Not all CPUs are created equal

I have recently come into some socket 939 3800+ 2.4GHz ADA3800DAA4BW (two with steppings of LBBWE and and four with NBBWE) CPUs, and today I set out to test them.

Testing methods:
Stage one: pop in the processor, spread some paste on it, lock down the heatsink, then once she's up and running let it sit at idle for 10 mins to get all of the startup processes out of the way so as to minimize unexpected variances between tests, then start up HWMonitor and let it sit at idle for another ten minutes to ascertain the average idle temps.

Stage two: run SuperPi 1m and record the time to completion.

Stage three: run AquaMark3 to get CPU and GFX ratings (perhaps not the best program for rating a CPU, but it's what I've got), and also max load temps.

Now, the first three chips' (two LBBWE's and an NBBWE) test results came out as I'd expected all scoring 40.9-41s on the SuperPi 1m calculation, scoring 8921-8947 in the AquaMark CPU rating, idling around 34-35°C, and having load temps of 47-50°C.

The fourth and fifth chips surprised me though, now they scored the same 40.9s on SuperPi 1m, and the same 8890-8900+ in the AquaMark CPU rating, but boy did they run cool, 30-31°C average idle temps, and load temps of 42-43°C. Oddly though they did actually have slightly lower ratings in AM, but they were 8897 and 8938 respectively as compared to 8899-8947 on the other chips, but I think that difference is low enough to be within the margin of error considering my small sample size.

So, then I test the last chip, and its scores are about the same as the first three with 35°C idle, 46°C load, 40.9s on SuperPi 1m, and 8899 on AquaMark.

Then, thinking perhaps I had put on too much thermal paste in the earlier tests I decided to retest the hottest-running chip in the bunch (35.5°C idle and 50°C load), and did get slightly better results on the retest with 34.5°C Idle and 49°C Load. That kind of difference is barely noticeable under normal conditions though, and if you told someone to try and find what was different without giving them that hint they probably wouldn't be able to spot it. I was looking for a difference of 3-4° minimum to confirm my theory that I was doing a poor job of applying the thermal paste initially, which I clearly didn't find.

So it seems that now and then some chips run cooler than others, even when in the same case, under the same heatsink/fan, and using the same thermal paste, when applied by the same person.

All tests were run with a stock AMD S939 HSF for single core processors and Arctic Cooling MX-2 thermal paste so that I wouldn't have to deal with a curing time.

Yep, it is quite noticeable. Generally (at least in my limited experience) later-released chips can run a bit cooler. For example, I do believe each revision of the i7 920 series processors got cooler (less leakage) than the previous revision. I believe my D-revision runs the coolest of the bunch.

Let's have something clear: do the chips with the same revision score identical or not?

Because if the chips with the same revision score different - that's what i would call "not all created equal", but only if the chips with different revisions score different stats - that is expected in some ways...

Mounts can result in difference in temps, if you did it more than once and it is consistently better then that really runs cooler.

Kuroimaho said:

Mounts can result in difference in temps, if you did it more than once and it is consistently better then that really runs cooler.

Click to expand...

Yes, on one of the cooler-running chips I've now mounted it two or three times on two different boards and it idles around 28-32°C on both (on one system around 25-29°C, and the other around 28-32°C).

Artas1984 said:

Let's have something clear: do the chips with the same revision score identical or not?

Because if the chips with the same revision score different - that's what i would call "not all created equal", but only if the chips with different revisions score different stats - that is expected in some ways...

Click to expand...

In at least one case, no.

I did have one chip that scored over a hundred points higher than any other chip, actually it was almost two hundred points higher than some of them.

The lowest scoring chip was at 8,897, and the highest scoring one was 9,104.

There is more than just the score though, I find temperature to be an important factor as well, because it could indicate a greater chance for overclock-ability and greater stability at higher speeds.

Those scores are well within the margin of error of AQ3. AQ3 while CPU bound is a GPU test. Why not try Wprime 1024M? Or Geekbench, or...something made for CPU's to test a CPU.