I think we need to sticky this thread.
Printable View
I think we need to sticky this thread.
I think that we've to send the link directly to AMD to see what's going on :D ;)
OK you impatient types. ;)
I had not luck at 9 x 270mhz so I gave up on that. I know this memory will do it but I decided to see what the CPU wil do. Only tried at 1t.
9 x 260htt seems to prime blend fine. Of course only overnight will tell.
I am working on 260mhz x 10 to see if 2600mhz will run. I am upping the voltage slowly to see what happens. So far P95 burps at 3-5 min. Nope 2600mhz is a bust.
Hmm just had something strange happen. Running the 1.41b1 BIOS I just set the CPU volts to 13%, 15%, etc and when I rebbot it defaults back to auto or 1.42v. Anyone else find that? Volts seem to be best at about 1.53 anyway, so maybe it's a 90nm limitation in the BIOS.
Well the latest is 10x 257 and it seems to be priming along. I'll let you know if it's stable after a while.
I am working on 260mhz x 10 to see if 2600mhz will run. I am upping the voltage slowly to see what happens. So far P95 burps at 3-5 min. Nope 2600mhz is a bust.
Yeah mine won't do 260 x 10 regardless of voltage - it will post but isn't stable.
Hmm just had something strange happen. Running the 1.41b1 BIOS I just set the CPU volts to 13%, 15%, etc and when I rebbot it defaults back to auto or 1.42v. Anyone else find that? Volts seem to be best at about 1.53 anyway, so maybe it's a 90nm limitation in the BIOS.
The beta 1.4 only goes up to 10% additional. There are modded ones floating around that allow 1.4 to go 13%, 15% but it's just cosmetic, it doesn't actually work. The only bios that allows more than 1.55 +10% is the 1.3b6. You may be running one of the modded ones, I shy away from those.
Well the latest is 10x 257 and it seems to be priming along. I'll let you know if it's stable after a while
Nothing wrong with that - most I can do is about 2.520, even that isn't 100% prime stable. What voltage are you using in bios to get that?
My guess is that a high fsb on the memory, on the winnies often fails b4 the max CPU score is reached.
I get a idle temp(MBM 5) on the CPU of 34C with 250x10, and a idle temp of 41C with 278x9, with ram 1:1.
also the CPU need a notch more Vcore on the 278x9 setting (1,5V).
running with a divider on the memory reduces the temps on the 278x9 setting.
the cpu is prime stable @ 260x10 1,52V (divider on ram) only tried Large FFT, (user stopped after 8hours)
Rather than starting a new thread, my question is a result of prime not passing and forget about prime for a sec, coz when it doesnt pass my comp is unstable anyway.. random freezes at settings like 10*250 2-2-2-10
So my question is, should i buy a newcastle 3500+, or an FX-53, or FX-55 or stfu and stick with my winnie
I really want something like 10*260 2-2-2-10 as memtest passes at this level
Im leaning toward 3500+ cost its obviously much cheaper
What is the big deal about being ""PRIME STABLE"". My AXP is running at ( and has been running at) 2400-2500 Mhz. At this speed my computer is not PRIME STABLE. I can play any game with no problems at all, I can run all kinds of other benchmarks (CPU stress) with no problem. I can loop 3D01 for days, run super pi and pifast till I am blue in the face---Again no problems. Computer never crashes, No BSOD's, etc..
You know what I do ?
I do not run Prime 95 on this machine.
Nuf said.
Regards
Tritium
You missed the point I'm afraid. Certain date codes of the 90nm Winchesters will not run Prime 95 at 200fsb. That is indicative of a problem and we are trying to troubleshoot that.Quote:
Originally Posted by tritium
This really started with RMA's being requested for OCZ memory that was failing P95. If the CPU won't run it then the memory sure won't.
I am not an advocate of P95. I will say that I defend the rights of someone to have the option of using it to insure a conservatively stable system if they desire that. :)
Yeh i understand what you mean but my system does crash at most prime unstable levelsQuote:
Originally Posted by tritium
I would say that getting a 90nm 3500+ 0441 or later would be a fairly safe bet. This 90nm 3200+ 0441 runs P95 blend just fine so far. They may have mixed up the cores and some weeks 41's might still not run, we can't tell at this point although dnottiss has a week 41 that is failing.Quote:
Originally Posted by Orker
I have a replacement 90nm 3500+ coming and will thoroughly test that when it shows up.
Interesting point. Some further testing along those should be done.Quote:
Originally Posted by v142
I will keenly be waiting to see that results. thx :)Quote:
Originally Posted by andyOCZ
Per my previous post, it also appears that some 3000+ >= 41 also are ok.Quote:
Originally Posted by andyOCZ
Has anyone received concrete conformation that the P95 code is somehow suspect (or not) with 90nm AMD products?
Regards,
Tritium
I have 2 CPU's, both 90nm Winchesters. Week 37 no P95 blend. Week 41 P95 blend runs 11+ hours so far. I used the same board, memory and settings. This is proof that the code is OK in my mind. We have aslo tried several past versions and a beta that has no P4 optimizations. No love on many CPU's.Quote:
Originally Posted by tritium
Andy,Quote:
Originally Posted by andyOCZ
I can confirm that week 37 is a no go with Prime95. I pull my chip out today and it is:
ADA3500DIK4BI
CBBFD 0437WPGW
I can't not run Prime95 at all with the stock setting. Prime95 bombed out at one minute mark with blend test.
I will RMA my chip tomorrow... Looks like I will be spending my Xmas and new years with my old girlfriend.... Pentium 4 @2.4 533FSB!! :mad:
man you need to get out more lol just kiddingQuote:
Originally Posted by p0k
wish i rma'ed my winchester now as mine was a week 37 non primer and all, ah well its moved on now at a financial loss :mad:
Most of us are using the MSI K8N Neo2 Plat..Maybe thats the problem with prime ??
I had missed this. Good to hear another confirmation.Quote:
Originally Posted by phobix
I do want to say that I am not advocating mass RMA action. If you can live without P95 and your computer is otherwise stable you should do fine.
It's not like AMD will recall all these "bad" CPUs - :rolleyes: I've had 4 of these CPUs, one was clearly defective the other 2 sucked for overclocking. Really seems like alot of bad parts are making it to retail.
only because CPUs "suck for overclocking" doesn't make it "bad parts". come on guys, you're just geting ridiculous here :rolleyes:Quote:
Originally Posted by dnottis
I'm talking about the ones that aren't priming @ stock. Out of 4 CPUs I had 1 defective one myself. Just from my first hand experience a 25% failure rate isn't exactly steller. Have a 3000+ 939 .90 coming from Monarch this week - gonna roll the dice one more time ;)Quote:
Originally Posted by bachus_anonym
Ok, after cleaning my chipset off that looked like this -
http://www.3dxtreme.net/images/oc/chipset.jpg
I went back to the basics, 280 x 9 and primed -
http://www.3dxtreme.net/images/oc/11...41.07%20AM.jpg
dnottis
You added some thermal paste and what - it primed fine?
Quote:
Originally Posted by dnottis
Hey guy ! My CPU, Mobo ( chipdate ) are same urs... With Aircool I can Run 300x9 evryday... I think U should use 1.41 Mod BIOS or 1.36Mod ( NF Raid ) to get higher... :toast:
Quote:
Originally Posted by bachus_anonym
Hi
I agree with you, couldnt have said it better myself....
Menty
It did last night :)Quote:
Originally Posted by IvanAndreevich
I think you missed a very important reply...Quote:
Originally Posted by Mentaly
"I'm talking about the ones that aren't priming @ stock. Out of 4 CPUs I had 1 defective one myself. Just from my first hand experience a 25% failure rate isn't exactly steller. "
Of course we arent calling poor overclockers defective, just the ones that wont prime at stock settings.
Quote:
Originally Posted by Zanr Zij
Well you got a better overclocking CPU than me then - I cannot run at 2.6Ghz no matter what voltage I give it. I can run 300 x 8.
Quote:
Originally Posted by dnottis
Hi
I know what you mean, I just had the feeling some of the postings sounded like it was because not all clocked well, I have watercooling on my own chipset, and I must say the little goble of goo didnt look to good on my board either, so I replaced.
Menty
I contacted the AMD tech support and requested for RMA. I just want to let you guys know that if you are going to RMA your processor, it will take months. Since AMD do not have any 90nm in stock. AMD offered me a replacement with the 130nm but I decided to wait.
It is time for me to get a new A64. Is the 3200+ good to overclock?
Did AMD have any thoughts on any of this?
-haha this could get good...
AMD REP "It's Prime95, the program is no good - use OUR stability testing software"
Quote:
Originally Posted by dnottis
You wouldnt happen to know where to get a copy of the AMD stability testing software as it isnt posted on AMD's website or at least I didnt see it.
LOL, it's was a joke! heheQuote:
Originally Posted by edfcmc
ummm :slap:Quote:
Originally Posted by edfcmc
he was being sarcastic... ;)
like: asking a Ford salesman if a Dogde is better... :toast:
First things first.. does anybody has a NON-MSI K8N Neo2 board and a CPU that fails @ STOCK SPEEDS?!
Hey guys,
I would say that our "thoery" is pretty well proven. Under stand that my original testing was done on the Gigabyte K8NS Ultra-939, so any doubt about the MSI K8 K8N Neo2 are untrue.
It seems most that have newer 0441 Winnies are ok with Prime95 at stock and also at decent overclocked speeds. Most earlier 0437 won't prime at 200fsb and certainly won't when overclocked. Again this was tested and confirmed on the GB and MSI boards.
The bottom line is that most guys are stable and happy without P95. This argument has been going on for years. If you are happy without Prime95, then I personally have no problem with that.
I appreciate the great responses and testing that you all added.
This screenie shows my 3200+ 90nm 0441 passing P95 blend at 10 x 255mhz running +5% on the CPU, 1.45 CPU VID, 2.5-3-3-10 1t with OCZ 3200 Platinum R2 at 3.0v. The memory doesn't need that much voltage to run 255mhz but the booster was in there and I just left it alone.
I will check back, but this will be the last of my testing for a while. I have allot to do :)
Read the above post. Yes, a GB K8NS Ultra-939. It was the first board I used for testing. :)Quote:
Originally Posted by IvanAndreevich
Right, so there goes the $hitty board theory :)
You shouldn't test blend, cause as you can see, it is was testing small FFTs which doesn't have anything to do with RAM. As I understand, the CPUs fail @ large FFT sizes (in excess of L1+L2). If you want to test quicker, you run Custom with FFT size 2048-4096 and manually lower the amount of ram a bit so there is no swapping happening. IMO this would allow you to stress test the memory controller a LOT faster than Blend mode which also tests the core.
Well, we were not trying to design tests to stress anything in particular, We are not sure that the issue is the memory controller, so it just became a pass or fail situation after we had reports of P95 blend not running.Quote:
Originally Posted by IvanAndreevich
I like you idea and will give it as try soon. Thanks.
My week37 CPU usually fails between the 2048-4096FFT's. Stock speeds and SPD timings (Some old OCZ PC3500 EL DDR 2x512MB), 2T enabled:
Actually, there were a few people reported to AMD tech support regarding to Prime95. Obviously, they will not make any official statement about it. The software CPU Burn-in is what AMD tech support use for stability testing.Quote:
Originally Posted by dnottis
well now that newegg is dealyed with 3500's whenever they do get them i know i will have atleast week 41 lol
So it seems you are basing that all on this:Quote:
Originally Posted by andyOCZ
andy, give up the crusade! Look, you found a chip that runs Prime stable, which disproves your theory, but instead of saying, "hmm okay maybe I'm wrong" you make up new conditions? Come on...Quote:
Originally Posted by andyOCZ
I heard AMD may have pulled their new 0.09nm chips from stores in Europe.....anyone else heard anything?
you work for ocz lol, cant you email AMD?Quote:
Originally Posted by SteveOCZ
I spoke with AMD tech support in USA, and they said they have received a few complains from the customers about Prime95 related trouble. Obviously, AMD tech support did not make any official statement regarding to this matter while I was on the phone with them.Quote:
Originally Posted by SteveOCZ
Anyway, how are you doing? I saw your post regarding to your accident... I have a Jetta V6 myself and it is a great car. I hope you are getting better. What a terrible accident!!! :eek:
Let me out this in a way that we can all understand. :)Quote:
Originally Posted by WiCKeD
This is no crusade. I am in fact am a huge AMD fan. I work for a company that sells memory. We recently had several RMAs's based on P95 not working at overclocked speeds. Thorough testing by myself and many others has shown that there are issues with P95 and at least week pre-week 41 90nm CPU's.
If a CPU will not run Prime95 at stock speeds it seems logical that it will not run it at 260mhz. We needed to know what was going on and asked for help from our buddies here.
The results form this testing will save my company $$$$, our customers will save time and $$$$ from shipping costs. They are free to use whatever stabilty programs and CPU they like, BUT if their RMA is based on P95 results then we all need to be aware that a problem does between P95 and some CPU's.
I personally do have a perfect 3200+ 90nm week 41 that will run any program I choose for hour on end. Please read the entire thread and follow along to see the purpose and results of all testing.
I suggest that anyone failing Prime95 at overclocked speeds reset their BIOS to default and run the test again for a minimum of 7 hours. If you pass at stock speeds you should go to go, Test away.
Thanks
Ok guys, fair enough, i think that selling the whole system will take much less time and buy something that will run at stock speeds :).
Anyways just for your interest, my cpu controller maxes out at 2.46G no matter what i do, i've primed for lots of hours with large fft's without any errors. But even the slightest fsb more than 246 it will fail within couple of minutes.
Before i hooked a northbridge waterblock on at 9x270 (2.43G) i had some stability problems while was running prime... such as monitor was freaking out with no reason and then it was comming back to life, also prime was failing after few minutes.
Now with the waterblock on using artic silver ceramique i run prime for hours just fine at that speed!
I've noticed when i removed the stock northbridge heatsink that there was not sufficient ammount of thermal paste between chip/heatsink. In some way the thermal paste was very sticky which was causing no good contact.
Conclusions are yours...
P.S.: with the waterblock on the northbridge the cpu max speed didn't changed, just stability at high fsb's.
Can anyone confirm if a non Nvidia based chipset with a 90nm 939 has failed prime at stock speeds? This would further reinforce the theory in this thread that the CPU is the culprit.
I think it's a very good idea about testing this on a K8T800 PRO board with a winch. I have a AV8, but haven't bought the CPU yet because I have exams and am studying for them. I would be happy to test on Dec 20 :toast:
Anybody have one? ;)
well, I will do some extensive testing over the next couple of weeks...
I have both the old version (with the single test) and V23.8 (with the 3 tests) and see what happens.
I have 3 3500+'s to use 1 winnie and 2 newcastles.. also have 2 Mobo's (1 Asus a8V and MSI Neo2)..
I was looking over at the GIMPS forums and versions 23.x Are optimized for The P4.. there is no code for the athon64 in Prime95. They are also having a problem with False prime numbers being reported on non-overclocked systems of all types (P4, AthlionXP and AMD64), so they could have some bugs in the code.
ALSO.... Remember that Memtest does not perform any Math calculations. it is Just reading and writing test patterns to memory and back. I have trashed my OS disk many a times finding Memtest never erroring, but Cpu was overclocked too high and trashed the OS...
Also, Make sure you get Prime95 ONLY from their website.. I have already found some copies that the files were corrupted, and would not run correctly at all..
Quote:
Originally Posted by Byron
Byron -
First of all overclocking results are going to vary - this is for issues with Prime at STOCK settings. If you are overclocking and failing prime95, then that problem is your own, the CPU isn't defective. My stability at 280 x 9 increased greatly once I cleaned / applied AS5 to my chipset as well. Just wanted to clarify that for people overclocking and not having prime stability is NOT the issue that we were getting at its the stock clocked prime failures.
hag6br
>>ALSO.... Remember that Memtest does not perform any Math calculations. it is Just reading and writing test patterns to memory and back.
And what do you think calculating something looks like at a lower level? Just like that.
>>there is no code for the athon64 in Prime95
A64 supports SSE2 so what's the problem? It might not run optimally, but it WILL run.
Please test the winch on the A8V and the K8N Neo2 @ stock. What week is it, btw?
ALL
Good news - local store (NCIX) has 3000/3200's 0441's in stock. I will be joining you in exactly 7 days.
What I am saying is that memtest only does memory reads and writes using a set pattern and compares them... Prime95 does floating point calculations and also uses Virtual memory (which memtest does not)..Quote:
Originally Posted by IvanAndreevich
ANd yes it has SSE2 code which he is still optimizing. Also, I was looking at his version info, and FFTS over 2048 have a time limit on them. If they do not complete in that time, they will fail.. (but do not know if this is implemented in the torture test).
My Winnie is a 0441. But I have a friend with a 0437 that I might borrow..
my Newcastles are 0427 and the other is an Original (bought it less than 2 weeks after they were released.
I am wondering on Prime95 though.. when they switched over to the New version systems that were Prime95 stable for 24hrs+ with the old version, Failed Prime95 within 15 minutes under the new version.
also, the version Presently in Beta is going to be a Huge memory Hog.. It is going to require 1GB+ of ram (both Physical and Virtual) to complete a run..
I havent tried priming at stock: A64 3000+ Winchester CBBFD 0442SPMW as you can see some of my results here
I have tried priming at 2520Mhz 280HTT 1:1 "Blend" and it fails after 3hrs or so...everytime i prime, always after 3hrs.
As you can see from my tests the pc is completely stable and has not crashed onces during testing and/or during everyday use/games.
I also had a 3200+ week 41 cpu that was in the same situation.
I will try and prime at stock speeds and report back.
Hey r2 :D
read up a lil futher, some one posted that by cleaning off the white goop underneath the MCP HSF and applying AS5, they could prime stable at a setting that would not have passed before...
Quote:
Originally Posted by jlccarv
That was me. Would crash priming at 278 x 9 sometimes 2 hr, sometimes 3 hr, now I'm at 280 x 9 now and primed 2 times over 12 hours - JOY JOY!
When i was using 3200+ week 41 i did apply as5 on the NB chip which made no difference to me.Quote:
Originally Posted by dnottis
This time i was experimenting as i applied the paste that came with xp-90 hsf (white goo) and i applied it to the NB and cpu (just wanted to compare to as5)
Ill tell you right now nothing compares to as5. However im going to remount the HSF and apply as5 and at the same time will fix up the NB with as5 is well.
Just currious, by why isnt AMD shipping 90nm CPUs in december?
over here the stock got empty in the first days of december, and dates on the different webshops now shows no stock b4 january...
Tip: try to LOWER the CPU frequency to 1200 MHz, 200*6, to see whether it fails, if it doesn't, then your CPU is defective, if it fails, it is something with the program or something is wrong with your system, and not the CPU (of course there could be problem with something another in the CPU, but not the operating frequency of CPU itself, maybe onboard memory controller).
:toast: :toast: :toast: :toast: :toast:
EVERYONE WITH PRIME95 FAILING AT STOCK !!! TRY THIS PLS ;)
OK, here's some more data into the equation:
Abit AV8 + 3200+ Winne Week 41 (0441MPMW)
260X10, vCore 1.55v = P95 stable depending on mem settings and timings (with divider definitely stable, no divider depends)
270X10, vCore 1.6v = SuperPI 31s, Benchs good, P95 blows immediately.
Sweet spot should be somwhere in between.
sorry for "late" responce...it was OCed...i will be doing stock testing on a new MSI mobo today...Quote:
Originally Posted by andyOCZ
o BTW Suma over on OC found P95 version: 2.46 which "seems" to be the latest versoin...? (DL from linked post)
just what we needed...more variables :p:
Fair enough andy, but if you can find more than one Winchester that can run Prime stably (and that's already been shown), then you have already shown it's not a problem with the software code, since all Winchesters are created using the same blueprints. Sure it's possible there are problems in AMD's manufacturing process with X week CPU and that could be causing some memory returns. I just want to make it clear that, unless there is some randomness in the stress test, the fault is not in the software.
Exactly, so no code specifically for A64's to go wrongQuote:
Originally Posted by hag6br
That's because of this: http://forums.extremeoverclocking.co...d.php?t=143717 :toast:Quote:
Originally Posted by v142
Ref
You're assuming that a CPU that is faulty @ 2.4 Ghz will not be faulty @ 1.2 Ghz, which is not necessarily true.
Quote:
Originally Posted by IvanAndreevich
Exactly - I see what he's getting at - but it's more important to Prime at STOCK to determine if a CPU is "defective" than to cripple it and run Prime95. It may prime fine at 1.2 Ghz and not be able to Prime at it's "stock speed".
I have a Winnie 0437RPCW and had no problems with Prime95 at 2313 (9.0x257) @ 1.575V. Now I keep it at 2250 (9.0x250) @ 1.475V with mem. @ 204. Made a 11 hour Prime at this speed and had no problems.
Quote:
Originally Posted by scorp
Interesting - heres a wrench in the week 37 theory -
what if it's only week 37 (and earlier) and MSI Neo2 Platinum boards?
Well...as I have mentioned before in this thread, my week 37 works fine too :), with 512MB RAM that is (2x256 DC in my case). 2x512 fails instantly. Scorp seems to run only one 512MB stick in single channel mode.
That's true :). I'll have to wait a little bit till I can get my hands on a pair of TCCDs :(.
I should try mine with 1x512 too :)
My system has been stable at 2.5ghz in dual channel. After turning off nv/ati speedup and removing the bad memory moduel (need to rma to consair) I got the sytsem stable up to 2.789ghz. that is prime stable for 7 hours along with 2 hours of 3d01 and super pi 32m. I dont seem to be having the prime bug... The only bug i have is that my chip doenst want to hit 2.8ghz!
BTW, im runnning on a week 41 3200 winnie.
scorp
Can you run the CPU and mem 1:1? Maybe it's a 1:1 glitch!
lsevald
..which strongly points to a $hitty memory controller.
MaxxxRacer
Seems like many newer CPUs don't have this problem.
Remember I started off with ny 3500+ week 37 and a GB K8NS Ultra-939 board. So the MSI only can't be true. I have nor read the whole thread but did anyone try a VIA board?Quote:
Originally Posted by dnottis
Ah, true - I forgot about that....Quote:
Originally Posted by andyOCZ
There's a guy stable on a VIA board on page 11.
OK, here's even some more data into the equation:
Abit AV8 + 3200+ Winne Week 37 (0437RPDW)
250X10, vCore 1.65v = P95 stable depending on mem settings and timings (RAM can not run 1:1) Verry stable
255X10, vCore 1.65v = Benchs good, P95 blows immediately.
Quote:
Originally Posted by Badge56
Sorry to hear that - I owned the ABIT A8V it wouldn't run 1:1 over 210 FSB. Personally I was disgusted with it and sold it.
Badge56
Can you test the highest your combo will run 1:1? It seems that people having this glitch are all 1:1!
well, I ran my Winchester on the MSI Neo2 for almost 23 hours straight with no errors..
My cpu is:
ada3500dik4bi
cbbfd 0441tpaw
ram was gskill Tccd..
My Winnie 3000+ fails prime every time at 1hr and 59 mins. Does anyone know what test this is running as its strange it fails at the same point every time :confused: (soory havnt a screenie of the test)
now try and OC it if you know it will pass prime stockQuote:
Originally Posted by hag6br
To those of you who think it is a hardware MODEL specific problem. Think of the following:
In order for a failure problem to be easily attributable to a piece of software, the failure:
1) It needs to fail in exactly the same test and exactly the same spot: it does not
2) It needs to fail on ALL units of a particular model, not just one or two or three units: it does not
3) It needs to fail consistently, and not on every fifth run or randomly: it does not.
However, as we have seen, Prime95 does pass several hours of various different prime95 runs (yes, even on some people's 90nm Winchester). Also, not all failures are consistent or in the same spot.
Thus, I still believe that the most likely case for this particular bunch of failures is pushing your system too far (regardless of whether 90nm Winchester or otherwise).
Only when Prime95 fails CONSISTENTLY, every time, in the exact same spot, on every piece of hardware that is of type X, is there a strong reason to believe that the problem is with Prime95.
Do you have reason to think otherwise from my above reasoning?
If so, what is the chain of reasoning?
I also suspect it is a specific production lot that has the problems.
what we do need to see is the third line on the chip since this is the production number.
Now has anybody seen a chip with a production date of 0438 through 0440? seems odd that they would not produce any chips for 3 weeks in a row.
I am going to see if I can borrow my friends 0437 chip..
Very true, but i guess AMD is working hard to release revision E 90nm cpus and i sure hope it shows great improvement over the current 90nm cpus :rolleyes:Quote:
Originally Posted by pc ice
hag6br
I think those other weeks just go to different parts of the world. For example, there is a guy here with a 0442 ;) But our local store around here has 0441's in stock.
We are not discussing overclocked failures - we are talking about STOCK CLOCKED Prime failures.Quote:
Originally Posted by halcyon
So unless you can "push" a stock system too far.....
>>So unless you can "push" a stock system too far.....
I suppose you CAN with timings. However, this is not the case here.
My 3200+ 441 will do 2.5Ghz (1.45 +5%) - any higher and prime isn't stable a few others here have the same exp with week 41. Also, it will do 2.4 Ghz on stock volts - making a $200 CPU a 3800 at stock volts is pretty cool too if thats all you are shooting for. Overall 441 is a decent overclocker.Quote:
Originally Posted by IvanAndreevich
Stock speed or not, the fault is in the HARDWARE (mem incompatibility, bad mobo, bad cpu, bad mem, bad timings, etc).
http://www.mersenneforum.org/showthr...6440#post46440
Only if you can make it fail consistently every single time in excactly the same manner in the same very spot of calculation with every single Winchester 90nm cpu, then it could be a prime95 bug. Do note that those (in/with) are logical ANDs and not ORs in that sentence.
But that does NOT happen.
Some 90nm Winchester perform ok.
And the failures are NOT consistent nor happen in the same spot of calculation.
regards,
halcyon
I agree, halcyon. I'm waiting for George, the maker of Prime, to reconfirm this.
Andy emailed about Prime95 and was told the same thing, don't know if it was George though.
BTW I've a 3500 winchester that fails Prime95 but it also fails SuperPi :(.
Hi all,
George Woltman, prime95 author, here. After reading through most of this thread, I agree with those that state this is not a software problem.
There has been some discussion of two previous software problems. The first occurred when running the blend test with a specific amount of memory (930MB IIRC). The result was a failure in the same spot with the same error message every time (roundoff = 0.5 in the first test of the 20K FFT). The bug was in my memory allocation code which caused the same memory corruption every single time. The second bug was in prime95's guess as to how much memory to use in the blend test. It was guessing too high and thrashing rather than torture testing. The code still "worked", just very slowly.
To those that say, "but I'm running at STOCK SPEED!", that does not change the argument that it is a hardware problem. Either the CPU is bad, memory is bad, the timings on the memory are too aggresive, the motherboard is bad, or the power supply is too weak or bad.
This forum is great for users to compare notes and figure out the most likely culprit. I think you've reached the conclusion that the CPU is the cause. If so, AMD has shipped some chips that they should have thrown out or downgraded. It happens. I recommend RMAing any parts that cannot pass stress tests at stock speed.
Intel had to pull production of its last P3 for similar reasons - programs failing at stock speed. See http://www.tomshardware.com/cpu/200008281/index.html
begins to ramble
Posted by saaya post# 70
so memtest works but prime fails? then it sounds like a program related error, i never heard that memory passes metest but fails in prime before... weird...
---- Memtest doesnt test a cpu the way prime95 tests it actually memtest doesnt even do anything really with the cpu. prime95 checks your answer agnist the real known correct answer
Posted by WICKeD post# 77
I think even a stock system is going to give small errors after continuous usage. Prime is probably the best test for stability, but I think there should be a limit on it. Most people are not going to run their system @ 100% for over 8 hours, without letting it cool. Unless you fold or do media encoding, I'd say it's fair to claim stability if it can run overnight. That's just my opinion.
----to answer your comment on limit if you can go over 8 hours on prime 95 theres something not right prime95 is running longterm on a lot of computers.
http://www.mersenne.org/primenet/status.shtml go to bottom of the page you will see a hourly updated info section of the project also susgest you visit 15 minutes after hour because server takes a few minutes to rewrite new data.
Posted by Byron post# 119
WiCKeD,
my mem is fine and my board, can run dual channel at same speeds no problems.
Must be prime problem, because system is rock solid, can run all benchmark programs etc, or bad mem controller... but if the mem controller is bad, then why it my system is stable?!
---- prime 95 tests the whole cpu some programs dont totally test it prime 95 looks for the smallest problem if there is a problem it will proibily be found. just because a system is stable doesnt mean it is truely stable. if you dont run proc 100% all the time then so be it but if prime 95 finds a bug then you know there must be something wrong prime 95 does advanced calculations and checks the answer your computer outputs agnist a true well tested and known answer.
Posted by [ R2 ] pst# 136
I dont know why so many ppl are so concerned with p95, just dont use it, its that simple.
There are other programs to test stability of your cpu and etc...
---- yes but they dont do a complete stress test.
Posted by Paa' post number 177
I would think there was'nt anything wrong at all with prime95 just the fact that a64's have built in memory controllers on the cpu means prime can fail for either and it's hard to tell..
i need to up my vcore to reach 250 fsb.. even with a below default speed.
that beta prime95 im running seem's to do one thing different ive NEVER noticed in using prime95 for 3+ years..
blend uses 99% cpu and all your ram.. where as before it would use only a little bit of the cpu and lots of ram..
now it's near 100% both because and i guess? they saw that with the mem controller being on the cpu they need to tax that aswell even when testing memory.. which would make memtest86 useless ?
---- when you say it used to use a lot of ram and low cpu your settings are off. your spilling over into virtual memory and that causes bottleneck problems prime 95 accesses the data in ram 100 times a second or more and uses a great ammount of bandwidth if you lower the ammount of ram you let it use youll notice it wont bottleneck thats because the ide/sata wont be interfearing.
Trying to answer a few misstated ideas after fully reading thread also check out
www.mersenneforum.org there are proibily answers you are looking for in the hardware or software thread.
Crunchy
Superpi @ stock?
prime95
Have _any_ AMD representatives contacted you about this? Intel made good on fixing the problem with the PIII's, so I hope AMD makes good on fixing their problem. However, I think they would need to care about the problem first and try to understand what's going on.
Thanks for clearing up the situation.
Just exactly how many ppl here have problems with prime stability on -stock- setting?
and stock meaning: auto settings on -all- the memory setting, not running 2-2-2-5-1T on ur PC4x00 just coz other TCCD chips can do it, for instance.
and if u have more than one set of memory try that too on rated timings too?
how many bad CPUs are we talkin about here, that cant run prime on real stock settings? 1 or 2? I read one was RMA
My best idea to sort this out is too let everyone who has problems with prime start a new tread, and list all components and settings used, too see if there is any combination of hardware that could be the cause.
Regarding OCin
My guess is that the memory controller is the limiting factor her, but the fact that the CPU "seems" stable on much higher speed, leeds too high hopes,
Just face it there IS a reason AMD didnt include the 3800+ in the first 90nm series launch.
On the other hand it does indeed seems strange why there hasnt been 90nm CPUs in stock the last 2 weeks, with still a week or two too go b4 new stock is said too arrive...
One question for everybody with Prime failing. What if you run 2T @ stock - does it become stable, or does it still error?
No problems here at stock with my winnie 3200+, it can run Prime95 for over 24 hours @2.4GHz with 10x 240HT @1.4v (with 166 divider), any higher at any vcore voltage and it would fail Prime95, tried lowering the overall CPU speed and increasing the HT but it won't go past 240MHz even with very relaxed memory timings @3-4-4-10 @2.85v!
I initially didn't expect great overclocking potential from my winchester chip compared with the mature Newcastle cores and I was right, I'll probably sell my chip for a later revision of these winchester chips probably sometime after Jan/Feb. :)
IvanAndreevich: My system fails Prime95 with everything in the BIOS at stock speeds. I've tried it on everything auto and I've tried it by setting to manufacturers recommended settings (ie 2-2-2-5 timing, 2.75v etc). Details about my system are in my sig.
The thing I'm really worried about though is that it's failing SuperPi as well :(. Everthing else is fine, 3DMark 2001, playing games such as ET, PiFast, no crashes at all. :stick:
I'll try 2T command rate, see if that helps.
No one from AMD has contacted me, nor would I expect them to.Quote:
Originally Posted by IvanAndreevich
This may or may not be a serious problem. The Intel problem occured because they tried to push the P3 process too far. Isn't this the first of the Winchester line? If so, it is more likely a startup problem with improper binning.
The best chance to get AMD to notice that they have a problem is to get the press involved. If one or more of the online news outlets report that the chip doesn't run properly at stock speed, then AMD will take notice.