Slashdot Mirror
1.13GHz Pentium3 Processors Unstable? Answer:Yes
brennan73 writes "Last week, Tom's Hardware wrote a very critical review of the Pentium3 1.13GHz, specifically mentioning serious stability issues. Today, HardOCP wrote an article of their own backing Tom up and describing their own experiences; they even got Anand in on it. Probably the most interesting thing to me is that none of the tested sample CPUs was able to run Tom's Linux kernel compilation test. After Rambus, this is the last thing Intel needs. Are they shooting themselves in the foot AGAIN?". Update by: HeUnique : Tom has updated the story, so I recommend to read the update.Update: 08/28 06:44 PM by H :Check out the latest burst from ZDNet - it appears that Intel will be recalling the Pentium3 1.13 Ghz. Thanks Evan.
It seems that Intel thinks the only thing we care about is raw speed. Trying to one-up AMD is simply going to get them into a hole in terms of public perception. I'm surprised that a company as big and diverse as Intel would stake their reputation on a speed pissing-contest.
-This sig intentionally left blank
They announce a product before it's completely designed. They "release" an unstable, unworkable version 1.0 of something, just to get their name out on a press-release. All to divert attention from a competitors product and get some cash flowing in to fund the development of version 2.0 (or 3.1) which is the actual product.
The competitor releases their offering later, but it's of higher quality. Still, the vapour has already dulled people's minds to the competitor's efforts and value.
Techies know better, but how many home-users and managers do? We have to make it clear to the business decision makers and the casual users, that this is not about Intel being technologically superior or better able to deliver on schedule than AMD or Transmeta or anyone else - it's marketting and it needs to be stated as such.
Not that I think Transmeta is a God-send, but let's let numbers and solid products do the talking. A 1.3Ghz CPU from Intel, even though it's been demoed, is still vapour-ware, until it's solid and readily available.
The REAL jabber has the /. user id: 13196
The REAL jabber has the user id: 13196
What you do today will cost you a day of your life
Yeah there are quite a few UK magazines and websites citing similar problems and even DOAs with the 1.13ghz chips.
That isn't to mention the severe shortage of the actual chips. UK Mags dont even get their own chip, they have it for a few days and then instructions on who to post it to next.
I have a sneaking suspicion that this is just an overclocked 1ghz chip, and they've picked the best few to ship at 1.13 in a desperate attempt to one-up AMD.
However AMD today announced real tangible availability of their 1.1ghz ath's and Compaq, Big Blue, HP, Gateway and Fujitsu Siemens are amonst the big names about to ship this in systems.
Looks damn good for AMD
This actually makes me nervous about all the Gig chips being built. What kind of testing is really going on with these chips and is it adequate enough or are the manufacturers so eager to get another press release out in order to get a boost to their stock they are failing to realize their design is flawed?
Intel has been suffering some serious blows, both in the technical world and in the business world and cannot afford another setback, especially since these new gig plus chips are supposed to be selling for the holiday season (as gamers are one of the first to eat up such advancements).
If I had any advice for Intel, it would be to retrench and spew out tons of 700+ chips at low prices to build up marketshare while focusing on turning the Gig plus chips into something worthwhile. This is just an off the cuff suggestion as IANA-MBA.
This is not the way to build a lasting empire.
The odd thing is that we're seeing stuff go wrong now with corporate capitalism what we saw go wrong with communism.
These huge corporations keep getting bigger and bigger, allowing themselves a stranglehold on the industry, and in doing so, they become hugely beauracratic, heirarchical, and conservative.
So some horrible design issue is found in one of Intel's products, something that would guarantee failure for a smaller company, and what happens? Intel denies a few allegations, issues a few workarounds to Microsoft, and hires a few new spin doctors to make sure everything works okay.
The computer industry is just catching on to this. The oil and tobacco industries have been doing it for years. Microsoft shows an uncanny brilliance for turning a bumbling mistake into a "feature." But at least, unlike Shell, they're covering up system crashes and not genocide.
It took Communism around 80 years to become so big and unwieldy that it collapsed under it's own weight. After 114 years of corporate rule (SANTA CLARA COUNTY v. SOUTHERN PACIFIC RAILROAD COMPANY ), it seems that the incredible arrogance of corporate capitalism is putting it on the same route.
Michael Chisari
mchisari@usa.net
In related news, Intel Corportaion recently purchased "Tom's Hardware Guide" and "HardOCP" for a record 1 billion dollars. Intel's spokesperson had these comments early this morning:
"We really like their [Tom's Hardware] input, and we wanted everyone in Newfoundland to be able to recieve the bounty of his wisdom. That's why we've purchased Tom's organization, and we're strategically relocating Tom to Newfoundland, where he will perform a streaming audio broadcast to everyone within earshot in his new deep, dark, hole in the ground."
Further updates as events warrant. Rambus Corporation (Nas: RAMBUS) had no comment, but their stock spiked 30 points on the news.
--
What happens when you outlaw guns
I thought I read this two weeks ago... Tom trashed Intel, then contacted Intel to report a possible flaky sample. Inappropriate and hardly objective behavior. It's worth noting that Dell is claiming to offer the 1.13 Ghz systems.
Vote Naked 2000
A feeling of having made the same mistake before: Deja Foobar
Disclaimer: I have AMD and Intel CPU's at home. I'm agnostic when it comes to processors, give me speed and stability and I don't care who it is from!
p u/thunderbird_1-1ghz/
Check out the review of the 1.1 ghz Athlon at
http://www.sharkyextreme.com/hardware/reviews/c
Good marks, great performance. However, look at this blurb from the review about the high return rates of high speed Athlons:
"We haven't had any more AMD CPUs fry on us, but we did talk to a mail order house that said some disturbing things. They told us that they have had approximately a 10% return rate on AMD Athlon Thunderbird CPUs. That is a painfully high failure rate. They said that some distributors place bar codes on the underside of the Athlon Thunderbird CPUs and that these barcoded CPUs often come back with their barcodes burnt. Their Duron failure rate, however, has fallen to normal and acceptable levels."
Could it be that both Intel and AMD are too busy racing each other to take proper care in the Quality Control arena?
Also, it's a software problem
</FUD>
icqqm [ICQ:11952102]
"We asked for 6 months to work on this, and you gave us one month. Right now, all it does is erase your disk drive. If you're on a network, it erases everyone's disk drives. And god forbid if you have a modem, it calls up your friends and erases their disk drives."
"We'll call it 'Quick Protect'"
"If you have a soundcard, it swears at you."
I wish to make a call to the hardware designers to stop releasing alpha stage hardware, and stop mis-labeling them with "1.0" version numbers. For instance, the Linux Kernel 2.2.14 should really be 0.6.02.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
I found out about this from Tom's review of the Athlon 1.1ghz, here is the key quote I forwarded to a few friends:
l on-02.html
It turned out that the kernel compilation failed on all three Pentium III
1.13 GHz samples in Kyle's lab, right in front of the eyes of Intel's
engineer. My CPU happened to be the flakiest of the three, as it failed in
other tests on different platforms as well.
http://www.tomshardware.com/cpu/00q3/000828/ath
AFAIK, they aren't really behind at all, because its a faster cpu clock for clock.
So sure, you may only have 500mhz, but if you can do twice as much with 1mhz then an Intel cpu can, then really there is actually no difference except that one sounds nicer on a box.
I don't know what the exact numbers are, but the situation is something like that.
Unless of course you look at RC5 benchmarks, the Altivec PPC's just destroy anything Intel can make.
-- "So they told me that using the download page to download something was not something they anticipated." - Bill Gates
While I don't discount the numerous problems that people have had with the 1.13ghz PIII, let's not pretend that Tom isn't strongly biased against Intel. He always has been, which in my book casts doubt on his credibility.
That being said, there is obviously a problem with the 1.13ghz part. I would not just take Tom's word on it though. The other articles/reviews mentioned add credence to this.
Difference is that the 500Mhz G4 processor is a completely different architecture.
Are you going to tell me a 1Ghz x86 is faster than a 750Mhz Alpha?
Clock speed doesn't matter near as much when comparing architectures. You can't say "This is faster cause its 1Ghz"
From Tom's....
:)
Intel's Next Paper Release -
The Pentium III at 1133 MHz
[an error occurred while processing this directive]
you may be using AMD chips and not even know about it. Since your chips come from 1981-83, then at least most of them came from the time when AMD was liscensed to make exact replicas of Intel chips (As was required from IBM when they chose Intel chips for the IBM PC.)
also, I do remember back in the 386-486 days getting PCs that just stated that they had a x86 processor, and opening them up and finding out that I got a amd processor. Back then the brand didn't matter. Intel made that a issue when they started advertising in 1994
I agree that this article taken alone might be considered bias, but part of his job is opinion based on numerical facts.
He has proof which he (and after reading all the articles I also) believe in. He has historically not shown a preference to Intel or AMD, but has been wary of Intel's pressure to churn out the "best" chip.
What you are describing is opinion, which he bases on fact (his tests). Compiling a Linux Kernel is a good way to test a system and if that Sig11's then there is a hardware problem. People read his articles because they respect his opinion and his tests.
Please don't confuse this with bias (like slashdot has against Microsoft), which is arbitrary and not always based on specific fact.
News for UW students
its pretty well known that if you run SMP on a bx-board, you better cool the hell out of that bx chip. and even this won't stop the inevitable lockup that will occur days or weeks after the last reboot. its a fact. and if your SMP bx system isn't locking up, you're not loading it down enough. try 4 instances of seti@home (2 per cpu) and wait a few weeks. it will lockup.
given this track record, I don't doubt that the ghz p3 chip will also be unstable.
--
--
"It is now safe to switch off your computer."
the above links on Tom's Hardware were published a month ago. For an article published today on thje AMD 1.1 Ghz, go here
The issue is that heatsinks for PIII/Celeron and Thunderbird/Duron arn't interchangeable due to different physical dimensions of the chips. If you use an AMD approved heatsink or one specifically designed for T-Bird/Duron (e.g. there is a special Golden Orb version for them) then it'll be fine, but if you use one made for an Intel CPU then you may well damage the CPU.
Agreed. I've read both and a few other articles on Tom's site and he has reminded me how important it is to take all things with a grain of salt. Good, bad or ugly.
Vote Naked 2000
A feeling of having made the same mistake before: Deja Foobar
As long as AMD (or someone else) won't produce an SMP chipset, there's a definite need for Intel processors. I'd rather have two or four slower processors than one blindingly fast.
...as long as they provide a way for me to prove that 2 plus 2 is still equal to 5.
Send your friends messages of love at fuck-you.org
[I'd look this up, but I'm too lazy, atm] the key here is factors like Instructions Per Cycle, speed of Write and Fetch and a few other real measures of CPU horsepower, as well as mating that CPU to a good motherboard architecture. The end user benchmarks are what consumers should pay attention to, and something Apple and publications make available, regarding time to do image operations, spreadsheet calcuations, spell check documents, etc. For gamers, how many frames per second of Quake III or some other game.
The burden for Wintel systems is having to tune the hardware and OS to work the best. Not something a lot of cheap clone makers do well. You may buy a 1Ghz PIII, but if it's thrown together it may run like a well tuned 200MHz PII.
Vote Naked 2000
A feeling of having made the same mistake before: Deja Foobar
I wonder if the problem with Intel right now is simply that they're trying too hard.
Maybe they're trying to stay ahead of AMD, and thats causing them to do things more quickly then they should, resulting in mistakes. It probably can't be good for morale over there to see that seemingly out of nowhere AMD has come in and almost taken over the high end single cpu market.
So the guys at Intel simply want to be better then AMD, but they don't have time to do what that would actually require (some pretty major architechture changes), and so we end up with stuff like this P3, since they have to release *something*.
I bet if management were to walk down there and say "ok, we're not going to release anything new for the next few months, I want you guys to take as much time as you need to redesign things so we're on top again", they could probably do it.
Of course, they aren't being given a mandate like that.
-- "So they told me that using the download page to download something was not something they anticipated." - Bill Gates
I think this is a huge blow to Intel. Speeding up big, hairy compiles (like the Linux kernel) is one of the few reasons I could think of for getting a 1GHz+ CPU. If it can't do that what good is it?
Molog
So Linus, what are we doing tonight?
So Linus, what are we going to do tonight?
The same thing we do every night Tux. Try to take over the world!
I'm not going to get in-depth on this, but the primary issue with Windows is that it usually comes installed in Jack-of-all-Trades manner. For those familiar with the end use of a PC or workstation a custom install can be performed to omit certain features and designate how resources will be used (memory, caching, paging, etc.) The less tasks the OS has to manage, the more efficiently it usually accomplishes each. (My PC at work is usually paging (swapping memory to disk) with only a couple small applications running (but I didn't set this beast up.) Problem is, most users just assume when things get slow that they need more power. Often better drivers become available for controllers or other devices which can reduce memory usage or run faster. The typical end user doesn't know and usually can't be bothered with such details.
Vote Naked 2000
A feeling of having made the same mistake before: Deja Foobar
AMD CPU's require VERY GOOD cooling, and many people would try to use either cheap heatsink/fans or ones meant for celeron or FCPGA P3. These are NOT interchangeable generally. In some cases contact will not be made with the chip (e.g. some Alpha coolers have small feet). AMD specs require higher contact force to be applied. Without heatsink/fan, the 1GHz cpu will burn in 8 seconds (source : AMD). There really is no chance to see this coming (like, system hanging or failing to boot); if you screw it up, it will be dead right there and then.
Second problem is that cpu core is exposed and extremely fragile. Couple that with very hard to clip on heatsinks (see contact force above), and you get many cases of cracked core. I did it myself, luckily it was just a small scratch and it still works. The biggest offender was the very popular Golden Orb.
AMD chips runs very hot. VERY hot. So cool them well and they will serve you well.
It has split 3 times over a 3 year period. And the stock is doing extremely well right now (and for the past couple of years). :-)
- sigs are for wimps.
"Satan" Clara? Grow up, Tom. That's just as juvenile as referring to Microsoft as M(dollar-sign).
Commenting only on this article, and not the follow-up, let's review a couple of points:
1)Tom didn't test on the Motherboard he was supposed to test on
2)He didn't have the required microcode update
3)Tom DOES NOT receive ANY MONEY OR ENDORSEMENTS from ANY competing hardware manufacturer, ESPECIALLY NOT AMD
What kind of excuse is it for Tom to say he doesn't believe in Microcode updates because they supposedly disable certain functionality?
Now, let's review a couple of points from the follow-up article
1)Tom glosses over reports of other people doing well with their testing, and gives well overdone attention to problem cases
2)Tom claims that Intel COULD BE shipping faulty chips to OEMs with little factual basis, then treats his claim as a FACT for the rest of his article
3)Tom heroically posts his "scathing" letter to Intel PR, but oddly enough "forgets" to post Intel's reply. Perhaps he considered Intel's reply long-winded, but after the two previous full-text quotes, I somehow doubt it could possibly be any worse.
4)Tom seems to think himself somehow godlike in his heroic letter, saying that Intel couldn't possibly disagree with the opinions of the great Thomas Pabst.
I used to respect Tom's Hardware, but have you ever noticed that whenever there are banner ads for one product on his site, the competition mysteriously receives bad reviews?
This is a manual virus. Copy it to your sig and help me spread!
... real world tests. Compiling a Linux Kernel is not a bogus WinBench / WinTune / WindTunnel / Optimized-for-Intel-because-they-paid-us benchmark tool, it's a real-world CPU stresser. Going through the hundreds of thousands of lines of code is a tough job for the CPU, and therefore is what I consider a valid test.
Thanks Tom!
Dimension XPS B used to support 1.13GHz...
i men_desktops.htm
_ dim_b.htm
http://www.dell.com/us/en/dhs/products/series_d
But 1.13GHz choice has now been pulled:
http://www.dell.com/html/us/segments/dhs/choose
Remember that Tom's was the first one to speak out against Rambus memory when everyone else was claiming how great it is. Now even Intel have vindicated Tom's Hardware, and are moving from RDRAM to DDR.
Looks like Tom's may have been the first to point out the PIII 1.13 GHz problems too. I think Tom's anti-Intel reputation is just because everyone else has an unrealistic (i.e. out of date) idea about how good Intel are, while Tom's isn't afraid to call it how they see it.
>Here's the thing I don't really understand (well, I understand it, but I don't accept it): Most programs and games don't need that fast a
>processor, so why are people buying it? The need for speed?
'Cause they're told they need it.
A recent issue of _Consumer_Reports_ stated that a new buyer should expect a minimum speed of 1 GHz in her/his new computer (unless it's used for non-CPU intensive tasks like word processing & Email).
Sheesh, I've been using computers with an average speed in the 400-600 MHz range (well, my home computer has two 400 MHz), & I have seen no problems with any of those being ``too slow". (For once, Andy is ahead of either Bill or Linus.)
The only reason I can think that _Consumer_Reports_ would state that is because the OEMs' PR flacks talked them into it.
Geoff
I think I see a trend here. Maybe for them it really would be easier to muzzle the entire internet than to produce p
There are lots of things that the CPU can do wrong that won't cause a crash. Floating point errors, for example. MMX problems. Trouble in 3D extensions. Those are features that generally don't affect the control path in system software. You need to run diagnostics, 3D renders that check that the output matches a standard image bit for bit, and such. For that matter, when they run the "Linux recompilation test", do they check that the generated code matches some known-good reference? Probably not. Like the overclocking types, if the whole system doesn't crash, they're probably happy; all they wanted was to see the CPU benchmark run anyway.
However, from what they wrote, it looks like Intel reached 1.13MHz by narrowing the temperature margins, not by actually making it faster. That's a desperation move.
I think AMD will keep Intel in their current tizzy right up until they deliver their death stroke in the form of sledgehammer. Intel will again be taken completely by surprise as they were focussing on gaining their bragging rights back in the 32 bit world, which will be inconsequential in just a few years. I don't expect they'll last much longer after AMD catches them flat footed in the 64 bit arena.
Of course, Intel's been doing a lot of stuff right with Itanium. They've been making sure that a lot of OSes will run on the chips before the chips are even out the door, and they're making sure heavily optimizing compilers are available for the chips. AMD could still screw it up and become the technological almost-ran in the 64 bit arena. And Intel still has a lot of influence in the motherboard/chipset manufacturers. It doesn't matter how technologically great your chip is if there isn't a motherboard you can put it in (Witness the continuing lack of Athlon SMP motherboards.)
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
Last week, the fastest PIII speed on Dell was 800MHz. This week, it's "NEW! 1GHz." That seems to reflect what you can actually get that works, as opposed to selected samples shipped to journalists.
I'm running Windows 2000 on a K6-2 with 64MB of RAM, and it runs a lot faster than Win 98 did.
Yes - because Windows 2000 has better page memory management on small memory systems. Shame that it is less fast for more memory systems - check MS's own benchmarks. 128MB is where you start to lose out and it gets worse above there relative to Win NT 4.0.
Cheers,
Toby Haynes
Anything I post is strictly my own thoughts and doesn't necessarily have anything to do with the opinions of IBM.
I'm not sure what you're looking at.
Try this link (same as I posted above).
Select Dimension XPS B Processor
If you can find a place on Dell's site that lets you select a 1.13GHz processor, then please post the link!
I hate to whore onto an early comment, but the full ZDNet story is available now.
--
sig:
sig:
See the "..for smart people" banners Wired runs here? Look elsewhere guys.
Toms opinion hasn't changed, depending on how you look at it, he is either biased, or seriously Pissed Off at Intel, and it does come through in what he writes.
Of course, this time apparently he was right.
I wouldn't be surprised if the Hard OCP article had something to do with this, apparently they had an Intel Engineer there to watch as the Kernel failed to compile rather dramatically.
You know, he probably went back to Intel HQ and mentioned this little problem.
Of course, it could be that Dell simply couldn't find a way to make a system using these things that was at all stable, and they complained to Intel quietly about it.
-- "So they told me that using the download page to download something was not something they anticipated." - Bill Gates
Instead of people admitting that Tom was right in his previous review, the same people that claimed the supposed instability was just a product of Tom's bias now are attacking him for bias again.
Well, Intel has recalled the chips. Could it be then that the sites that did not discover the instability have a pro-Intel bias, instead, and that Tom was just calling it as it was?
Oh, sure, Tom has a bias. Guess what -- so do a lot of people criticizing MS software. A bias doesn't mean you're not telling the truth.
Intel's chips had flaws serious enough for Intel to recall them. Tom was right, and those who accused him of distorting the truth owe him an apology.
Steven E. Ehrbar
I guess dates in the update section don't get changed for my timezone like dates on stories and comments?
I'm a leaf on the wind. Watch how I soar.
The 604e had a huge aluminum heatsink but no fan- it ran fairly hot for a PPC chip, being a 166mhz 604e hopped up to 200mhz by Daystar Digital. After a while, I got a 300 mhz G3 card in another upgrade (still have the 604e card, too). The G3s are said to be the next-generation equivalents of the old 603es- power-efficient, higher-Mhz, that sort of thing. Mine ended up roughly twice as fast as the 604e for most tasks, most easily measured by POV renders which take a good long time- also, FPS framerates at least doubled, when Q3test came out I was unable to try it as the 200Mhz 604e made it a slideshow. The 300Mhz G3 started to make things vaguely playable.
The G4s apparently are to the G3s as my old 604e was to the 603s. Be afraid. Be very afraid. That, or gloat a lot if you have one ;) lord knows I am tempted to waste a bunch of money and spring for a G4, even though not a thing that I do really requires one.
One factor that should be considered is how much code is tailored to the weaknesses of x86. As things develop, this should even out- apart from a tendency in the high-end x86 world to turn to increasingly byzantine compilers in desperate attempts to make these complicated mechanisms work :) such things have little appeal to _me_, though, and I enjoy running a (admittedly 300Mhz-slow) computer that uses comparatively simple and straightforward compilers and happily spins its tiny wheels for me, cooled only by a purple heatsink the size of an ornate postage stamp. Small is fscking beautiful, I say. :)