3.06@3,560 vs 2.66@3,300 an observation

[PClark99]

This is Technonut's chip. It came in the mail today.

I did a fresh install of Windows XP Pro, installed SP1, chipset drivers, DX9, vid card drivers, IAA, and other drivers in that order. I am using the Catalyst 3.1 version.

My board is running the BT modded F3 bios. It is overvolting the chip to 1.61V actual. Have 2 x 512 MB XMS3200C2 in dual channel mode. I am using a Coolermaster IHC-H71 with the 6k rpm stock fan.

This board ran 3dmark2001SE at 150 FSB at 1.61V. Thats 3,456 on air. Sandra mem score 3657/3653, 3dmark2001SE 15,636.

Next I bumped it up to 155 FSB at the same voltage. Thats 3,570 on air. Sandra mem score 3769/3766, 3dmark2001SE 15,940.

CPU temp is idling around 35C. Under load I am not sure. When I dump out of 3dmark2001SE and check EZTuner it is reading about 40C.

Now here is where I am a little confused. After reading the article over on overclockers.com, which in a nutshell purports that CPU speed is more important than memory bandwidth, I was expecting a higher score now than with my other CPU, which is a 2.66@3,300. With that CPU in the same setup, my Sandra memory score was around 4300/4300 and the 3dmark2001SE was 16,258. I know it is only 2% less, but I was wondering whats up with that. Is 3dmark2001SE more biased towards changes in memory bandwidth than CPU speed?

Now there is always a chance I borked something in the install, but everything seems stable, even moreso that with the 2.66.

I am not sure how much I want to push this chip. I have some more testing to do, but at 155FSB it looks pretty good.

I recognize that 3dmark2001SE is not the end all be all of benchmarks, is there something better I should be using to gauge overall system performance? Also I might add that the system seems more responsive with the 3.06 but that could just be me.

Any thoughts are appreciated.

 

[Jojo1971]

kewl

 

[PClark99]

Yeah, I might be keeping this CPU for a while.

 

[THUGSROOK]

low fsb is limiting your scores

if it was 18x198= 3564 it would be much faster.

 

[PClark99]

Thugs,

thanks for the input.


I have some more testing to do. I plan on working from home tomorrow. I will probably loop 3dmark all day to allow the thermal paste to set and to see if it will run the day.

 

[oldfart]

Exactly, its 150 Vs 165 FSB. The FSB speed also gives you a boost. The OC.com article was comparing max CPU speed Vs max memory speed on an SC DDR setup using mem ratios.

 

[THUGSROOK]

thats what the whole "P4m" experiment was about....

3DMark2001
P4 12x200fsb 2.4ghz = 12417
P4 18x133fsb 2.4ghz = 12001

quite a boost eh?

 

[PClark99]

I see.

Hmmm. Still, there is only a few % difference and it seems that the CPU at a higher clock makes the system more responsive (purely a subjective feel/observation) so far.

 

[machu]

Yeah, I read that article and I totally disagree with it. I think FSB makes all the difference
My test using Barton 2500+ on NF7-s
3DMark2001
13x170fsb 2.2ghz = 17041
10x220fsb 2.2ghz = 17828

 

[PClark99]

Just a thought, but would I be better off with this CPU if I had say a 8SQ800 that I could run at Dual Channel DDR400 or so?

 

[EdipisReks]

Originally posted by: PClark99
Just a thought, but would I be better off with this CPU if I had say a 8SQ800 that I could run at Dual Channel DDR400 or so?

my guess would be yes. also, keep in mind the 3.06's Hyper Threading.

 

[Technonut]

Nice job so far PClark99... I kind of miss the 3.06, but am glad it has a good home.

 

[PClark99]

Thanks.

I gotta leave you some heat.

I was really sad to hear about your 2.53. Did you ever figure out what caused it to go buh bye?

Left you some heat.

 

[Technonut]

Thanks for the Heat... I will post some for you as soon as I get the password I have fogotten e-mailed to me..

I was really sad to hear about your 2.53. Did you ever figure out what caused it to go buh bye?

Yes, it was a bummer, but nothing that I have not been through before in years of OC'ing. I probably will not know for certain what caused the meltdown.. I do know that I will not be throwing another processor in that BH7 to find out..

Things are looking up... I have another 2.53 from Outpost scheduled to arrive today, and another BH7 on the way... Call me a glutton for punishment, but I like the BH7.

 

[PClark99]

The BH7 has the features and you are probably right, there are som many things it could have been.

Well this CPU is good, but it won't loop 3dmark more than 3 times before kicking back to the desktop at 3560,

I am dropping it back to 3500 and trying it there.

I am at 1.61V and my temps seem ok. Maybe more voltage is the answer, but I want to find the ceiling at the voltage I am at now and work from there.

 

[PClark99]

Looped 3dmark2001SE for 8 hrs straight. No problems at all at 3500.

Checked core temp on CPU right after I dumped out of the benchmark, 42C.


All in all not too bad.

 

[stardust]

wow u get pretty darn low cpu temps for such an overclock...i used to have my old 2.66 @ 3300 and it wouldn't dip lower than 46degrees...hmm and i was using a TT spark 7 fan too with arctic silver thermal grease...vcore: 1.65

tell me more about overclocking ur 2.66...ever try anything higher than 3300 on air?

 

[PClark99]

I did try 3400.

at 1.6V it would boot into windows fine. But 3dmark would make it lock.

 

[stardust]

i reread sum posts above and noticed the whole higher FSB=better performance issue. ur running a P4...so how would u decrease ur multiplier?
is the processor unlocked? I was only aware of athlons that had this issue.