A64 winchester has issues with Prime95

[flexy]

Thread at xtremesystems

this made big waves - the INQ even reported about this.

AMD gave a statement denying there are problems with the A64.

The issue looks like that there's huge number of early A64 winchester cores which cannot even pass a Prime95 at *stock* speeds.

Please...(also in the other thread)..lets not make this a thread re: "The sense of running Prime95". This is irrelevant.

As a long-time overclocker i ALWAYS used Prime95 to confirm whether my system/CPU is stable.
If i get a new CPU and it would not be able to run Prime95 at STOCK (!) then i can definatly call this CPU "defective" - it doesnt matter whether all other "applications and games" run fine.

Overclocking is much more complex than what some people think.

I had my share with people claiming their overclocked systems being "stable" just because they can run 3dmark or their favorite game...but as said its not THAT easy.

ASSUMING that its not a problem with Prime95s code - so what do we make of this ?

Remember: SOME CPUs cannot pass P95 at stock - we're not even talking overclocking here.

---

I will continue to follow the discussion over at xtremesystems - i do not agree with everything said. Especially "conclusions" which dont look too "scientifically" to me.

If ONE person has a week 37 CPU which fails, and then puts a week 44 CPU is which is fine..this proves of course *nothing* 🙂 But there seems to be anumber of people with this problem...

 

[Duvie]

I have a week 36 cpu...Want to see the pics of the prime95 passing????

I think there is a lot of speculating at this point. We don't know the status of a lot of those ppls other components and whether they have properly configured things....Proper bioses??? Proper driver updates??? conflict with other software, perhaps an anti-virus software interfering??? It all happens....

 

[Dethfrumbelo]

Interesting question: Does failing Prime95 at stock qualify a CPU as defective?

If the CPU can run Windows normally, but fails all the stress tests, would AMD allow for return/replacement under warranty?

 

[flexy]

Originally posted by: Duvie
I have a week 36 cpu...Want to see the pics of the prime95 passing????

I think there is a lot of speculating at this point. We don't know the status of a lot of those ppls other components and whether they have properly configured things....Proper bioses??? Proper driver updates??? conflict with other software, perhaps an anti-virus software interfering??? It all happens....

I just posted almost the SAME over at xtremesys in the thread - you are totally right. I was criticizing their "semi-scientific" approach - with the same points and concerns you just laid out here 🙂

 

[Markfw]

But I can;t pass F@H at better than 2.2 on a 3200+ winnie. There may be a "little" truth to the early cores (don;t know what week mine is, but I got it a month ago.)

 

[flexy]

Originally posted by: Dethfrumbelo
Interesting question: Does failing Prime95 at stock qualify a CPU as defective?

If the CPU can run Windows normally, but fails all the stress tests, would AMD allow for return/replacement under warranty?

I'd say...if its PROVEN there is no problem with the P95 code, and a CPU fails at STOCK - yes, it is defective.

As for the other question: I dont know...even worse for me, i got a OEM CPU and the warranty run out *yesterday*. (I couldnt test the CPU yet since i dont have a board since i cannot test my CPU on a frickin' paper-launch motherboard 🙂

The question for me would be if there is a way to exchange/RMA a PROVEN faulty OEM CPU/..eg. if it turns out there is indeed a problem with these chips.

This needs a LOT MORE good, systematically investigation (and not only guesswork like in the orignal thread)...in case there *IS* really something to it.

Otherwise it might be difficult because AMD of course denies problems..so is the programmer of Prime95. Doesnt change the fact that some chips fail at stock - as it seems.

 

[Nyati13]

Originally posted by: flexy


Otherwise it might be difficult because AMD of course denies problems..so is the programmer of Prime95. Doesnt change the fact that some chips fail at stock - as it seems.

You're assuming that it's a fact that the Prime95 is failing due only to the CPU. That can only be determined if a CPU is failed, and then immediately replaced with a second CPU while changing absolutely nothing of the entire rest of the system, and the second CPU then passing Prime95. Has even one person done that yet?
It's actually just as likely that the problem is behind the keyboard of whoever is having problems.

 

[flexy]

Originally posted by: Nyati13

Originally posted by: flexy


Otherwise it might be difficult because AMD of course denies problems..so is the programmer of Prime95. Doesnt change the fact that some chips fail at stock - as it seems.

You're assuming that it's a fact that the Prime95 is failing due only to the CPU. That can only be determined if a CPU is failed, and then immediately replaced with a second CPU while changing absolutely nothing of the entire rest of the system, and the second CPU then passing Prime95. Has even one person done that yet?

yes ! people swapped week 37 winchesters for week 44...and then it worked. Or replaced winnies with newcastle...then it worked too.

 

[Link]

I had week 37 3500+ running at 2.5ghz, and it passed Prime 95.
At 2.6ghz, it needed 1.55~1.58V to pass the Prime95, and temp reached 45C.

 

[Sparky19692]

will they RMA the chips? Mine at stock or even below stock WILL NOT EVEN START PRIME 95 everything else runs Great! I have been using F@H for all testing because of that. F@H will run for days on end. Stock or O/C. How do yoou know what week it was built?

 

[Duvie]

It is written on the integrated heat shield....So if you are running the chip then you will need to pop the hsf and cleanup the thermal paste to read it...

 

[Sparky19692]

I wrote it all down before I installed it. now I know why what is the week?
ADA3500DIK4BI
CBBFD 044TPAW
1121205J40127
I can not make a date out of that.

 

[Duvie]

YOu missed a number!!!

CBBFD 044TPAW??? The 044 number portion should have been 4 numbers...04 would be the year and the next 2 digit number would be the week of that year....

 

[beatle]

Interesting, that looks like it could be a week 44, but there's usually an 04 prior to the week on the same line. Are you missing a zero? 🙂

My Winchester is a week 41.

 

[Zebo]

God I hope not. This could seriously screw AMD in public eye just when they stood to make some little entry into mainstream.

I definity want some scientific testing by reputable sources than these crazies (in a good way) at xtreme who slam these processors to ridiculus levels before saying anything.

 

[Zebo]

Flexy no offense but you guys are worse than a sewing circle over there... I mean come on the first page and on is littered with 40-50% overstock overclcoks and some are questioning prime stability? Well duh. Where are these case(s) of stock failing? I bet you it's a NF3 ( here's my entry into the circle: ever wonder why enthusiast kings like DFI/ASUS/ABIT never made a 939 NF3 board?) or mem issue or even worse user error.

I think it's memory for these stock guys if anything ....on A64's is much more difficult.

check this out from mushkin:

One of the most critical issues in this resepect relates to the first step in the memory access sequence, that is, the opening of a memory page and subsequent steps until a read command can be issued. All of these steps fall into the time frame referred to as RAS-to-CAS delay or tRCD. In a typical system, the memory controller on the chipset is running at bus speed. On the Athlon64, it is running at CPU speed, which means that it is A LOT faster. That means that the internal sequence of commands also goes faster, and that's where all systems are getting hiccups if the memory chips used can only run at a high tRCD.

A very trivial explanation is that the controller expects that the memory is as fast as it itself but we don't have memory yet running at a 2 GHz and beyond clock rate. What it comes down to is that, depending on the BIOS tuning (by the manufacturer), the read command is issued a bit too early and violates the tRCD. The workaround is to add one additional cycle to the actual memory latency, that is, a 2:2:2-rated DIMM will run stable at 2:3:2; a 2:3:2 rated module will, most likely, need to be set to 2:4:2 in order to run stable (where the latencies are CAS:tRCD:tRP). Bottom line is that when shopping for memory for the Athlon64, it is highly advisable and in most cases mandatory to only consider memory with a tRCD of 2 or, to make it simple, memory that is rated at 2:2:2 a the designated speed. Alternatively, registered DIMMs can be used as well without problems, even though they are slightly more expensive and will give lower performance.

 

[Sparky19692]

well that is all that I have witten down maybe I missed a number I was going to take a digital pic #^%$&I knew I should have.
Zebo as far as Memory at stock it should not be giving me any probloms. As compared to Mushkin its every bit as good, TCCD chips and all. I have yet to find a negative review on them.
I reloaded my PC last week maybe I will try installing it and running again. Don't get me wrong I'm happy with the chip I had just figured right from the start that it was a Prine 95 problom because everything else runs great.

 

[sandorski]

I think Zebo is onto something. Most of us know that "best" Memory Timings on Athlon64 are not what most would think(lower is better), but that for at least one Timing option a higher setting works "best". How many threads have we seen of people complaining about their memory not working properly at their "best" Timings, then having it explained that they don't want that Timing anyway as it is not the "best" for their Athlon64/mobo combo that they have?

Sure, it's possible there is something wrong with the early batch of current Athlon64 cpu's However, especially since it seems Overclockers are having the problem most, it is equally likely that other issues are causing the problem. It is also possible that AMD made a slight change after the first batches that minimized or eliminated the issue, but that in itself doesn't mean the CPU is/was defective.

 

[flexy]

>>>
It is also possible that AMD made a slight change after the first batches that minimized or eliminated the issue, but that in itself doesn't mean the CPU is/was defective.
>>>
there are people who claim they cant even run Prime (and Knoppix Linux and various other CPU taxing stuff) at stock speed.

Give me a break...we're not even talking overclocking here...but i would have reason to be VERY concerned if there's a bunch of programs a certain CPU just is UNABLE to execute.

For all the other people who clain "...but its stable running Folding@Home/Seti etc." <-- this is the PERFECT example of mistakes being made by people who overclock but have a lack of certain insight. (Sorry, dont want to sound like a ahole - but its true nevertheless. Its just more to overclocking than what many people think).
If you run Folding/Seti and it does not CRASh - well it doesnt mean your system is stable. Does Folding test against fixed/known results internally ? Not to my knowledge !
One of the most overseen facts about overclocking (see: Prime95 !) is that the system/CPU very well *works* - but it calculates WRONGLY due to overclock.

You will never know whether your results are actually pure nonsese/wrong results with Seti/F@H...you will think you machine is fine....but in reality the results you submit (especially for distributed computing) are WRONG/invalid.

Thats the beauty of Prime....it gives you an error message at this point - WAY before your system actually would crash.

Unless someone lists a another, alternative program which does simioliar calculations like prime and VERIFIES these achieved results you can only GUESS whether your system/CPU is stable - "Not crashing" does NOT mean your system is stable/calculates right.

I still think this whole matter deserves more attention since i would feel like a fool spending $315 on a CPU (warranty is already run out) which COULD turn out a dud once i receive my motherboard and am able to test.

 

[sandorski]

Originally posted by: flexy
>>>
It is also possible that AMD made a slight change after the first batches that minimized or eliminated the issue, but that in itself doesn't mean the CPU is/was defective.
>>>
there are people who claim they cant even run Prime (and Knoppix Linux and various other CPU taxing stuff) at stock speed.

Give me a break...we're not even talking overclocking here...but i would have reason to be VERY concerned if there's a bunch of programs a certain CPU just is UNABLE to execute.

For all the other people who clain "...but its stable running Folding@Home/Seti etc." <-- this is the PERFECT example of mistakes being made by people who overclock but have a lack of certain insight. (Sorry, dont want to sound like a ahole - but its true nevertheless. Its just more to overclocking than what many people think).
If you run Folding/Seti and it does not CRASh - well it doesnt mean your system is stable. Does Folding test against fixed/known results internally ? Not to my knowledge !
One of the most overseen facts about overclocking (see: Prime95 !) is that the system/CPU very well *works* - but it calculates WRONGLY due to overclock.

You will never know whether your results are actually pure nonsese/wrong results with Seti/F@H...you will think you machine is fine....but in reality the results you submit (especially for distributed computing) are WRONG/invalid.

Thats the beauty of Prime....it gives you an error message at this point - WAY before your system actually would crash.

Unless someone lists a another, alternative program which does simioliar calculations like prime and VERIFIES these achieved results you can only GUESS whether your system/CPU is stable - "Not crashing" does NOT mean your system is stable/calculates right.

I still think this whole matter deserves more attention since i would feel like a fool spending $315 on a CPU (warranty is already run out) which COULD turn out a dud once i receive my motherboard and am able to test.

CPU at Stock, but RAM at proper settings as well? Although, it is important to note that certain ram sticks at Stock don't work well. My point is that there are a number of things that could cause the error, not just a defective cpu.

edit: I agree though, this issue should be thoroughly investigated.

 

[Duvie]

Originally posted by: Duvie
I have a week 36 cpu...Want to see the pics of the prime95 passing????

I think there is a lot of speculating at this point. We don't know the status of a lot of those ppls other components and whether they have properly configured things....Proper bioses??? Proper driver updates??? conflict with other software, perhaps an anti-virus software interfering??? It all happens....

Again I say.....