AMD, Vcore and aircooling

[felix88]

i know this question has been asked before, and i'm pretty sure i already know the answer. i guess i just need to hear it again.

i just picked up a new XP1700, to get it stable at 2.3GHz i need 2.1v. i've got a PAL8045 with a 45-50 CFM fan on it, and my temps are respectable(52c in the winbond hardware monitor).

needless to say, i'm dissappointed with this chip, i was expecting much better. :mad:

[Swiftdeathz]

Originally posted by felix88
i know this question has been asked before, and i'm pretty sure i already know the answer. i guess i just need to hear it again.

i just picked up a new XP1700, to get it stable at 2.3GHz i need 2.1v. i've got a PAL8045 with a 45-50 CFM fan on it, and my temps are respectable(52c in the winbond hardware monitor).

needless to say, i'm dissappointed with this chip, i was expecting much better. :mad:

eeeeek 2.1v... :eek:

On prometia or vapochill that's maybe ok but on air I don't think the chip will last long at all...

[felix88]

yeah, thats what i figured.

i'm just pissed that this chip won't do more than 2.1GHz with a respectable amount of voltage. maybe i'm trying to kill it... :hehe:

[felix88]

trying it again at 1.95v, and it seems stable but it won't do 3dmark.

[Lunatic]

burning it in might help.

[luihed]

Looks like you picked up a DUT3C cpu which requires alot of voltage to get maximium mhz.... my 2100+ needs 2.22v (2.17v in bios) to reach 2550mhz and its aircooled.... seems like these DUT3C loves high vcore :D

[felix88]

no, it's a DLT3C. just a crappy one.

maybe i'll try selling it and i'll pick up a XP2500.

[GVCryan]

I have a DU that runs pretty good.

My sig is out of date but it runs around 2.3 at 1.8v, and still under 45c.

2.1v isn't that bad, it doesn't give you enough room to play on air, but it won't light your house on fire and kill your family.

I ran an Aroia 1600 pally at 2.4v on air for several months. Still works, just in an office machine.

[SpiderMAN]

Originally posted by felix88
i know this question has been asked before, and i'm pretty sure i already know the answer. i guess i just need to hear it again.

i just picked up a new XP1700, to get it stable at 2.3GHz i need 2.1v. i've got a PAL8045 with a 45-50 CFM fan on it, and my temps are respectable(52c in the winbond hardware monitor).

needless to say, i'm dissappointed with this chip, i was expecting much better. :mad: Too high VCore for AC :)
Don't be stupid :)
My JIUHB runs stable @ 2144 MHz @ 1,75V
DUT3C - default VCore 1,6V

[Liquid3D]

Originally posted by luihed
Looks like you picked up a DUT3C cpu which requires alot of voltage to get maximium mhz.... my 2100+ needs 2.22v (2.17v in bios) to reach 2550mhz and its aircooled.... seems like these DUT3C loves high vcore :D
It's not necessarily so the DUT3C's require so much voltage. Mine as shown is Sig is running now at 1.75V @ 2443MHz. I agree the 1.5V do seem to overclock higher, and hold better temps.

I have definately noticed a serious increase by running the chip 24hr period, with no need for excessive clock-cycle burn in. Simply running a High FSB inititially, and then backing down just slightly on the FSB, and raising the multiplier in steps, alternating, all the wile maintaining a speed a few hundred MHz shy of the first highest overclock. This has essentially allowed my chip which originally needed 1.8Vcore to run stable at 2400MHz, to now run stable at 2443MHz at 1.75V.

I used to overclock, benchmark, and bring the chip/memory back down to default. I did this to preserve, and prolong the life of the chip. While in actuality I was stressing the chip more with intermittent excessive voltage, and speeds. I cannot stress the importance of "burn in" periods. They should high, but not to the maximum stable, but just under that stress level.

[felix88]

Liquid3D, what do you use to burn in your CPUs? Prime95?

maybe i'll try 2.2GHz, right now i'm at 2.1GHz with 1.85v. or would you say the current speed i'm running would be sufficient?

[Liquid3D]

I think you said the voltage your at currently, is the minimal you can run with stability? What's your FSB right now?

[luihed]

Felix88, sorry for assuming that you have a "DU" chip, Ive never had a "DL" chip but from what ive seen in the forums, they dont even boot past 2v at any speed..... which led me to my assumption

Liquid3d, you have a good chip in your hands there..... ive had a 1700+ and a 2100+ and both of them requires 2v+ to hit its maximum mhz.... I can only do just shy of 2400mhz at 1.85v with my 2100+ so I need ed the extra voltage..........

And Spiderman..... its fun to be stupid at 2v+.....

[felix88]

Originally posted by Liquid3D
I think you said the voltage your at currently, is the minimal you can run with stability? What's your FSB right now?

well i believe 1.85v is the minimum stable voltage. last time i tried 1.825v, i got a prime error. i'll try it again, maybe it failed 1.8v.

[Lunatic]

i wouldnt go over 1.85 for 24/7 use

[felix88]

Liquid3D, i forgot to say, it's at 210MHz FSB at the moment.

i wonder if it would help to put some better cooling on the NB? i thought that only helped very high FSB overclocks.

[Liquid3D]

Most nForce2 chipsets are overclocked past 200FSB where voltage is required for true stability. Here's a link to VRZones 8RDA VDD-mod (voltage to NB), it also mentions proper NB cooling. Since it's easier to add a HSF to the NB then it is to mod the circuits, perhaps that is what you should try firts. I'd be willing to bet my infamous "solder dropping collection" this will give you another 10MHz. It really depends upon the motherboard but the nForce2 chipset tends to run very hot, and in most cases 200 FSB is the maximum sttainable speeds. In fact I'm shipping out my 8RDA tommorrow RMA.

I purchased the Soltek SL75FRN2-L specifically because it offered BIOS adjustable VDD from 1.6VDD to 1.8VDD. ANytime I get over 205FSB I'm using 1.7V, and when surpassing 115FSB, i crank it to 1.8V. The highest I've been able to attain on this board is 229FSB, just as the review stated. It's purported to be the highest attainable FSB without modifications to the NB. It also has an excellent NB cooling solution. You'll notice motherboard makers are placing tiny heast sink, fan combos, instead of tiny little fans which were used on the first graophics cards 6 years ago. Peopla will disagree, but i feel the Epox passive cooling is totally inadequate for such a board, especially the newer versions which have VDD adjust.

In summation, I'd definately place the biggest -"ass" HSF combo you can fit on the NB without crushing it. You should carefully remove the NB heat sink and clean off the goo they put on there, and replace with a thin layer of Thermal Paste. If your running the Epox, once you remove the NB heatsink, you'l see the main reason I label it inadequate. The NB only covers 1/2 the total die space of the nForce2 chip. here's that link;
http://www.vr-zone.com/guides/EPoX/8RDA+/

[felix88]

i've got a NF7-S V2.0 actually, it has VDD from 1.5-1.7v. i believe it's stable at 200MHz with 1.6v, but as soon as i kick in the agressive RAM timings, i need 1.7v.

anyway, do you think my NB is actually holding back my total CPU speed? i thought only FSB affected the NB.

i'm gonna run a few tests with lower FSB.

thanks for your help so far Liquid3D!

[felix88]

ok, i'm priming it at 2.2GHz(200MHz x 11x) with 1.85v. i had to back my ram timings down, but it is 3DMark stable.

hopefully better NB cooling will allow me to have my RAM timings back. or maybe a better PSU? my 12v rail looks like it's hovering around 11.8v, within 5%, but still a bit low.

[Liquid3D]

Try dropping the multiplier to 10x, and shoot for 10x210 @ 1.8V. First up the VDD. I'd also suggest removing the side panel from your case, I'd almost gaurantee an overall system temp drop of 10C. If you have a few minutes, try the following experiment. Just for Sh_ts and giggles once you've removed the side panel, place a floor fan about two feet from the case blowing into the open side. I'd bet my "solder dropping" collection you'll see 15C drop acroos the board (pardon the pun). And will attain higher overclocks across the board as well. Seems like your memory may be holding you back as well as the NB, and CPU. However, there's one universal element which transcends a;ll others, and either limits, and allows increase, and that is thermal dissipation. More heat which can be dissipated = MHz. people don't spend $600 - $900 for a Prometeia for not.

You have an excellent mobo there, and I believe a potentially decent overclocker. Actually there';s no potential about it. I beleive we become overly compacent with the overclockability of TBred-B's. I'm not implying your obtuse, only weve come to expect 2300MHz to 2500MHz from 1433MHz CPU's. As overclockers we take risks, and beyond the prima facie frying of a CPU, we forget there are other risks as well. That the CPU we purchase will not meet any expectations. Which is why I try to think of it, as icing on the proverbial silicon. Still, I feel ya my brother from another mother. try to remember you've already attianed a 700MHz OC, and if we all attained these speeds, pretty soon ALL CPU's would be locked, FSB and Mulitpliers. Definately try the fan experiment. Just for comparitive reasons. So I think you don't really have to worry. Exactly what HSF are you using , memory, and PSU?

[felix88]

Originally posted by Liquid3D
I'm not implying your obtuse, only weve come to expect 2300MHz to 2500MHz from 1433MHz CPU's.

i bought this CPU used from a guy who claimed it would hit 2.3GHz on a Volcano 7+ with "about 1.8v". this is the only reason why i'm expecting so much out of it.

oh and it's 1466MHz. :p ;)

Exactly what HSF are you using , memory, and PSU?

i'm using an Alpha PAL8045 with a 47CFM Panaflo, 512MB Twinmos PC3200 with Winbond CH-5 chips, and an older Enermax 431W PSU.

i'm working on a deal with a guy currently, for a Vantec 520w PSU. i'd rather have a PCP&C, but thats a bit out of my price range at the moment.

oh, and i need 1.85V to hit 10 x 210MHz. i can't go any higher unless i drop the FSB and run really loose timings.

[Liquid3D]

I'd recommend passing on the Vantec, and trying the TTGI 450 4fan silent PSU, from Directron for $62; http://www.directron.com/sf450ts.html Trust me, post a topic about TTGI 520W (which is $89) or this one, and you'll find of the best overclockers on this forum nearly 70% either own, or will recommend TTGI regardless of their being less expensive then most. The four top PSU's on the market currently are; PCPower&Cooling (I have the Silencer 400W), Antec True Power 550, TTGI 520-TT, and Sam Cheer 475W. I'd almost gaurantee if you were to post about the TTGI 520-TT 520Watt it be the most recommended, among the top overclockers here!

I didn't really mean to be critical towards you, and I do see why you would expect the CPU to do 2300MHz given those circumstances. My statement was not passing judgement towards you, but to "ease" the pain. However, given the fact you were sold this used, and it was claimed you would attain 2300MHz, I'd say don't buy anything else from thast person. It just sound like they must have known immediately it required 2.1Vcore to hit 2300MHz, and even if they had a Prometia, It would have required 1.9V to reach 2300MHz/2400MHz.

I'm sorry I forgot your running the Alpha PAL, that's one of the top three heat sinks, continually juxtiposing between 1st, and 3rd in most HSF round-ups. Jeez, at this point, I'd recommend first buying a Vantec Tornado 80mm/84CFM fan (since you plan on one anyway for your side window, if the Alpha can take a 80mm fan size) prior to ordering the Thermalright (to save money) and see if that's the gist of the problem. It usually is. Whether you have the MCX4000 or SLK900U, these are FAN dependent!

And with memory, this is why I recommend, and stand behind brands like OCZ, while Twnimoss, Kingston, A-Data, Mushkin, Crucial, etc. will come across great steppings periodically, OCZ works to consistantly be the best, and their life-time warranty doesn't hurt.

BTW although loose timings, and low Muliplier, wont yeild any significant bandwidth, keep dropping the multi, and loosen the timings just to see where your FSB limit is! In this way, you can at least determine where the that limit is. It's also very important (perhaps more important then the burn-in process on other devices) to burn in the NB nForce2. ONly in this 2way, will you be able to readily acheive 230FSB and above, otherwise it's just another ceiling you'll hit regardless of replacing memory, CPU, and cooling. So long as your burning in something, your accomplishing an important function towards higher OC's.

[felix88]

once again, thanks for all the help Liquid3D.

i guess i'll stay away from the Vantec. i was looking at it because i could get it without spending any money, and because it is probably better than my current PSU. i have heard lots of good stuff about the TTGI units.

i might look into a higher CFM fan, but i don't know if i'd go with a Tornado. might be a bit too noisy for me.

i just lowered my timings to 2-3-2-7, and it's still stable at 2.6V. VDD is still at 1.6V(i believe this is stock on the NF7-S V2.0). and suprisingly, my CPU now does 2.2GHz with only 1.8v. i haven't tried lower yet.

maybe i had my voltages too high? or maybe i finally burned something in?

if i have a little time tonight, i'll shoot for a higher FSB, otherwise i'll have to do it tomorrow.

[Liquid3D]

Sorry i thought you had adjustable VDD. SO either cooling, and/or a VDD mod, are your only alternatives to exploiting those limits your FSB can ultimately acheive. Still it's safer to juxtipose the higher voltage required to maintain a high clock rate (CPU speed), for dropping the multiplier, and working on increasing the FSB limit.

I glad to hear it's burning in, and I think it will continue to improve. As we (you) try different solutions, what we have to remember is, whatever changes are made they should be given time. Only time will allow the physical (electronic) resistance along the voltage pathways to undergo the metamorphosis needed. As the sub-atomic particles along said pathways become more malluable, burn-in occurs. Forgive my terminology (descriptive) for the specifics, however; I'm going to set aside several hours today to research the physical changes which occur along the voltage "gate lengths" within the silicon, and how these are affected during the "burn-in" process.

Speaking of computers, I have to go re-set the "Dosing" PC at the Methadone Clinic. I volunteer my time to do this every day at 5:50AM. I may be disabled, unable to work, but it's essential to keep myself (mind/body (problem)) occupied. Otherwise I'd be jumping off the Newport bridge with my Monitor tied to my leg. :p
Be back in an hour.

[felix88]

ok, cool.

i actually do have adjustable VDD, my choice of words earlier just sucked. i meant to say that the NF7-S V2.0 comes with the VDD set at 1.6v.

anyway, i tried 215MHz FSB and i keep getting a prime error, regardless of the timings. i'll give it some more time tomorrow, but for now i've got to get some sleep.

funny, i'm going to sleep about the same time that you are waking up. :hehe:

[felix88]

after priming all night long at 11 x 200MHz at 2-3-2-7, i decided i'd try 10.5 x 210MHz. suprisingly(to me at least), it was stable at the same voltages(CPU: 1.8v, VDimm: 2.6v, VDD: 1.6v) and timings that 11 x 200MHz was stable. previously i wasn't able to get the RAM/board to take anything above 200 with decent timings without upping the VDimm to 2.7v and VDD to 1.7v.

so 2.2GHz might be the sweet spot for this CPU, i'll just have to find the highest FSB.

Liquid3D, do you run Prime95 to test for stability? or maybe memtest86? it seems Prime reveals instabilities in the memory just as easily as memtest.

[OPTPrime]

Well, at 2.1 what type of temperatures are you getting? 2.1 for benching is fine...

[felix88]

i was getting temps in the low to mid 50s.

i think if i keep burning in, i might be able to hit 2.3GHz.

[Liquid3D]

I haven't tried prime 95, I usually perform a Burn-in based on Drempel's screensaver (which actually uses some cycles, and definately works te AGP bus, and card.) I also run Folding @ Home around 50% concurrently.

Very good! I actually think (and will again bet my Solder Dropping collection) your sweet spot will be 2.3V at 1.75V in one week, if you continue to burn in at this rate. What i would do now, is up the FSB just 2-3MHz, and/or lower the Vcore one increment. Continue to run whatever your using. If she's not stable raise the vcore to where she is, but focus on keeping the Vcore at or under 1.8V. Continue to push the FSB 2-3MHz daily (without a Vcore increase). When you reach another 10MHz, back it down and raise the internal clock (multiplier) one increment.

In other words if it's at 10.5x210MHz, try 10.5x213MHz, 24hrs later 10.5x216MHz, etc until 220MHz, then drop the FSB raise the multi again. I think 2.3GHz willbe sweet, but 2.4GHz will be stable as well by next week.

[felix88]

well, i may have hit some sort of a wall here. i had it 100% stable at 210 x 10.5 until yesterday, and then it started locking up. now the best i can do is 200 x 10.5, and i have to up the Vdimm to 2.7v and the VDD to 1.7v. previously it was at 2.6v and 1.6v, but with the same timings and 10MHz faster.

i have no idea what happened. :(