[JigPu]

I hadn't a clue that Whichesters had this kind of a problem (since I don't follows CPUs nearly as much as I should ), but has anybody brought up the possibility of a tweak to the Athlon64's FPU causing the difference? It may very well be stable even with the error, and that Prime95 is just too sensitive (read on if you think I'm loosing my marbles )

As I'm sure you all know, floating point math is inherently error prone. Rounding errors from loss of precision creep in all the time, though are usually kept below notice from additional bits (for example, my calculator will display numbers with 10 digits of precision, but internally does all calculations with 13 digits to prevent errors from being noticible).

The math is still deterministic since the same error will be introduced every time, though it isn't nesscaraly deterministc across kinds of chips. Every manufacturer implements the FPU logic a little different from everyone else, meaning that the way each chip creeps in error will be different. Could it be that Prime95 actually causes the precision error to bleed into it's results, and thus fail when it encounters a different precision error because of different FPU implemetation? It wouldn't really be a sign of instability (since would be the fault of the creators of Prime95 that they don't use precice enough variables), but would still cause it to fail from "rounding errors" and such....

Just random ponderings...
JigPu

[HousERaT]

I'm asking this question in case it hasn't been asked already......

Will the Venice cores suffer the same fate as the Winchesters? Will the Venice chips be Prime95 stable?

[Sentential]

Yes. The issue primary lied with the memory controller which has since been fixed

[HousERaT]

God I hope so..... I've spent a good part of the day looking over all the posts at XS about this...... I'll be real ****ed if my first Venice won't run prime at stock.....

[deception``]

It shouldn't matter too much.

There are plenty of individuals who enjoyed completely stable Winchester systems that wouldn't run Prime worth a damn. That alone should tell you that Prime 95 is not the end-all for stability.

deception``

[edwardaune]

Amen

[Scar]

Originally Posted by HousERaT

Will the Venice cores suffer the same fate as the Winchesters? Will the Venice chips be Prime95 stable?

they always were. Sure, it's not the "end all" but those of us who hold a higher standard of stability still rely on it. At first people were trying to blame it on prime saying it was outdated, not optimized for a64's or a software glitch. then they started blaming it on the memory controller. i blame it on user error and higher than achievable expectations.

[deception``]

Originally Posted by Scar

they always were. Sure, it's not the "end all" but those of us who hold a higher standard of stability still rely on it. At first people were trying to blame it on prime saying it was outdated, not optimized for a64's or a software glitch. then they started blaming it on the memory controller. i blame it on user error and higher than achievable expectations.

Just because I don't care for Prime....that does not mean that I don't have my own regimen for stability testing. On the contrary, I am certain that PCs I build for myself and others undergo much more thorough testing than just leaving a program running for 24 hours. I have learned in my experience that a PC can be game, encoding, multitasking, etc stable and yet can still error out in Prime. Hence, it is a moot test for me.

deception``

[HousERaT]

Originally Posted by Scar

they always were. Sure, it's not the "end all" but those of us who hold a higher standard of stability still rely on it. At first people were trying to blame it on prime saying it was outdated, not optimized for a64's or a software glitch. then they started blaming it on the memory controller. i blame it on user error and higher than achievable expectations.

that sounds like a corporate explination for chips that don't exactly meet up to spec..... and if you've read as many threads as I have on the subject you'll know there were plenty of winchesters that couldn't run prime at stock levels.

[hitechjb1]

I define a system including all its components (CPU, memory, chipset, video subsystem, ...) stable at certain frequency, voltage and temperature only if it can run ALL "well tested" and "error free" programs with properly installed OS for certain long time.

If some programs, drivers crash the system with properly installed OS (with fixes if necessary), and if it can be established that the abend, crash, exception are not due to the software, I do not consider the system as "stable".

If a CPU or a memory module or something else is isolated and proven to not run certain "error free" and "well tested" programs (including proper drivers, OS) at stock specification, I would request RMA.

If a system does not pass Prime95 which is consider as a "error free" and "well tested" program, the system is not stable.

If a system does not pass program XYZ which is consider as a "error free" and "well tested" program, the system is not stable.

[nicknomo]

Its only stable if it doesn't consistently produce a self caused error on ANYTHING within a reasonable time period... Otherwise its not stable...

[trueplaya4ever8]

Originally Posted by deception``

It shouldn't matter too much.

There are plenty of individuals who enjoyed completely stable Winchester systems that wouldn't run Prime worth a damn. That alone should tell you that Prime 95 is not the end-all for stability.

deception``

I argee with deception, as long as it works when you do the things u want on ur comp, who cares if prime 95 fails.

[HyperYagami]

I also consider data integrity as part of stability as well. Stabilty means a lot more than just "uptime". You don't see any system crash or visible error right away doesn't necessesarily mean everything is fine. Now granted there's no way to make sure a system is "100% error free" either, but if it can't even run a simple program error-free that's a pretty bad sign.

[yeha]

well let's look at the prime95 situation - the program as stored on the hard drive doesn't change between runs, so it's sending an identical stream of instructions to the cpu time after time after time after time after time. the cpu, given identical streams of instructions, is producing different results in the case of a prime95 error. if a cpu receives a set of instructions, from the same starting point, it should give the same end result every time. what would you say if your calculator, one out of every 10 times, gave 2 + 2 = 5? if you start with a constant number and set of calculations, no matter how many times you run it you should always get the same answer. anything else indicates a lack of determinism - if you can't guarantee determinism, you can't guarantee stability.

prime95 is definitely exploiting a bug or very weak path inside the winchester line. those same instructions should always give the same answer - the math is deterministic, the only way an error can creep in is if the cpu is giving a wrong answer to an equation at some point, scheduling uops in the wrong order, incorrectly combining writes, reading invalid values from cache, etc. - that's unstable in my book. the fact that no one's been able to pinpoint what exactly generates the winchester-prime95 error just worries me more - how do you know what other programs are silently flipping bits without raising errors? i know all the code i've written assumes that the cpu is calculating things accurately, it's not often you see code like:

foo = bar + 5;
if (foo != bar + 5) die("your cpu sucks");

[rseven]

I have often said that I'm from the old school and believe that a system is not stable until it can be Primed for 12 hours. This seems to continue to be a heated controversy among Winchester users, many of who complain that theeir system won't pass Prime at stock speeds. What strikes me as strange is that others like my own will pass Prime with a 40% overclock. This is a rather huge variance in systems that no one has been able to explain. If the problem does lie with AMD let us hope that it is resolved with the Venice.
Side note: I worry that the Venetian canals will short out the new core! Also, how will they fit a Gondola in there?

[j3lly]

those p95 errors most likely due to crappy ram people usally run "blend" which p95 even tells you "lots of ram tested"

the way to isolate teh cpu for p95 stabilty test is to use custom FFT according to to L2 cache,IE: minimal FFT size 256k/512k/1024k -> maximum FFT size 256k/512k/1024k

[Scar]

Originally Posted by HousERaT

that sounds like a corporate explination for chips that don't exactly meet up to spec..... and if you've read as many threads as I have on the subject you'll know there were plenty of winchesters that couldn't run prime at stock levels.

actually, i have read alot. it's all random problems. surely you realize p95 can fail for more reasons than the cpu right.

[HousERaT]

sure but if a person can't get a cpu primed at stock then they switch out the cpu for another and all of a sudden prime works that's a strong indicator something is going on with the cpu. For whatever reason some A64 cpus don't want to work with prime even though they appear to be stable using other programs. My original question is whether or not this same phenomema will rear it's ugly head with the Venice.......

[dropadrop]

I'm probably one of the more oldschool guys, and would not consider keeping a system running that does not pass a day of prime95. What was said above is completly valid, and it cannot be put down to selecting the wrong test in prime. People with bad processors have done alot of differant tests with prime, not only blend. There have also been alot of cases where people have changed the processor and things have been fine.

If you don't care about the possibility that your computer is slowly degenerating under use, then this is a moot point. If you prefere to never get suprises then the only way you can be sure is by keeping the computer at a level that passes prime. Alot of code is not dependent on small errors, but if you are getting them you can count that sooner or later you can run into problems.

[Scar]

Amen to that! ^

but to HousERaT, to answer your question in my pov, i doubt it. the venice is a core revision of the winchester. so