Athlon 64 3500+ overclocking results and help

[phrowzen]

Hey,

Just got my a64 3500+ 90nm winchester yesterday to pair with my msi k8n neo2 platinum. Today I have it at 240x11 (2620MHz or so), and its been priming on blend mode in prime95 for a little over 4 hours, no problems. 230x11 ran prime95 blend for over 7 hours, no probs. Right now I am doing 240x11 in prime95 on small fft mode, overnight.

This is all on stock voltage with a thermalright xp-90 and tornado 92mm. Does anyone have any idea what the average oc is for a 3500+ winchester? I want to know possibly how much more ocing headroom I've got, on stock voltage at least.

Thanks

 

[Megamixman]

It all really depends on your chip. Some are good and can OC a lot. Some can?t budge 100 MHz. As long as your ram can take the extra FSB, just keep upping it. When you can no longer up the FSB to get higher speeds, you could run it with a divider.

What I would recommend is get it to the higher Speed you can till you can?t post. No priming. Once you can't post. Reset your Cmos and bring it down a little and try to prime that. If that doesn?t work just keeping going down little by little till your stable.

 

[Turin99]

most chips that I've seen people post on the boards are hitting around 2.6
A few people say they can get higher, and statistically speaking someone has to be able to, but I would think the mean would be around 2.6

 

[Duvie]

I think the mean is a bit less then that but each passing week 2.6's are more and more common and 2.7's are being reach with a bit more frequency....

Figure in the fact that 50% are either embellishing or do not have the same standrad as stability as maybe you or I......

I post my prime95 with mine as well as my superpi scores....A system can be stable without those but since I do some DC I want to be sure I am not wasting my time or others....

gamers need to obvioulsy do rigorous testing to include graphices....

 

[phrowzen]

Hey,

Thanks for the responses guys. Megamixman -- I tried as you said. I attempted to post at 250 htt, which seemed successful, it showed various BIOS messages, but then stalled with a cursor in top left corner.

Right now im at 246 htt = 2706MHz, and prime95'ing with small fft.

Im hoping that works out, ill let it go overnight, and then maybe ill bump the htt up 1 or 2. Ram at this point, 2*256 corsair xms3500 c2 pt, is being ran on 166 setting (193 i believe with respect to increased htt), 2-3-3-7, which I know is quite agressive as I am only testing the cpu at this point, but bleh -- its already priming, if it fails, i know what to check first .

1GB OCZ is coming next month I believe, im thinking OCZ 1GB PC4200 enhanced latency -- any thoughts?

Thanks,
phrowzen

EDIT: Well, it doesnt appear it likes 246htt all too much -- I tried again with relaxed timings to 2.5-4-4-8 -- same deal. I can almost tell when its going to fail, as if I play an mp3, i can notice some popping/ticking in the audio stream, which seems thus far only to be present when the overclock is failing.

Back to 240htt and testing overnight @ small ftt, I guess 240htt is still nothing to complain about. Ahh well.

 

[ts3433]

You don't need to get expensive RAM to overclock an A64. Save some money and get 2x512MB of value RAM from Corsair, Crucial, or Mushkin for $140. Link

By the way, to get over 240HTT, try lowering your HT multiplier to 3x, and moving any SATA drives to known locked ports.

 

[phrowzen]

I tried as you said ts3433, dropping to 3x HTT multiplier and going back up once again to 246 HTT. As I was running prime95 and writing my anandtech reply, the machine reset . This was with ram @ 2.5-4-4-8 at about DDR386.

Welp, I suppose ive found the range of my overclock then, without voltage that is.

Im going to prime this @ 240HTT/4x HTT multiplier overnight just to make sure it is in fact completely stable at this.

EDIT: Welp I don't know what I've done, but prime95 is no longer stable for 240HTT. I won't say prime95 isn't, as it doesnt seem to report error on screen, but I get a winxp blue screen, "MACHINE_EXCEPTION_CHECK" and then it reboots. The testing I had done earlier was on blend mode, but the testing im doing now is small fft, i wonder if blend wasnt revealing the problems...

 

[Turin99]

I would stick with prime fft just to test your cpu. I wouldn't mess with any other system settings that might destabilize the system. I suggest you start with a low multiplier and a safe oc of like 2.0 or 2.2 or even stick with 1.8 and just test dif htt settings to see what speeds are stabile and you can use memtest to find what memory speeds are stabile. Then when you know the rest of yoru system is stabile, you can use the fft test to test the processor alone and that should be your final seal of approval.

 

[Duvie]

It needs more vcore...This board undervolt about .06-.07v...You likley are only giving it 1.34v if set to 1.4v in the bios....BUmp it up to 1.475v and it should give you about 1.4-1.41v actual.....Then rerun...

small fft is the 2nd most for heat produing to the large fft but places a lot of stress on the cache of the chip...i dont like the blended as it test ram...2 things...It doesn't hold high temps like the others and therefore does not stress enough, and doesn't isolate the cpu. I would set the memory to like 166 ro 133 with cas 2.5,4,4,11 setting for know with 2.7v....Use memtest to test memory....

 

[phrowzen]

Hey,

Looking in CPU-Z, my voltage is usuallly 1.41V-1.44V. However, i still chose the over vid 3.3% option, which is putting the cpu at about 1.472-1.49V.

Priming now, I just wish I knew why sometimes my sound will begin to crack. As I said earlier, this usually leads to a machine lockup/bluescreen/crash.

There is no PCI lock option I see in the BIOS, i am just assuming it is always 33MHz?
I saw a section for AGP frequency, which I set to 67MHz instead of 66MHz just to make sure.

phrowzen

 

[Duvie]

That is good, though I think newer bioses have taken care of that....

What you want to watch for is make sure you are using SATA ports 3 and 4 as they are the only ones with working locks....

Also slots 1 and 2 in all of my experience watching ppl here and elsewhere are better then using slots 3-4....

Restes will usually be heat or power..since you say BSOD it means windows is not auto restarting which is good....

 

[phrowzen]

Hey,

Yes I learned of this SATA 1/2 problem a little earlier yesterday, which gave me a brick wall as I jumped from 230HTT to 240HTT. I moved over to SATA 3/4 and I could easily push my 230HTT marker.

Anyways, yes about my BSOD, Im really not sure it has to do with the overclocking potential of the CPU itself. I honestly think it can easily do the speeds im putting it at, prime95 does not complain at all, and temps are about 49-52C for the CPU according to core center or MBM5.

As I said in last post, im priming small fft 240HTT @ stock cpu voltage chosen, but then hit the over vid by 3.3%, to bump it up a bit. So far its good and I havent noticed any clicking or popping in my tunes, so maybe its ok? I also upgraded the BIOS to latest before I did this as well, Im doubting that has had any impact though. If this primes ok, maybe ill remove the extra over vid 3.3%, I possibly just had some setting chose wrong earlier (im hoping).

Ram at this point also is a 2.5-4-4-11 @ DDR311 -- so plenty room there

EDIT: Also as I said earlier im a pretty big noob to a64 overclocking, I dont like pushing the voltage on my cpus very far, is under 1.5V considered a safe zone for me for a 3500+ 90nm winchester?

Thanks

 

[Duvie]

10% over stock 1.4 fro 1.54v is as high as you should go..I have never seen 10% kil a chip.....Those are actual vcore not set since 1.55v set will likely only be like 1.47v actual...

What you will notice is that the neo does not recognize the winchester right and at default it is setting 1.5v...you can verify this by switching manually to 1.5v and you will get same 1.41-144v you were see as stock bios setting...

Therefore you have 1.5v + 3.3% which is actually setting 1.55v but since the board undervolts you likely get 1.47-1.49v.....You probably would not want to go over 1.5v + 5.0% which shuld give an actual of 1.52-1.54v....

 

[phrowzen]

Ok great.

So I am hovering at about 1.47-1.49V now, peaking sometimes at 1.5V -- so I guess I am good As well, I see you are running 1.57V, so thats reassuring as well

I just got back from school and I see it is still priming, so maybe once more ill try and bump it to 2.7

phrowzen

EDIT: Last, once I can get my machine stable (hopefully at 246HTT) -- what process should I do to acchieve the greatest bandwidth out of my system? Should I be dropping the cpu multiplier and increasing the HTT further?

 

[LouPoir]

Mine is running at 11x230. Same exact set up.

Lou

 

[Sparky19692]

About all I can get and remain Prime stable is 237 X11 2.607 that may be somewhat do to keeping mem timings tight 2-3-2-5. With lose timing I can remain stable in 3dmark03 looping 5 to 10 times (usually all th longer I run it) up to 2.7 but I cant pass prime for more that 30 min. I have read all the posts on memtimings they are right there is little gain by remaining tight or 1:1 however at the same time there is little gain by O/C 4 percent higer and loose timings. For benchmarking I like to O/C just to see what it will do. for day to day running I keep it at stock 2.200 with my voltage right downI can pass prime and anything else at 1.25 vdc. Meaning case fans are off cpu fan is at 11/16 speed (no I'm not using cool and quiet why would you?) and still max temp underload is 37c 30c when doing normal tasks.

 

[phrowzen]

hmm, 2.7GHz 246*11 and cpu overvid 3.3% just failed after 12 hours 47 minutes .. back down to 2.6

 

[Icepick]

I've got my 3500+ running Prime95 stable at 235*11 (2585). CPU Voltage: 1.525 +3.3%, DRAM voltage: default, DRAM timings: 2-3-2-10 1T, 166.

No matter what I do I can't seem to break the 235HTT barrier. At 236 with 1.55 & +5.5% it fails Prime95 within 30 minutes or less. 3DMark05 runs fine at that speed but I can't get Prime95 to run more than 30 minutes. Even with loose timings 3-4-4-10 2T.

Does anyone know if bumping the chipset voltage would help?

 

[phrowzen]

icepik you sure you have your samsung 160gb SATA on SATA 3/4?

Anyways,

I'm still priming 240HTT at 4x LDT for 2640MHz at default voltage. It's been going about 21 hours or so now, so looks promising. But again, my goal was 2.7GHz 246HTT. It had primed at 246HTT with 3.3% cpu over vid but failed after 12 hours 47 minutes, thats pretty long. Does anyone have any tips of what might help me squeeze an extra 60MHz out of this pig? I wouldnt like to increase voltage any further as I was recommended above not to go over 1.5v + 5.0% (actual 1.52-1.54V due to undervolting for winchester) -- but may if this is the last option, and hopefully I get some feedback from others that this voltage will infact do no harm.

Thanks,
phrowzen

 

[Shimmishim]

I'd say 1.6 is safe.

have you tried lowering the LDT to 3x?

your chip looks promising though.

 

[phrowzen]

Hmm nope ive only ran it at LDT 4x, i thought that would be sufficient, I guess its worth a shot though. Would the decrease in bus bandwidth be worth that 60MHz though?

EDIT: I have primed small fft for 31 hours on LDT 4x HTT 240 @ completely stock voltage. No problems. Wishful thinking I put it to LDT 3x as you said and no cpu overvid, and primed small fft for about ten minutes. I did not have the prime95 window open but all of a sudden a blue screen of death and instant reboot. I guess it likes the extra 3.3% over vid I had given to it earlier when I had 246HTT running LDT 4x that had lasted for 12 hrs 47 min.