Message boards :
Number crunching :
Optimized app for Opteron Dual core?
Message board moderation
Author | Message |
---|---|
KWSN THE Holy Hand Grenade! Send message Joined: 20 Dec 05 Posts: 3187 Credit: 57,163,290 RAC: 0 |
What's a good optimized SETI app for the AMD Opteron-165? according to CPU-Z, it has SSE3, but all the apps I see for SSE3 are only for Intel chips... I've downloaded a SSE2 app, BTW, but won't use it 'till I get a response. . Hello, from Albany, CA!... |
htrae Send message Joined: 3 Apr 99 Posts: 241 Credit: 768,379 RAC: 0 |
The SSE2 Generic 2.2B Chicken application is the best you can do for the Opty 165. SSE2 Generic 2.2B without Graphics |
KWSN THE Holy Hand Grenade! Send message Joined: 20 Dec 05 Posts: 3187 Credit: 57,163,290 RAC: 0 |
The SSE2 Generic 2.2B Chicken application is the best you can do for the Opty 165. Thanks, that's the one I downloaded. Anyone working on a non-Intel-only SSE3 app? . Hello, from Albany, CA!... |
htrae Send message Joined: 3 Apr 99 Posts: 241 Credit: 768,379 RAC: 0 |
The SSE2 Generic 2.2B Chicken application is the best you can do for the Opty 165. I believe (correct me if I'm wrong) that a SSE3 version was coded but the SSE2 version turned out to be just as fast, if not a little more. |
KWSN THE Holy Hand Grenade! Send message Joined: 20 Dec 05 Posts: 3187 Credit: 57,163,290 RAC: 0 |
The SSE2 Generic 2.2B Chicken application is the best you can do for the Opty 165. Thanks, I was wondering if the AMD chips' SSE3 instructions were screwed up, like the some of the FPU instructions on the early versions of the Pentium® were. (Yep, some of us have long memories, and won't let Intel live that one down!) . Hello, from Albany, CA!... |
Francois Piednoel Send message Joined: 14 Jun 00 Posts: 898 Credit: 5,969,361 RAC: 0 |
The SSE2 Generic 2.2B Chicken application is the best you can do for the Opty 165. FUD |
ML1 Send message Joined: 25 Nov 01 Posts: 20443 Credit: 7,508,002 RAC: 20 |
...instructions were screwed up, like the some of the FPU instructions on the early versions of the Pentium® were. (Yep, some of us have long memories, and won't let Intel live that one down!) Very much fact. Worse still, followed up by some very crass marketing that it was all a billion-to-one-chance that noone could possibly ever notice!!! Shame it screwed up many calculations for many people. It was uncovered due to some astronomy simulations giving wildly unexpected results... Many hours wasted and the cause was proved to Intel whom then grudgingly admitted the obvious. There's even a special flag that notes whether or not a software patch is needed for the Intel FPU bug! Hopefully, lessons have been learnt so that such a fiasco is never again repeated. All good silliness! Happy crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Francois Piednoel Send message Joined: 14 Jun 00 Posts: 898 Credit: 5,969,361 RAC: 0 |
...instructions were screwed up, like the some of the FPU instructions on the early versions of the Pentium® were. (Yep, some of us have long memories, and won't let Intel live that one down!) what about your start crunching instead of FUDING! Your RAC is 2! we should make it a rule, number of posting per month = RAC /100, it could avoid this Kind of FUD. who? |
Spear Send message Joined: 15 Nov 01 Posts: 49 Credit: 6,365,604 RAC: 0 |
He's stating a fact. http://en.wikipedia.org/wiki/Pentium_FDIV_bug |
Francois Piednoel Send message Joined: 14 Jun 00 Posts: 898 Credit: 5,969,361 RAC: 0 |
but only present one side of the story! AMD has bugssss too: read it here and that is on much more recent product. this kind of FUD is not very constructive, the Pentium bug is a long time ago, this one is much more recent ... what about this one: AMD opteron , Athlon 64 bug, mobile, the all familly has a bug (on this one, the Opteron hangs ... the pentium bug was 1 out of few billion chances of making a mistake, and very few divisions ... humm hummm choose what is worst! Witch one is really a big deal? hanging? or making a little mistake one time every few century? end of day, try to understand, we are building a system is 220 000 000 transistor now, AMD and Intel are trying their best to be close to perfection, and we both are. wiping a bug in our face is not very smart, it shows how little you know about about microprocessor design, if you compare to the software people, we are millions of time more realable, so, pleaseeeeeeeeeee this is FUD! AMd did fix gracefull their bug, and you did not see the Intel people screaming it all around, we are gentlemen. I just posted them here that you know it happen in both side. I do not blame them for having it because I understand the size of the problem. Educate yourself! I forgot, wiki is getting full of FUD, fanboys are polluting the encyclopedia. we are gentlemen enough to do not go and enter our competitor bugs into it. I recommand that your read the book from Andy "only paranoid survive", you ll then understand the bug of the pentium history, by the man who had to deal with it. As we should all know, Internet is not a safe source of information. I could publish a web page saying that SETI found aliens, but the USA administration catched the bit before it when into the data base server and locked it into area 51 ... Got the point? The computer industry is victime of a tabloid effect, where may people get so emotional about it that they stop being racional. don't throw rocks before you know if somebody is guilty, and how much he is, and is he responsable? Microprocessors are a great archivement for human kind, and we keep getting hate from people who choosed blue or green ... understand that we are dedicated to improve microprocessors, in doing so, some time we have informed risk, it is call "innovation". give us the credit for what we are doing, green or blue, and respect us: I work between 12 to 15 hours per day! And get your RAC higher than 2! who? is your daddy now! |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
Agreed, but I wouldn't say it isn't FUD per se since the defect does exist, and yes all processors have errata. It seems to me it's more "axe grinding" in the way some folks try and spin a situation to influence the opinion of people who don't how the semiconductor industry works. Suffice to say the reason errata are published by the manufacturers is so they can worked around in either the BIOS and/or the operating system. As I'm sure you're aware, many get fixed in the next stepping, but a fair number are insignificant enough they neve get dealt with in the silicon. IMHO, it's a tribute to all semiconductor manufacturers they are so open about publishing information this regard. Also, whereas the Pentium FDIV errata was a particularly nasty one in certain specific cases, AFAIK Intel made good to anyone who had a chip effected. Also, as you point out it's ancient history at this point. Alinator |
hiamps Send message Joined: 23 May 99 Posts: 4292 Credit: 72,971,319 RAC: 0 |
Also, whereas the Pentium FDIV errata was a particularly nasty one in certain specific cases, AFAIK Intel made good to anyone who had a chip effected. Also, as you point out it's ancient history at this point. Alinator [/quote] My neighbor had a bad chip and they sent him a faster one to replace it. He was very happy as back then even little upgrades were expensive. Official Abuser of Boinc Buttons... And no good credit hound! |
zoom3+1=4 Send message Joined: 30 Nov 03 Posts: 65825 Credit: 55,293,173 RAC: 49 |
Yeah I remember that, But back then I didn't have a PC so It was only something to read about then. The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's |
KWSN THE Holy Hand Grenade! Send message Joined: 20 Dec 05 Posts: 3187 Credit: 57,163,290 RAC: 0 |
[snip] The Pentium® FPU bug is famous, anyone in the computer industry (as I've been since 1973!) would know it - and that Intel™ made it worse by (initially, at least) asserting that a) very few people would notice [probably true... but publicity poison!] and (worse) b) that to get the chip replaced you had to prove that your calculations would be affected. This made it a public relations nightmare, that changed the way the chip-making industry handles errata... Yes, AMD has had its errata also - I'm not disputing that... but none has been quite as well reported! With the number of transisters, and other components, on chips today I'd be suprised if there weren't errors on every chip out there... (there is sometimes art as well, in electroniclly unused areas of the chip...) As far as FUD goes, I was reporting a well-documented, well reported (at least where I live - 50 miles north of Silicon Valley) old case - no dis-information there. I'm glad to say that the chip I have (a 939 Opteron-165) is not affected by the Opty bug you mention. (or at least it isn't on the list in the article...) As far as RAC goes, mine's 500+ and climbing... I haven't plateau'd with the new platform yet. . Hello, from Albany, CA!... |
Francois Piednoel Send message Joined: 14 Jun 00 Posts: 898 Credit: 5,969,361 RAC: 0 |
[snip] Who ever had it and wanted to get it replace did get it replaced. Nobody had to prove anything, the serial number of your CPU was enough. I still remember the statistic: In 2007, we are 3 years away from getting the 2nd person to meet this "feature" again (Assuming all those Pentium are still used today) ... Intel is a pioneer, and had to learn for the entire industry, Do you keep reminding your daughter about her 1st car accident every morning? The Pentium Bug feel a little like this to me. who? |
ML1 Send message Joined: 25 Nov 01 Posts: 20443 Credit: 7,508,002 RAC: 20 |
...end of day, try to understand, we are building a system is 220 000 000 transistor now, AMD and Intel are trying their best to be close to perfection, and we both are. Which is why there is such extensive automated test. However, there is still the Human failings and Management pressure to 'short-circuit' some of the test design time... With full test coverage, then the chips can be tested to ensure that the logic operates as designed. It is then just a question of the design itself, and of circuit timings... All tremendously detailed and difficult, but all completely useless if it does not work completely predictably and reliably. ...I still remember the statistic: In 2007, we are 3 years away from getting the 2nd person to meet this "feature" again (Assuming all those Pentium are still used today) ... And you must equally know that the statistic is completely contrived and pointless. (OK, very bad floating point pun!) So you go to your bank and you ask for your pay check... and they reply that "Oh sometimes we will get it right, and we know that sometimes we will get it wrong. You check it and tell us for every one of our customers for when a few of them trip over the known error... We know there's only a few that might find it..." The reality is that the known error, which Intel kept quiet about, caused seriously wrong results for real people whom did find it, and whom had found it expensively the very hard way. Simply: Intel made an FPU that was incorrect and that could not be relied upon to always give correct answers. Your Marketing people can play all the silly games they like with the vastness of numbers to play silly games with statistics. Some numbers still came out wrong and the Marketing people didn't care about anything other than their own favourite numbers... But in any case, this is all a very old story, and a very well known old example. This is all a game of big numbers and incredible detail. With a complete design and a complete independent test for that design, the chips can be tested to be 100% correct for their logic. Reality is that there is a game of always trying to 'cut a few corners' without losing 'the game'. Aside: I run diagnostics software that checks that the operation for the CPU and RAM is as 'expected'. Its then a game of chance whether I might catch a hardware failure or a hardware design problem... Note that I have equal suspicion for both Intel and AMD. Good luck for your design working, Regards, Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
KWSN THE Holy Hand Grenade! Send message Joined: 20 Dec 05 Posts: 3187 Credit: 57,163,290 RAC: 0 |
[snip] yeahbut- Intel™ marketing division is really what made this famous, as (as I stated above) initially their response was: 'It probably won't occur in your computer - but it might' (paraphrase) and 'we won't replace it unless you prove to us that your calculations are affected'. (again, paraphrase) That initial response is what made this the computer industy's answer to the Ford Edsel (or Chevy Corvair - but that was a real safety issue!) The fact that Intel later had to replace most of the initial batch of Pentium®s because of the bad publicity of the inital response has nothing to do with how much infamy this bug produced for them. Of course, part of the reason that some people wanted the Pentium®-60 when it came out in the first place, is that it was the first of the x86 line with an on-board FPU, instead of having to buy a seperate x87 chip. To learn that that FPU wasn't flawless left most people stunned. (Intel had done a good job with the x87 chips until then.) As far as being "3 years away from meeting the 2nd person to meet this 'feature' again", there are probably fewer that 500 of the buggy chips left computing in the world... I didn't have one of the chips, BTW - at the time I was using a Cyrix 486. . Hello, from Albany, CA!... |
kittyman Send message Joined: 9 Jul 00 Posts: 51469 Credit: 1,018,363,574 RAC: 1,004 |
I had one, Intel replaced it post haste, no BS involved. As I recall, Intel issued a program to test for the fault, and gave a replacement with no other strings attached. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
KWSN THE Holy Hand Grenade! Send message Joined: 20 Dec 05 Posts: 3187 Credit: 57,163,290 RAC: 0 |
I had one, Intel replaced it post haste, no BS involved. As I recall, Intel issued a program to test for the fault, and gave a replacement with no other strings attached. It may depend on where you were living at the time: I was then (and still am...) living about 50 miles north of Sunnyvale, CA; (Intel HQ) so I heard all the above of Intel's initial response to the bug. (I was also an IT professional at the time!) If you lived farther away from Silicon Valley (and weren't in IT) then the initial response may have passed before you heard about it... It only took about a week for Intel to reverse field and start replacing the buggy chips... . Hello, from Albany, CA!... |
Francois Piednoel Send message Joined: 14 Jun 00 Posts: 898 Credit: 5,969,361 RAC: 0 |
I had one, Intel replaced it post haste, no BS involved. As I recall, Intel issued a program to test for the fault, and gave a replacement with no other strings attached. You "heard" ... that's your problem! I heard that some time , chicken have teeth ... did you know? We all understood, Intel is very very very bad, because we had a bug one time in our processor. Happy? who? |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.