David's Learning and Rambling Thread

[hokiealumnus]

Now that's money well spent. Great benching!

 

[David]

So, 540 run through the mill. Completely heat limited (needs moar voltage, can't give it moar voltage or it shuts off... sits at 70C in the BIOS with 1.7 V ).

The 521 and 551 are also heat limited I'm sure (impending P4 dice-fest in April I think).

I'm back to playing with the Celly D - got a few more MHz out of it (went from 66/78 to 62/78) but I'm definitely being held back by the amount of juice I can feed it, but I can't boot with anything other than BIOS defaults The Formula doesn't go below 200 MHz FSB, but I can't get it to boot at 4.2 GHz, even if I throw silly voltage (1.8 V) at it.

On the plus side, I have a Celeron 420 on the way to play with. Narrowly missed out on a cheap E1500, forgot to bid in time! Fail.

 

[David]

Ugh... RAM running well below its capability at default as well, which is going to shaft benchmarks. I think I need a different board for Celeron 3xx-bashing ...

 

[Woomack]

I'm not sure if you will be able to make good clocks with 533 FSB cpus ( Celerons/P4 ) on X48 board. I have ~220-230MHz FSB wall on most "newer" boards ( P35/P45 ) with Celeron D326 and P4 2.66.
Most boards have also max cpu voltage at about 1.7V in BIOS and ~1,65V under load. I think that you won't have that problems with your board ... you will need about 1,8V or more to make good scores on dice ( I mean like something near to 5GHz on P4/P-D/Cel-D ).

good work so far

 

[David]

This 521 is no fun. I can't figure out where my limits are, and I'm all over the place :-\

I managed to get a 4.2 GHz valid previously, but can't get past 4 GHz at the moment. I'm having issues staying stable at 275 FSB, more voltage makes Windows fail to boot, less voltage makes SuperPi crash more.

I've backed off right down to 266, hitting heat issues under load so I'm hoping to find a sweet voltage spot where I can still run at 266 without crashing, so I can run the RAM at 1066 1:1 and hopefully register some boints. We have few benchmarks done with the 521 so I should net some TPP either way.

 

[David]

I'm not sure if you will be able to make good clocks with 533 FSB cpus ( Celerons/P4 ) on X48 board. I have ~220-230MHz FSB wall on most "newer" boards ( P35/P45 ) with Celeron D326 and P4 2.66.
Most boards have also max cpu voltage at about 1.7V in BIOS and ~1,65V under load. I think that you won't have that problems with your board ... you will need about 1,8V or more to make good scores on dice ( I mean like something near to 5GHz on P4/P-D/Cel-D ).

good work so far

Dziekuje!

I do plan to put everything under DICE eventually, but I want to wait until I've time to be properly prepared, ideally with a decent SSD (looking at this at the moment: http://www.scan.co.uk/products/60gb...orce-2281-read-525mb-s-write-475mb-s-80k-iops) set up with XP/7/Vista for the relevant benchmarks so I can make the most of the DICE.

 

[Woomack]

Dziekuje!

I do plan to put everything under DICE eventually, but I want to wait until I've time to be properly prepared, ideally with a decent SSD (looking at this at the moment: http://www.scan.co.uk/products/60gb...orce-2281-read-525mb-s-write-475mb-s-80k-iops) set up with XP/7/Vista for the relevant benchmarks so I can make the most of the DICE.

Agility is bad for pcmarks. Better get Crucial M4 or look for something like Force GT or Patriot Pyro SE. Prices are dropping so difference shouldn't be so big for ~60GB SSD. Remember that pcmarks = 99% read speed so slow writes won't make any difference.

I was benching P4 520 some days ago and I also didn't like it . It's almost the same as 521. Max on SS and really limited board voltages was about 4.5GHz / 325FSB
P5K WS is really bad board for tweaking. I couldn't even change memory timings or it couldn't boot ( on auto all was fine ).
I remember that some P4 had something like FSB hole ... when you couldn't boot @255MHz but you could @280+. It's something about FSB strap. To try that you will need better cooling.

 

[David]

Cheers for the advice on SSDs - I'll keep my eyes open. Pay day is Monday ... so I'll look for an M4 or Force GT, ideally for < £1/GB.

The 521 is infuriating. I've managed to run all the shorter tests at 3.73 GHz with ~ 1.46 V, but it wont stay stable for minutes at that voltage if the HT is used (wPrime 1024). I've had to gradually back it off.

Just going to try and knock through each of the tests then fire it back in the box until I can have a crack at it with better cooling.

 

[keny]

I have been pondering over this but wether it is worth the extra tenner over the agility I'm not sure.

 

[Woomack]

Point is that asynchronous NAND is noticable slower than synchronous in some operations. Compare tables for Agility and Vertex 3 60GB, especially sequential read for AS-SSD or IOPS
http://www.ocztechnology.com/res/manuals/OCZ_Agility3_Product_sheet.pdf
http://www.ocztechnology.com/res/manuals/OCZ_Vertex3_Product_Sheet.pdf
If you want SSD for benching then synchronous one is what you are looking for.

Agility 3 = Corsair Force = Patriot Pyro
Vertex 3 = Corsair Force GT = Patriot Pyro SE
Revo 3 X2 = 4x Agility3

Also there is no point to get more than 2 SSD if you are going to use onboard controller. Both AMD and Intel platforms make max about ~1050MB/s no matter if you have 2 or 6 SATA3 ports. In my case adding 3rd Crucial M4 gave me ~80MB/s more max sequential read and read is almost all you need for pcmarks ( at least these for which you get points ).

 

[David]

The Celeron 420 is rubbish. Feels like a wall at just short of 2.8 GHz. More CPU volts and/or more NB volts do nothing

Running the tests just because we don't have them done for a 420.

http://hwbot.org/submission/226179_


EDIT: Actually, it's just UC and PCMV we don't have :-\

 

[Bobnova]

Mine got higher, but still hit one hell of a wall. DIce increased clocks by like 70MHz.

 

[David]

Yeah, same wall here from stock volts all the way up to 1.75 V, with NB from 1.49 to 1.55 V. A bit peeved, but there's little I can do about it.

Needs moar chips!!

 

[David]

Took the 550 under (chosen randomly) dry ice* today - managed to squeeze another 400 MHz out of it, and get most benches into proper boint-scoring territory (in my book, that's >0.5 ). Very very happy. However, the 550 at 1.65 V is eating dry ice like it's going out of fashion. Running wPrime1024 at the moment.

Once I've finished wPrime1024 (not going to bother with the PCMark benches, they take too long) I'll put the P-D 820 under (most boints, at a glance).

I had half a bag of dry ice and I've used ~2/3 of it. I reckon I used a lot getting familiar with the whole setup, as I had some issues getting things seated properly for good contact.

* Can't get LN2 at the weekend unless I fill a big dewar the day before and get raised eyebrows from the postdoc and/or boss. They give away half-bags of dry ice on a Friday for free, because they get more in on Monday anyway and they'll never last over the weekend.

 

[David]

Well ****. I ran out of dry ice - I had *just enough* to get through the wPrime1024.

Newbie mistakes to avoid next time:
- Don't waste time on chips that aren't as popular. Get the high-boint chips done first.
- Properly allow the TIM to cure... I don't think I did.
- Don't go nip into the office for food, get distracted, and come back to find the computer off, the pot hot, and half of the acetone gone.

 

[Bobnova]

- Don't go nip into the office for food, get distracted, and come back to find the computer off, the pot hot, and half of the acetone gone.

That's an important one, lol.
I've found that I almost always either have too much DIce or too little, it'll drive you nuts

Glad the first session went well, it sounds more productive than my first was. That's the dirty little secret of extreme cooling, the first session is usually dubious at best. Yet it's strangely addictive despite that

Lastly and possibly most importantly: Congrats and welcome to being a genuine Extreme Bencher!

 

[keny]

Dry ice for free is always a bonus, but running out not so much.all of those "not going to go cold until summer" comments I will just ignore lol, I wish I could get dry ice as easy as you, it's going to cost me just shy of £50 to get 10kg, man this benching is pi**ing my wife off more each day lol , I'm hopeing to join the dice crew later this month,

 

[Bobnova]

It's summer somewhere, right?

 

[keny]

It's summer somewhere, right?

There's always an excuse somewhere if you look hard enough

 

[David]

Dry ice for free is always a bonus, but running out not so much.all of those "not going to go cold until summer" comments I will just ignore lol, I wish I could get dry ice as easy as you, it's going to cost me just shy of £50 to get 10kg, man this benching is pi**ing my wife off more each day lol , I'm hopeing to join the dice crew later this month,

Where are you getting it from? We get it in from BOC, if I remember rightly, and the in-house price is £4/scoop or £16/bag. If I could steal you half a bag one Friday I would, but I don't think I could sneak it off-site.

I don't know when I'll go cold again - today was ideal: dry ice left over this week, the boss down in England for the weekend, no-one in the lab. today. I don't know that I'll get the same 'perfect storm' for a while again yet ...

Ideally I'd love to have an SSD set up with three O/Ss all tweaked for the appropriate tests but it'll have to wait until this summer when I can afford it.