Sign in with
Sign up | Sign in
Your question

Various tests for q6600 overclocks. Critiques and opinions wanted.

Last response: in Overclocking
Share
May 22, 2011 2:22:15 PM

Regrettably, I've overclocked my system rather late (Purchased Dec 2007) and I had no idea of their tremendous benefits until I did it. That said, I have some questions that I would like to ask. First, the system in question:

CPU: Q6600 VID 1.3250 G0
Mobo: Asus P5Q Deluxe
CORSAIR CMPSU-450VX 450W ATX12V V2.2 80 PLUS Certified Active PFC
ZALMAN CNPS9500 AT 2 Ball CPU Cooling Fan/Heatsink
COOLER MASTER Centurion 5 Case.
Exhaust fans: SILVERSTONE FM123 120mm in the back, two SilverStone 92 mm Case Fans on the side.
Intake Fans: Stock fan by HDD, plus the old 120mm fan is in the front.

Ram: G.SKILL 4GB (2 x 2GB) 240-Pin DDR2 SDRAM DDR2 1066 (PC2 8500)
I also recently purchased Kingston HyperX 4GB (2 x 2GB) 240-Pin DDR2 SDRAM DDR2 1066 on sale. I wanted to see if it would be work well with my system. This question inevitably brought me to investigating how ram worked vis-a-vis the FSB. Of course, this took me right to overclocking my computer.

The two overclocks that worked best for me were: (edit: not much anymore, was able to bring case temps down a bit, some of the 3.1s below are also fine)
1)
FSB: 333
Multiplier: 9
CPU Volts: 1.26250
Ram 1:1, 4-4-4-12

(I did try a 4:5 ratio with the ram at ~416. It worked fine, but Asyncronous mode did add heat and vcores. If I want to do that, ill just up the FSB)

2)
FSB: 400
Multiplier 8
CPU Volts: 1.325
Ram 1:1 4-4-4-12

The only thing i set was the PCI e slot at 100. FSB strap was auto. I just set the ram to 1:1. All other voltages were left on auto. I did try to adjust the other voltages mentioned here, but i constantly ran into instability issues. I would love to drop the temps of the 3.2 overclock with some simple voltage tweaks, but I doubt its going to have a big effect. I don't know.

Temeratures were measured with Realtemp, coretemp (both with a TJMax of 100) and Everest (with a TJMax of 90). I used the higher temps, as there is quite a bit of disagreement on which TJ Max temp is correct. Temperatures were achieved with Prime 95 blend test. The 3.0 setup gave me temps that hovers around 56 with 59 peak degrees core temp. I used this to video encode a blu-ray last night and it worked great there too. Typical temperatures on a particularly hot portion of the test would be 58c, 57c, 51c, 55c. More normal prime 95 tests would be 56c, 55c, 49c, 54c. Idle temps with this are in the low 30's.

For the 3.2 setup, the highest core temperatures would move from 62-65 C with peaks going to 67c. Idle temps are in the high 30's.

I did try to find a "happy median" between the two, by experimenting with 350 351 355, and 360 FSB with a x9 multiplier. I tested these results on everest and found that each of these had relatively poor memory read, write, copy, and latency ratings. (Edit: I just whipped up the 3.1 test below, and its ok, too. Temps are 60, 58, 52, 56, with a 63 peak on prime 95)

3.2ghz: 400 FSB, 8 Multi 1:1 Ram @ 4-4-4-12
Memory Read 7000 MB/S
Memory Write 7288 MB/S
Memory Copy 7658 MB/S
Memory Latency 75.7 ns
CPU Queen 27102
CPU PhotoWorxx 18400
CPU ZLib 84501 KB/s
CPU AES 23823
FPU Julia 11861
FPU Mandel 5849
FPU SinJulia 2989

3.2ghz: 355 FSB, 9 Multi, 1:1 Ram @ 5-5-5-15
Memory Read 6257 MB/S
Memory Write 6473 MB/S
Memory Copy 6756 MB/S
Memory Latency 85.4 ns
CPU Queen 27056
CPU PhotoWorxx 15626
CPU ZLib 84318 KB/s
CPU AES 23769
FPU Julia 11833
FPU Mandel 5838
FPU SinJulia 2981

3.158ghz: 351 FSB, 9 Multi 1:1 Ram @ 4-4-4-12
Memory Read 5076 MB/S
Memory Write 3810 MB/S
Memory Copy 4891 MB/S
Memory Latency 92.9 ns
CPU Queen 26788
CPU PhotoWorxx 16593
CPU ZLib 83207 KB/s
CPU AES 23438
FPU Julia 11661
FPU Mandel 5761
FPU SinJulia 2942

3.1ghz: 345 FSB, 9 Multi 1:1 Ram @ 4-4-4-12 (345 Mhz)
Memory Read 6260 MB/S
Memory Write 6307 MB/S
Memory Copy 6656 MB/S
Memory Latency 83.6 ns
CPU Queen 26343
CPU PhotoWorxx 15895
CPU ZLib 82275 KB/s
CPU AES 23177
FPU Julia 11531
FPU Mandel 5678
FPU SinJulia 2907

3.0ghz: 333 FSB, 9 Multi 4:5 Ram @ 4-4-4-12 (~417Mhz)
Memory Read 6719 MB/S
Memory Write 6095 MB/S
Memory Copy 6634 MB/S
Memory Latency 76.6 ns
CPU Queen 25503
CPU PhotoWorxx 18551
CPU ZLib 79390 KB/s
CPU AES 21570
FPU Julia 11083
FPU Mandel 5480
FPU SinJulia 2765

3.0ghz: 333 FSB, 9 Multi 1:1 Ram @ 4-4-4-12 (333 Mhz)
Memory Read 6205 MB/S
Memory Write 6083 MB/S
Memory Copy 6475 MB/S
Memory Latency 83.7 ns
CPU Queen 25534
CPU PhotoWorxx 15878
CPU ZLib 79382 KB/s
CPU AES 22367
FPU Julia 11122
FPU Mandel 5495
FPU SinJulia 2806

I guess what i want to know is, am I overlooking something? Is there a MHZ combo that I should test before I settle on the 3.0 667 as a summer setup and the 3.2 800 as the winter? Something else? Is there something I am overlooking with the speeds voltages, not all uniform ram?

Added:
Just tooling around and a new answer came to me typing this all out

3.1ghz: 388 FSB, 8 Multi 1:1 Ram @ 4-4-4-12 (388 Mhz) (1.29 cpu volts)
62,60,55,58 with a 65C peak on prime 95
Memory Read 6933 MB/S
Memory Write 7066 MB/S
Memory Copy 7437 MB/S
Memory Latency 75.8 ns
CPU Queen 26324
CPU PhotoWorxx 17439
CPU ZLib 82044 KB/s
CPU AES 23172
FPU Julia 11506
FPU Mandel 5655
FPU SinJulia 2902
a c 197 K Overclocking
May 23, 2011 12:18:40 PM

OK. First of all, you load temps are fine. My Q6600, OC'd to 3.6 GHz at a vcore of 1.425 volts runs at 61 - 66 C under Prime95 load.

Second, look at your benchmarks for 3.2 GHz. Notice how the performance does not scale with the increase in FSB freq. 400 MHz is 12% higher than 355 MHz but it yields only a trivial increase in benchmark numbers.

I run my Core2 CPU's at max internal multiplier. The increase in stability is worth more than the very small increase in processor speed.
m
0
l
May 23, 2011 4:10:44 PM

jsc said:
OK. First of all, you load temps are fine. My Q6600, OC'd to 3.6 GHz at a vcore of 1.425 volts runs at 61 - 66 C under Prime95 load.

Second, look at your benchmarks for 3.2 GHz. Notice how the performance does not scale with the increase in FSB freq. 400 MHz is 12% higher than 355 MHz but it yields only a trivial increase in benchmark numbers.

I run my Core2 CPU's at max internal multiplier. The increase in stability is worth more than the very small increase in processor speed.


First of all, a big thanks for taking a look at this. My wife certainly isn't interested. :) 

Second, I just cleaned out the dust from the front filters and, lo and behold, temps dropped 5c or so. The 3.1 (344x9) setup never got above 60c.

Thirdly, on your point about benchmark numbers. I want to be sure that the decrease in ram speed (and photoworxx benchmark) is not a big concern here. Thats what scales differently with a smaller FSB and larger multi. The heat also disproportionately scales a little more with FSB over a focus on multiplier. I've read that ram speed here isn't that big a concern. Its probably true in my case, as Ill just game some. The bulk of my hard computer work is video encoding, and thats a processor show.

Heres a graph



That 2.4 266 12:8 (400 mhz ram) blue setting is just stock. I needed a baseline.
m
0
l
Related resources
a c 197 K Overclocking
May 23, 2011 10:29:22 PM

That illustrates my point. Notice on all the benchmarks except memory i/o. There's virtually no difference at 3.2 GHz. Memory i/o shows a small improvement at 400 MHz, but the difference does not show up in the application benchmarks.
m
0
l
May 23, 2011 11:05:09 PM

jsc said:
That illustrates my point. Notice on all the benchmarks except memory i/o. There's virtually no difference at 3.2 GHz. Memory i/o shows a small improvement at 400 MHz, but the difference does not show up in the application benchmarks.


Just to be clear, these differences in Ram and Photoworxx (which stresses the integer arithmetic and multiplication execution units of the CPU and also the memory subsystem.) are to be considered negligible between the 355 X 9 and the 400 X 8?




CPU Queen thrown in

More explanation of the benchmarks
m
0
l
May 31, 2011 2:11:12 PM

Well, I'm going with the 388 X8 setup for the summer. It peaks at 60c, runs at 1.29 volts, and theres a nice balance in the various benchmarks.

For the winter, I'll go 3.2 or faster.
m
0
l
!