I've been gradually increasing the FSB on my P4 2.66, and I think I may have hit the limit on what the CPU will run reliably at. Presently, I have the FSB set to 145, and the system runs great at 2.91, no problems. I raised the FSB to 150, which should have given me 3G. On rebooting the system, it blue screened, so I upped Vcore from 1.525 to 1.550, and rebooted once more. No problems loading XP, and it ran the Sandra tests with no problems. Once I loaded up Seti, to see what the temps were when under full load for a time, it ran for about 10 minutes, and then blue screened out once more. I've set the FSB back to 145, and decreased Vcore back to 1.525, and all is well once more.
I don't think the problem is temp related, as with both speeds, the idle stayed at 40C, and under full load, it runs at 55C. My memory is DDR400, and using the 150 FSB should have had it clocking at 375, which Sandra confirmed. The memory timing is set to 2.5-3-3-7, btw. Any ideas, or should I just be happy with 2.91 and leave things as they are?
The limit of most B-steppings is around the 3GHz area. You might have hit the wall at 2.9 with your 2.66, but most people are getting 3.0-3.1. Try upping the vcore a bit (if you can) and pay close attention to the motherboard power regulators. Check their temperature. Do not go over 1.65v.
If you can, try to sell your 2.66 and obtain a 2.4C. You're not obtaining the best performance possible on that chipset with a B core P4. And most 2.4Cs will reach 3.2-3.5 anyway, with much more bandwidth to boot. :t
Belgarius
09-14-2003, 01:24 PM
I've been running MBM5 alongside my tests, and have been following the system voltages as well as the temps. When running Seti to keep the CPU pegged, I've noticed that the PWM temps (those associated with the mobo regulators, per Abit) are running at about 48C, and the processor levels off at 55C. The system temp stays at 32C. One thing I had noticed that struck me was that during the full load, Vcore drops a bit, from it's present setting at 1.52 (MBM only shows 2 places past the decimal - it's set to 1.525 in the BIOS) to 1.50. I would imagine this is due to the load on the processor, and reflects normal operation. Basic Ohm's law, current goes up, resistance stays constant, voltage drops. While attempting the tests at 150FSB, the temps stayed about the same, although I didn't note the Vcore drop, didn't think to look at it. I raised the voltage from 1.525 to 1.550 in the BIOS, which allowed the system to boot with no problems, and run Sandra, but it crashed out running Seti. Since I am not in a hurry to fry my processor, I figured the safe bet would be to back things down to the last stable settings, and leave it be for a time.
I plan eventually to replace the CPU with at least a 2.8/800, but in the meantime, would like to tweak it up to the 3G mark. I've seen a good many using a similar setup to mine that have attained that with no problems. I may attempt setting the Vcore up another notch from 1.55, and giving that a go here later in the day.
One thing that has me a bit confused here. You mention B stepping, whereas Sandra shows my CPU having a C1 stepping mask. Am I overlooking something as usual, or is this a different stepping?:confused:
Thanks much for the info,
Belgarius
tasty danish
09-14-2003, 07:34 PM
im not sure bout your stepping, but that 2.8 will never obtain an overclock like a 2.4c
Belgarius
09-14-2003, 10:13 PM
:confused:
Hmmm... I must be missing something here, I was thinking that the 2.4c and the 2.8 were of the same family? If not, I'll surely not shell out the extra ducks for the 2.8! I originally intended to get the 2.4c, but unfortunately, it's not available retail in this lousy burgh I reside in. I opted for the 2.66 over the 2.4b, which was available, in the hopes the additional speed might prove a bit more overclockable in the long run. Seems that my logic must have been flawed. Eventually, I intend to replace the 2.66 with a CPU that supports the 800FSB and HTT, so would I be better off going for the 2.4, or perhaps one of the 3X flavor? I really don't want to set the world ablaze, just want to get to about 3.2 or a bit better without squandering my kid's inheritance in the process... :D
Belgarius
tasty danish
09-14-2003, 10:29 PM
well, all those chips are of the same "family" its just that between a 2.4 and a 2.8 at the same speed: the 2.4 will perform better because of a higher fsb. also, all those chips tend to top out (with stock voltage and cooling) at about 3.5ghz, at this speed the 2.4 would have a fsb of 290 and perform way better than a 2.8 at this speed with a fsb of 250. plus the 2.4c is really cheap for an intel: about 160usd online. if you have the memory to handle it (ddr460 with a fsb of 290) then there is no beating the 2.4c.
Belgarius
09-15-2003, 08:19 PM
Ah, now I see where the confusion may have started. The 2.8 I have in mind is the 2.8C. With that in mind, will the 2.4C still outperform it?
Belgarius
tasty danish
09-15-2003, 09:39 PM
i thought that it was a given that all the chips we are talking about are "c" family. i dont know hardly anything bout the b series.
Belgarius
09-15-2003, 10:40 PM
Okay... my bad, I suppose. My present processor, the 2.66, is what I presume to be a "B", as it has the 533FSB, and does not support HTT. On looking back over the postings, I see where I may have inadvertently started something unintentional, where it was mentioned about stepping masks... Mine shows as a C1 in Sandra, and Caustic Vapor had mentioned something about B stepping. I was a bit confused, and probably still am, about this aspect of things. As far as things go presently, my CPU shows C1 stepping, but has the 533FSB, and does NOT support HTT, regardless of what WCPUID led me to think. (standard flags indicate HTT support, which the processor doesn't have, nor does the BIOS indicate. From this, I think it's either a glitch in WCPUID, or an indication that the mobo supports it, which it does...)
I'm still a bit iffy about the 2.4C, though. If it will allow a higher and more stable overclock than the 2.8C, why would anyone in their right mind shell out the additional costs? As you mention, the 2.4 can be had for substantially less than the 2.8, and both support the 800FSB, as well as HTT... This bears a bit more looking into, methinks. If nothing else, I am known for my ability to overlook the forest for the trees at times.:D
Belgarius
tasty danish
09-15-2003, 10:53 PM
because most people are not do it your self-ers and do not want to overclock because intel says it's dangerous. that is why the would shell out the doe for a 2.8 because at stock, it whips the 2.4's ****.
Belgarius
09-15-2003, 11:09 PM
Sad, but true... I am by no means an expert in the field, but with a bit of digging and research here and there, I have been able to take my paltry 2.66 to a bit better than 2.9, and even without the 800FSB, or the HTT, that's a definite improvement over the "stock" specs that Intel posts for the CPU. I've been in hardware for a long number of years, literally since the days of the XT and it's predecessors, and have never really been long on taking things at face value. If the 2.4 can be taken to the extremes that I've seen, I only wonder what the 2.8 or higher versions are capable of. Hmmm... perhaps instead of that Vantec, I might opt for a water or vapor cooled system, and really make my power company happy.:cool:
Belgarius
tasty danish
09-15-2003, 11:39 PM
with an extreme watercooling or peltier system, you may be able to get really really high fsb's with the 2.8, but as for affordable, aircooled applications, 3.5ghz is bout as insane as it gets
causticVapor
09-16-2003, 01:03 PM
The only advantage the 2.8C has over the 2.4C is the higher starting multiplier. This means, for instance, to reach 3.6GHz, you're not gambling as much. A 2.4C requires a 300MHz FSB to reach 3600MHz, while a 2.8C only requires a 257MHz FSB. This has its advantages, as you're not pushing the northbridge as much. Not all 865/875 northbridges will reach 300MHz - some will, some won't, and some will need extra cooling to get there.
However, if you're content with 3.4GHz, then the 2.4C is the way to go. Most of the cores tend to max out at 3.4-3.6GHz anyway, so I don't think you'll be going on a shooting rampage if you achieve 200MHz less.
Oh, about the B - wrong B. Sorry. There are A P4's, B, and C p4,s according to BUS. The A's run at 400, the B's at 533, and the C's at 800. A-bussed P4's can be either willamettes or northwoods; B's and C's are Northwoods.
As far as STEPPINGs, the A is the first (and can do about 2.4GHz); the B is better (~2.53-2.8 for the later models) and the C1 is the stepping for the 3.06. Most C1 cores top out at 3200MHz, tops; the average for 2.4 C1s is 2.8-3.0GHz. Some really late C1s were lucky and could get up to 3500MHz. Overclocking a 2.26GHz to 3500MHz was more exciting than riding in a derailed rollercoaster car.
Then came the D1 steppings. All C-bus P4's are D1 steppings, and top out at 3.8GHz on air, tops. Most achieve 3.4-3.6, as I have said earlier. (Overclocking a 2.4C to 3.8 on the stock cooler is a lot of fun, too.)
BTW - 533MHz FSB P4's will move to the D1 stepping later this year. This will prove interesting to overclockers, as a 2.4B on the D1 stepping could get to 3600MHz by simply raising the FSB to a C-bus P4's stock (200MHz). This would allow for a really fast processor combined with relatively inexpensive 5ns memory.
Hope that cleared it up for you.
Belgarius
09-17-2003, 09:45 PM
Many thanks for the info. If I am seeing things correctly, my 2.66 has the same stepping as the 3.06, albeit without the HTT ability. With a bit of luck, and a case that doesn't run hot enough to fry eggs on, I may actually achieve my 3.0 to 3.2 goal, with a bit of adjustment to the Vcore, and perhaps bumping the DDR voltage as well... Time will tell.
Again, my appreciation for the insights!
Belgarius
SysOpt.com
Copyright Internet.com Inc. All Rights Reserved.