David's Learning and Rambling Thread

[EarthDog]

Green setting? Sir, you are benchmarking not surfing the web, grow a pair!

Being serious, where are you at with Vtt, and NB voltage?

 

[David]

Haha!

Checking my scraps of notepaper, I had:

• VCore: 1.6 V
• CPU PLL: 1.60 V
• Northbridge: 1.55 V
• FSB termination: 1.5 V
• Southbridge: 1.15 V

I didn't mess with DRAM setting as the RAM was running well within spec.

 

[EarthDog]

PLL I thought default was 1.8?

But yeah, that should be enough voltage to get you were you are at... thats for sure.

 

[David]

I'll check, but by default I'm sure the PLL was running around 1.5 V. What's the view on LLC? Yay or nay?

How high is too high for the NB? I don't care if I fry a CPU, but I'll cry like a baby if I kill the board.

 

[EarthDog]

I would imagine around there on air... Its been a while since I have been on s775 though...

Im learning again on e6600 and e6750.

 

[David]

The NB temp was pretty high so I have a fan aimed at it, which brought it back down quite a bit.

 

[David]

Got 4.347 GHz validated on 1.5175 V on the 925, but it's way too hot to bench. Trying to back off the voltage a little, I think I'm giving in and ramming up the VCore too quickly.

 

[Bobnova]

You need SetFSB, usually you can get a few MHz more fsb in windows with setfsb vs bios settings.

EDIT:
775 PLL is 1.5v stock. Over 1.6 rarely helps.

 

[Janus67]

yeah I've always gotten a lot more using setfsb than trying to boot with different FSB.

 

[David]

Got my 4.5 GHz, so I'm chuffed.

http://www.hwbot.org/submission/2244997_davejn86_cpu_frequency_pentium_4_d_925_4500.01_mhz

Definitely glad I have a pot on the way, because this thing is roasting.

EDIT: Yep - need more volts to stop it BSODing, but more volts = CPU overheating warnings. This one's likely spent until the DICE fun begins.

 

[David]

Was hoping my 630 would have arrived, but no such luck

Maybe that's fate telling me to write my damn thesis.

 

[David]

Got the Celly D 336 to work last night, but it only boots after a CMOS reset. I think on subsequent boots it tries to kick the FSB up to 200 and it can't handle it, even if I set vcore to 1.6 V and bump up the NB voltage.

I can boot into Windows and use SetFSB up to around 180 MHz FSB (http://hwbot.org/submission/2245359_). We don't have any team entries for this chip so I'll probably rattle through the 2D benches to get the ca. 2 TPP (per test) they are worth.

I also acquired a P4 550 @ 3.4 GHz which might help me get a higher valid on air (or not ). We're also lacking some scores for it.

I doubt I'll get many hwboints until I get them all under DICE, to be honest.

 

[David]

Hrmm... why is my Celly D going in as a 2x CPU?
http://www.hwbot.org/submission/2245755_

I'm unfairly getting 2 HW points for this ... can I fix it?

 

[Bobnova]

PM Massman, sometimes the the HWBot software screws up and nobody can touch it bit him and their mod people.
You did everything correctly as far as I can see.

 

[David]

I think a couple have boobed. I'll get my wPrime 1024M and Superpi 32M scores done then I'll see which I need to report.

My P4-550 arrived (fast!) but I'm holding off testing it until I get the complete set of Celly-D tests onto the 'bot. That, and I have a thesis to write

Mmmm... processors

 

[David]

More Google spreadsheets to keep track of targets, etc.
https://docs.google.com/spreadsheet/ccc?key=0AiJS0Yov_Q8jdF9fZV9pUFB1OVVTRWtqaFNwYmM4SXc

My P4-D 880 has also arrived, so I think that's everything for now. Should be plenty to play with

 

[David]

Had a quick crack at the 550 - got 4252 (250 MHz) FSB. I had to boot at 240 MHz FSB to get into windows, and got BSOD with 255 MHz FSB (SetFSB). All this with a vcore of 1.575 V and LLC enabled, plus 1.49 V on the northbridge.

http://www.hwbot.org/submission/2248026_davejn86_cpu_frequency_pentium_4_550_4252.81_mhz

Feels like a brick wall :-\


EDIT: Am I better with HT on or off? I imagine it'll help for wPrime, but not SuperPI?

 

[Bobnova]

Good for WPrime and UCBench, probably not for everything else.

 

[David]

Struggling to keep the P4-550 stable enough to bench it. My best validation was at 4340.6 (only 190 MHz away from getting > 0.1 HW points ...) but I've had to back right down to ~4.2 to get wPrime32M to run. wPrime1024M will maybe need to go down further - Windows keeps encountering critical errors that require reboots ...

The temperatures aren't hellish (ca. 65 C running wPrime32M) and I have the Vcore at 1.6 V with LLC enabled, and NB at 1.49 V. Any higher on the Vcore appears to be unstable.

Any tips on improving stability? I'm rapidly learning to redefine stability as "long enough to run a benchmark and nab a screenshot"...

 

[hokiealumnus]

Sorry, I don't know jack about P4's, but as far as...

I'm rapidly learning to redefine stability as "long enough to run a benchmark and nab a screenshot"...

...is there another definition?