OC settings advice from the experts

G

Guest

Guest
Archived from groups: alt.comp.hardware.overclocking (More info?)

I'm new here. So first let me say hello and thanks in advance to all who
spare a moment to provide their thoughts.

I have an ABIT IS7 motherboard with a P4 3GHz/1MB over clocked and stable
at 267Mhz. So my 3GHz CPU is running at 4.005GHz. I'm using the stock
heat sink and fan on the CPU. I'm running Windows XP Home Edition. I have
the following settings and temperatures at idle and at load while running
Prime95 v2.38 at maximum torture settings:

Bus setting: 267/76/38
Memory: 5:4 (512x2 DDR PC3200) someday I'll get PC4000 and try 1:1
AGP Setting: 7:2:1
VCore: 1.6v (Prime95 produces errors at voltages below 1.56v)
DDRcc: 2.70v
+3.3: 3.3v

Temperature at idle:

Sys: 34c | CPU: 54.5 +/- 5c | PWM: 50.0 +/- 5c

Temperature at load (Prime95 maximum torture setting):

Sys: 37c | CPU: 75.5 +/- 5c | PWM: 75.0 +/- 5c

The system is rock solid at these settings and Prime95 ran error free for
two hours. No crashes. No I/O problems with Sondra2005 disk drive test.

Are these settings and temperatures safe? Am I walking on thin ice, or
living on borrowed time? Any suggestion, comments, observations? Any
suggestions on a better heatsink fan combo that will fit on this MB
inside by Antec SLK1600 case (300watt stock power supply was upgraded to
450watts, hey what can I say, the case was free).

Thanks, -Barry
 
G

Guest

Guest
Archived from groups: alt.comp.hardware.overclocking (More info?)

"baj2k" wrote in message...
> I have an ABIT IS7 motherboard with a P4 3GHz/1MB over clocked
> and stable at 267Mhz.

That's a good overclock!

> Bus setting: 267/76/38

As a general point, you'd be better off locking the PCI and AGP buses at
33/66. Running an overclocked PCI bus doesn't gain you any significant
performance nowadays, and can cost you in component (primarily hard disk)
reliability.

> Memory: 5:4 (512x2 DDR PC3200) someday I'll get PC4000 and
> try 1:1

You're actually losing quite a bit of overall system performance by running
the async memory bus. As you say, PC4000 (or maybe better PC4400) is the
perfect answer, but you may find that, if your current memory will run a
little faster than it is ATM, it's possible that the best performance
compromise may come from running 1:1 at a lower FSB. Doing it that way would
also reduce the CPU's power demand, which would make the power supply
circuitry happy if nothing else! :)

> Temperature at load (Prime95 maximum torture setting):
> Sys: 37c | CPU: 75.5 +/- 5c | PWM: 75.0 +/- 5c

That PWM reading is very, very hot. Unsurprising given the amount of current
the CPU will be pulling under load. It's also possible that the CPU is
throttling, cutting back the performance. You can check this by downloading
ThrottleWatch and leaving it run during periods of extended intensive
utilisation.

> Are these settings and temperatures safe?

I'd be uncomfortable running this system with the PWM temps as high as they
are, while I'd also want to check for CPU thermal throttling before deciding
whether the existing thermal solution on that was sufficient.

At minimum I'd be thinking about heatsinking the power MOSFETS and directing
a source of cold air directly onto that area of the board.

> Am I walking on thin ice, or living on borrowed time?

Questions like that are always almost impossible to answer with certainty or
evidential backing, unless you're a motherboard manufacturer with access to
proper test data. Which, of course, none of us are.

> Any suggestions on a better heatsink fan combo

There are a few heatsink/fan assemblies that'll do a better job than the
stock product supplied with the CPU. Difficult to suggest one in particular.
--


Richard Hopkins
Cardiff, Wales, United Kingdom
(replace nospam with pipex in reply address)

The UK's leading technology reseller www.dabs.com
 
G

Guest

Guest
Archived from groups: alt.comp.hardware.overclocking (More info?)

"Richard Hopkins" <richh@dsl.nospam.com> wrote in news:4209fe75$0$16577
$cc9e4d1f@news-text.dial.pipex.com:

First of all thanks for the thorough and very informative reply. I've
done some case baffling to dirrect the air in a more efficient manner and
have droped average case, CPU and PWM temps under Prime95 maximum torture
settings by ~3c. While significant I'm assunig that's not enough. I do
have a question and I'm not asking for an implied warrenty, and if I blow
this thing up it's nobodies fault but my own, but what is the maximum CPU
and PWM temp's you'd be comfortable running. This thing ran in the high
60's C on Prime95 at stock 200MHz / 3.0GHz CPU speed. I'd say it gained ~
8-10C at 267MHz.

> ThrottleWatch

Here's the ThrottleWatch data @ 4.005GHz The first throttled data point
was at CPU temp of 68c. No throlling seemed to occur until that temp. Max
temp was 72.5C +/-5C on CPU and 71.5C +/-5C on PWM; temp fluctuates up
and down quite a bit on the ABIT EQ monitoring app, but never exceeded ~
77C.

ThrottleWatch Journal file
--------------------------

Time Load Throt. Clock
460 51 11 4034
1459 45 11 4047
2462 37 0 4009
3455 51 0 4009
4459 52 0 4020
5459 51 55 4036
6461 55 22 4009
7462 52 11 4032
8456 55 22 4010
9460 51 22 4036
10458 55 22 4022
11458 51 20 4036
12455 51 22 4040
13460 50 33 4033
14468 52 11 4010
15455 52 33 4043
16454 51 22 4038
17555 50 44 4009
18457 55 10 4009
19459 50 33 4040
20459 55 55 4009
21457 51 44 4040
22457 56 11 4013
23459 52 44 4029
24458 52 22 4036
25459 51 20 4045
26460 51 33 4041
27458 55 55 4009
28456 51 22 4031
29455 55 22 4009
30457 51 33 4038
31454 55 33 4009
32454 52 20 4037
33461 55 44 4022
34459 51 44 4029
35462 50 44 4028
36455 51 44 4060
37454 52 22 4047
38455 51 33 4045
39458 52 20 4032
40461 55 33 4009
41455 51 44 4014
42458 55 11 4009
43456 51 66 4015
44465 56 22 4019
45460 50 44 4037
46459 51 40 4028
47557 51 22 4009
48463 55 22 4009
49459 51 22 4028
50457 51 22 4042
51460 50 33 4031
52458 55 22 4006
53460 51 30 4027
54463 55 55 4009
55462 51 33 4034
56459 55 55 4023
57458 51 33 4048
58458 55 77 4012
59458 52 66 4040
60457 51 30 4035
61458 51 22 4041
62461 51 22 4039
63455 55 44 4002
64460 52 0 4019
65459 55 33 4009
66459 51 33 4040
67461 55 70 4008
68462 51 33 4048
69459 56 33 4015
70460 51 55 4049
71454 51 100 4045
72459 52 44 4027
73458 51 55 4044
74460 55 0 4009
75464 51 11 4936
76513 54 22 4009
77566 52 44 4020
78454 55 44 4020
79459 55 22 4011
80460 52 33 4047
81461 55 0 4018
82458 52 55 4040
83462 51 55 4044
84463 51 55 4037
85458 52 33 4039
86463 55 77 4009
87456 52 22 4044
88464 55 70 4001
89464 52 33 4050
90462 55 77 4009
91464 51 44 4027
92455 58 33 4022
93456 51 55 4036
94454 55 11 4027
95459 50 70 4029
96459 51 55 4038
97460 55 66 4009
98459 53 22 4029
99457 55 44 4009
100458 50 22 4025
101457 55 55 4009
102460 53 30 4034
103458 55 33 4015
104467 52 11 4035
105461 52 44 4031
106460 51 44 4030
107557 52 44 4020
108455 55 33 4020
109455 57 50 4009
110459 52 22 4031
111459 55 66 4009
112459 51 33 4033
113460 56 55 4017
114460 52 22 4034
115459 55 44 4019
116455 52 30 4032
117457 51 33 4032
118456 51 33 4018
119459 52 33 4037
120455 55 33 4009
121454 55 11 4042
122459 58 33 4009
123456 51 30 4041

Average load 52%
Average throttling 34%

Prime95 Benchmark Results:

Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 4009.14 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: unknown
L2 cache size: 1024 KB
L1 cache line size: unknown
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 10.206 ms.
Best time for 448K FFT length: 12.443 ms.
Best time for 512K FFT length: 13.979 ms.
Best time for 640K FFT length: 16.765 ms.
Best time for 768K FFT length: 20.182 ms.
Best time for 896K FFT length: 24.361 ms.
Best time for 1024K FFT length: 27.124 ms.
Best time for 1280K FFT length: 36.160 ms.
Best time for 1536K FFT length: 43.214 ms.
Best time for 1792K FFT length: 52.054 ms.
Best time for 2048K FFT length: 58.215 ms.

Thanks, -Barry
 
G

Guest

Guest
Archived from groups: alt.comp.hardware.overclocking (More info?)

"baj2k" wrote in message...
> First of all thanks for the thorough and very informative reply.

No worries. :)

> I've done some case baffling, dropped average case, CPU
> and PWM temps under Prime95 maximum torture settings
> by ~3c. While significant I'm assunig that's not enough.

Not really. Other folks here may have different opinions of course but if
that were my system I'd like to see the PWM temp significantly lower, and
the CPU cooling improved enough to eliminate the throttling.

> but what is the maximum CPU and PWM temp's you'd be comfortable
> running.

As I say, opinions may vary but I wouldn't be truly comfortable with the
whole thing until the PWM temp was down below 55C or so under load. The CPU
temperature is a slightly different case. Prescotts run hot, and there's
very little you can do about that. However, you need to cool it sufficiently
well enough that it doesn't throttle. If it does, there's very little point
running it at the sort of speeds you are, as, if all it does is throttle
whenever you use it, you're not actually getting the performance you should
be, if you see what I mean.

> This thing ran in the high 60's C on Prime95 at stock 200MHz
> / 3.0GHz CPU speed. I'd say it gained ~ 8-10C at 267MHz.

Yeah, this is the problem with these bloody things unfortunately. Fricking
silly power consumption and heat dissipation before you even start
overclocking.

> Here's the ThrottleWatch data @ 4.005GHz The first throttled data
> point was at CPU temp of 68c. No throlling seemed to occur until
> that temp.

Indeed. What you then need to aim for is a situation where you can keep the
CPU under this temperature, even under sustained and full utilisation.

> Max temp was 72.5C +/-5C on CPU and 71.5C +/-5C on
> PWM; temp fluctuates up and down quite a bit on the ABIT EQ
> monitoring app, but never exceeded ~ 77C.

The throttling prevents the CPU from lunching itself by inserting increasing
amounts of halt cycles once the temperature starts to reach danger level.
This is a handy safety mechanism, but it also means that any gains you make
from overclocking are going to be offset - and possibly even lost
completely - if the throttling is severe.

> Average load 52%
> Average throttling 34%

That's a significant statistic. I haven't done any in-depth testing with
ThrottleWatch, primarily because my setup doesn't run hot enough to
throttle, and thus don't know whether its readings pan out in reality.

If they do, in effect it means that your 4.005GHz CPU turns into something
like a 2.65GHz CPU whenever you try and do anything significant with it. If
the ThrottleWatch results are correct, it also, significantly, means that
you're getting less performance at 4GHz than you would at the CPU's default
speeds - provided of course that CPU then ran cool enough to keep it under
the throttle threshold.

> Prime95 Benchmark Results:

As a matter of interest, have you tried running Prime at default CPU speed?
It'd be interesting to see how the results compare, and whether the
throttling is indeed impacting performance to the extent ThrottleWatch is
suggesting.
--


Richard Hopkins
Cardiff, Wales, United Kingdom
(replace nospam with pipex in reply address)

The UK's leading technology reseller www.dabs.com
 
G

Guest

Guest
Archived from groups: alt.comp.hardware.overclocking (More info?)

"Richard Hopkins" <richh@dsl.nospam.com> wrote in
news:420aa151$0$16572$cc9e4d1f@news-text.dial.pipex.com:

> As a matter of interest, have you tried running Prime at default CPU
> speed? It'd be interesting to see how the results compare, and whether
> the throttling is indeed impacting performance to the extent
> ThrottleWatch is suggesting.


Here's the results, surprisingly no apparent difference when throttling
up to 33%.

Prime95 Benchmark results as requested:

200/66/33

Cold - No throttling - Before Prime95 Torture Test
--------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 3006.85 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: unknown
L2 cache size: 1024 KB
L1 cache line size: unknown
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 13.515 ms.
Best time for 448K FFT length: 16.461 ms.
Best time for 512K FFT length: 18.552 ms.
Best time for 640K FFT length: 22.104 ms.
Best time for 768K FFT length: 26.584 ms.
Best time for 896K FFT length: 32.062 ms.
Best time for 1024K FFT length: 35.689 ms.
Best time for 1280K FFT length: 47.582 ms.
Best time for 1536K FFT length: 56.890 ms.
Best time for 1792K FFT length: 68.379 ms.
Best time for 2048K FFT length: 76.474 ms.


Hot - No throttling - After 10 min Prime95 Torture Test
-------------------------------------------------------

Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 3006.90 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: unknown
L2 cache size: 1024 KB
L1 cache line size: unknown
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 13.522 ms.
Best time for 448K FFT length: 16.417 ms.
Best time for 512K FFT length: 18.537 ms.
Best time for 640K FFT length: 22.084 ms.
Best time for 768K FFT length: 26.578 ms.
Best time for 896K FFT length: 32.110 ms.
Best time for 1024K FFT length: 35.709 ms.
Best time for 1280K FFT length: 47.621 ms.
Best time for 1536K FFT length: 56.845 ms.
Best time for 1792K FFT length: 68.380 ms.
Best time for 2048K FFT length: 76.365 ms.


267/66/33

Cold - No throttling - Before Prime95 Torture Test
--------------------------------------------------

Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 4009.11 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: unknown
L2 cache size: 1024 KB
L1 cache line size: unknown
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 10.143 ms.
Best time for 448K FFT length: 12.348 ms.
Best time for 512K FFT length: 13.931 ms.
Best time for 640K FFT length: 16.638 ms.
Best time for 768K FFT length: 19.955 ms.
Best time for 896K FFT length: 24.166 ms.
Best time for 1024K FFT length: 26.890 ms.
Best time for 1280K FFT length: 35.734 ms.
Best time for 1536K FFT length: 42.728 ms.
Best time for 1792K FFT length: 51.370 ms.
Best time for 2048K FFT length: 57.345 ms.


Hot - Constant, up to 33% throttling - After 10 min. Prime95 Torture Test
-------------------------------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 4009.11 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: unknown
L2 cache size: 1024 KB
L1 cache line size: unknown
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 10.136 ms.
Best time for 448K FFT length: 12.358 ms.
Best time for 512K FFT length: 13.925 ms.
Best time for 640K FFT length: 16.712 ms.
Best time for 768K FFT length: 19.985 ms.
Best time for 896K FFT length: 24.193 ms.
Best time for 1024K FFT length: 27.678 ms.
Best time for 1280K FFT length: 35.815 ms.
Best time for 1536K FFT length: 42.793 ms.
Best time for 1792K FFT length: 51.682 ms.
Best time for 2048K FFT length: 57.333 ms.

Your thoughts, perhaps Prime95 benchmarking is not a good test for this,
maybe you could suggest another.

Cheers, -Barry