factor

 

[repoman0]

haha, what are you guys talking about? Both cores have the same data processing capability as long as they're stable because they run at the same speed and have the exact same architectural design and the same ability to access the rest of the system. Any speed difference between cores is up to your OS's process scheduling

 

[LuminaryJanitor]

Any speed difference between cores is up to your OS's process scheduling

Glad to see that someone in here knows what they're talking about .

EDIT: Actually, I should clarify a little... I doubt it's due to the scheduler splitting the workload unevenly between cores; rather, it's because the scheduler itself needs to run somewhere.

 

[Rydawg5143]

hell realtemp shows my cores idle as 31 and 25