It seems mobile 2600's are better than the others

[ectx]

What do you guys call "prime stable for x hrs"? You run (a) small ffts (b) in-place large fft's, (c) blended or (d) cusotm?

I have found that custom tests, setting min fft to 2048, really torture bartons (if you have ht p4 - you need to run another app to torture it). If you can run the big fft torture tests for >12 hrs than you have my congratulations. If you use small fft tests - please try the big fft custom tests.

If you have an earlier version of p95 - you can edit the ini file to do the custom fft tests. But it would be easier to get 23.8.1.

 

[myocardia]

Man, I'm forgetful! I'm now running at 12x210.8 (2,530 mhz), because although my board will now go at least as high as 222fsb, my memory isn't playing along. The memory won't go above 215fsb at 2.5-3-3-11 timings, even at max vdimm for this board. For anything above, I have to use 3-4-4-11, which is slower than 12x210.8 with the faster timings. Anyway, I'm quite happy with this chip, even though I had to raise the vcore another .025v to be stable at the speed I'm running it at now (1.760-1.776v actual, under 100% load).

 

[myocardia]

Originally posted by: ectx
What do you guys call "prime stable for x hrs"? You run (a) small ffts (b) in-place large fft's, (c) blended or (d) cusotm?

I have found that custom tests, setting min fft to 2048, really torture bartons (if you have ht p4 - you need to run another app to torture it). If you can run the big fft torture tests for >12 hrs than you have my congratulations. If you use small fft tests - please try the big fft custom tests.

If you have an earlier version of p95 - you can edit the ini file to do the custom fft tests. But it would be easier to get 23.8.1.

I had always just used the "blend" tests, until reading your post earlier. At the moment, I'm runnning "custom" with a minimum of 2048 and maximum of 4096, at 2530mhz. At first, I had it setup to do 4096 minimum and maximum, and it passed the first iteration of that, but I decided to try 2048 and up, to see what happened. Since this is what I plan on leaving it at, I'm going to let it run for at least 12 hours.

edit: Oh yeah, I recently wiped my hard-drive, and I can't find Prime95 2.37, so last night I downloaded 2.38, and that's what I'm running now.

 

[ectx]

Myocardia,

Thanks for giving it a shot. I made my observations over about 6 chips recently.If anyone else can give me some input, it would greatly help. I did get someone in the for-sale forum to test a barton chip and he got an error much quicker using big fft custom test.

I do believe blend tests are decent but the big test give me quicker results. Small fft tests are of little value (have a 1700+ that can pass small fft tests for 48hr but cannot pass the big test for 1 hr).

 

[myocardia]

Well, I run a DC (distributed computing) project, and I want my computer to be as stable as a mission-critical server would be, so anything that can help me find out if it's completely stable is a help to me. So, thank you for telling me about that, since I'm not going to devote 3 or 4 days worth of computing time, just to test for stability. As a matter of fact, I will use this way of testing for all computers that I test for stability from now on.

 

[Dantzig]

Wow, some of the vcore numbers in this thread are crazy, especially with air cooling! How long will a Barton chip last @ 1.9v?

My 2400+ does 2.4GHz @ 1.7v perfectly stable but needs a good amount more voltage to pass Prime95 for 24 hours at 2.5+GHz.

BTW, the blend torture test seems to work just fine for me. I always leave it going for at least 24 hours or until I get errors. Once I think I have a stable configuration, I go ahead and leave it running for ~72 hours while I use the system normally. If it passes that, it's golden.

 

[myocardia]

Hah, you should go browse xtremesystems.org! Nobody there is using less than 2.0V on their Bartons! Although, I have always tried to err on the side of caution, when it comes to voltage. That's why I'm "only" running mine at 2530mhz. Oh, and nobody knows how long an Athlon will run at any particular voltage, because nobody has ever killed one with less than 1.9v or 1.95v.

edit: It's generally agreed upon by those who have the most experience that anything less than 1.65v actual is safe long-term for a P4, and 1.75-80v actual is safe for an Athlon or Duron.

 

[ectx]

Originally posted by: Dantzig
BTW, the blend torture test seems to work just fine for me. I always leave it going for at least 24 hours or until I get errors. Once I think I have a stable configuration, I go ahead and leave it running for ~72 hours while I use the system normally. If it passes that, it's golden.

I am convinced that small fft tests are of little value but not 100% sure large fft custom tests are better. But I do know, after years of running p95 that (a) if you oc too much, any tortute test will give you an error really soon. (b) if you are on the margin, it almost always errors out on bigger fft tests. So I take a leap of faith and assumed it would be suffice to do bigger fft tests only. I maybe full of sh*t. But it would be easier to prove/disprove my theory if I get someone else to help me test.

 

[myocardia]

Originally posted by: ectx

Originally posted by: Dantzig
BTW, the blend torture test seems to work just fine for me. I always leave it going for at least 24 hours or until I get errors. Once I think I have a stable configuration, I go ahead and leave it running for ~72 hours while I use the system normally. If it passes that, it's golden.

I am convinced that small fft tests are of little value but not 100% sure large fft custom tests are better. But I do know, after years of running p95 that (a) if you oc too much, any tortute test will give you an error really soon. (b) if you are on the margin, it almost always errors out on bigger fft tests. So I take a leap of faith and assumed it would be suffice to do bigger fft tests only. I maybe full of sh*t. But it would be easier to prove/disprove my theory if I get someone else to help me test.

Hah, you already have! Prime95 errored about 10 minutes after I posted that, so I rebooted (you should always reboot after Prime fails), and have been running it at 12x210.0 for around an hour or so now, still using 2048 minimum and 4096 maximum, and it hasn't had a problem so far. But, without knowing this, I would have had to spend a few days finding out what it will run at, instead of less than an hour! Thanks, man! Oh, here, have a beer on me::beer:

 

[katcher40401]

my only concern witht he mobile bartons...i was looking at newegg and all the mobile processors are 266 bus while desktop bartons to my knowledge are 333 bus...someone mind clarifying for me?

 

[myocardia]

Yeah, the mobile Bartons are the best of the best Bartons, that's why they overclock so high. They will run at their "rated" speed, but with much lower voltage (1.45v for the mobiles and 1.65v for the desktops). They are also unlocked, which is cool, and helps overclock. Anyway, most of them will actually run a higher fsb than any of the desktops will, especially the newer desktops, which are having problems reaching even 200fsb. Mine will actually run at 222fsb, but either my memory or my board is holding me back. It will run it just fine, but it isn't Prime95 stable, and that's very important to me. So, I've chosen to run it at a slightly lower fsb, with a 12x multi. But there are a few people who've got theirs running 10x250. Oh, I almost forgot, they spec them at a 133fsb because the only chipset designed for laptops has a 133mhz fsb.

edit: Dang, I forgot to include the link that I finally found that shows a mobile Barton destroying a 3.4ghz Prescott in every test they ran, except for video encoding and a couple of memory-intensive things that the P4's quad-pumped fsb is able to take advantage of, so here it is: link. Oh, the benchmarks don't start until page 5, which you can jump directly to, with the drop-down box over on the right side.

 

[Tango57]

Originally posted by: myocardia But, if you are still considering buying a mobile Barton, and you are looking for 2500mhz or more, you should definitely consider that extra $8 (over the price of the mobile 2500) as insurance that you will be able to get there.

thanks myocardia! this is just the post i was looking for to read. i think what you said tipped me in favor of the mobile 2600+ as i've been thinking the past few days about which mobile barton to buy and reading all the reviews. i'm gonna go ahead and order one tonight from newegg! can't wait!

 

[myocardia]

Okay, folks, here's an update. I ran a custom torture test (minimum of 2048, maximum of 4096) in Prime95 for ~12 hours, and came very close to posting in this thread that at my slightly lower speed of 12x210.0, I had completely passed. Then, I got a bit busy, and when I came back, after 13½ hours, Prime95 had stopped, and said something along the lines of I had reached the maximum amount of warnings (no errors), which is 100. So, I rebooted, and decided to try 12x208.8 (I boot at 12x175, then raise the fsb using ClockGen, that's why it can have odd numbers on the fsb). I ran that for ~6-7 hours, and decided to stop it, to see if I had any warnings yet, which I didn't. Anyway, I've restarted it, and I'm roughly at the 9 hour mark now, at the same speed. So, it looks like it wasn't Prime95 stable like I had thought it was at 12x210.8, but, I'm still slightly over 2.5ghz (I lost a total of 25 mhz, making it completely stable), and I can live with that! Thanks alot, ectx, I will be telling others about your findings, for sure.

 

[ectx]

Thanks for your tests, myocardia.

It does look like Barton 2600+mobile is the best of the mobile barton as you suggested. I never noticed the price difference is only $8-$9, either. Makes the decision a no brainer in my opinion. (I run through too many chips and decided to buy used ones. I think 2600+ demanaded a premium in the used chip market, for a good reason.)

The custom test with larger ffts only detect problems sooner compared to blended tests in my opinion. My thoery is that the problems normally occurs in L2 cache and you want to run something that needs L2 cache constantly. But if my theory is correct, then you should be able to run test using say 384k-1024k and got similar, quicker detection. On second thought, I should not call it a "theory", I have nothing to back it up at all.

I started oc using AMD 586 and really got into it in p2-300 (to 450) days. In those days, stability tests were easy. Now the chipset, memory are also under a lot of stress at 200-300Mhz FSB. Stability testing is a lot more difficult. You really need some other tools (3d apps and econding apps) to make sure your system is completely stable, I am still seraching for the perfect test suite. But Prime95 is always my first test.

 

[ectx]

Sorry - double post.....

 

[myocardia]

Yeah, it's also my first test for stability. Well, I take that back, it's my first after Memtest86, then I move on to Prime95. Although, I've never personally had, for instance, a game fail on a system that is Prime stable. As a matter of fact, the reverse has always been true for me, I can play a game or run benchmarks on a system that's stable, just not Prime stable. Plus, since I STILL haven't bought a DVD burner, I don't do any video encoding, so I could care less whether it's stable running video encoding or editing software! If I did, then it would make a difference to me, of course! Well, I'm at the 11 hour mark, I sure hope it's had no warnings when I stop it in a little while.

edit: Did I ever say that all of this is with a not-so-good stepping? I'm thinking that if I had good stepping, it would run quite a bit higher with the voltage I'm giving it, or the same with less voltage. I'm also thinking that if this were a lower spec'd chip, I wouldn't be able to run at 2.5 ghz, without crazy voltages. Just my 2 cents.

 

[ectx]

myocardia

I got a couple questions for you:

I noticed that you use clockgen. What additiona benefit would that give you compared to changing fsb in bios? (Other than the convenience, of course). In the past I had the need to use software to control fsb (some mbs do not handle high fsb well at start up) but I have not used one for a long time. I am wonering if I missed some tricks here.

I ran prime 95 for years but never got warning messages (not even once). Do you have to enable it somewhere? I have gooten enoguh weird messages from prime 95 but any error would always stop p95 right away.

Thanks. Any update on your test?

Out to do some real work (clean up the yard) to make the wife happy.

 

[myocardia]

Well, unless you actually like unhooking everything from your computer, so you can take it to whatever carpet-free place that you work on it at, then I can't imagine anyone not using ClockGen to overclock with. There are many other advantages to using it also: it won't be overclocked (or not nearly as much) when you boot to Windows, so you can defrag without worrying, you can install software not overclocked or not very overclocked, and bigger than everything else besides it taking 3 days to find out how far your chip will OC, there is a 0% of having to wipe your hard-drive and reinstall Windows because you corrupted Windows by OCing. And yes, I did actually stop Prime95, and here is the screenshot: link to screenshot. I'm happy that it's stable now!

edit: I forgot about the Prime95 warning messages. I don't remember ever getting them with the 2.2 version, but both the newest version and the one before it give warning messages if something goes wrong that wasn't enough to make the program stop.

 

[kloostec]

I'm running a mobile 2500+ at 14x166 - 2333 MHz (maximum fsb 166 due to ddr333 ram... but it's cas2 ). I had to bump the voltage to 1.625V. It's in a Soltek Qbic EQ3702 SFF case, and runs at 60C at full load (burnk7). I had it running at 15x166 (2500 MHz), but I couldn't get it stable without bumping up the voltage a bunch, which raised the temps too much.

 

[CarpeDiem99]

Hmm, could be true, but I'm happy with my overclock right now, 250x10.5 2630 mhz 1.85v load
http://service.futuremark.com/compare?2k1=7720657
all air cooled,
Carpe

 

[Sniper82]

whats the difference in the mobile XP's and normal XP's? Do they just o/c better or something?

 

[fireontheway]

lower stock V = better OC

 

[Tango57]

the mobiles also have unlocked multipliers unlike their desktop counterparts.