Bugs in Conroe

[Trickster63]

Just seen this link

http://download.intel.com/design/processor/specupdt/31327902.pdf

And apparently one of these could be very serious with regard data loss

 

[Justintime]

All cpus have some errata, generally they try to weed those major ones out with new revisions etc..

 

[lay-z-boy]

cough*xs*cough

all cpu's have errata - fact

 

[NathanE]

As said, all CPU have eratta. Some of the late P4's had probably 5x the amount that Conroe has.

Although some of the wording in that document looks severe. You must remember it is written by some engineer and is intended for the consumption of CPU engineers.

 

[Servo]

Lol, I'm happy I bought a e6300, that ain't mentioned

 

[nicRob]

E6000 sequence = E6300/6400/6600/6700

 

[Servo]

Oh, you bust my bubble

 

[stoofa]

What you need to remember with Intel is that they were stung once before and it won't happen again.
If anybody remembers the infamous Intel P90 floating point problem where a number divided by itself was giving a result of 0 rather than 1.

Initially Intel denied the error existed.
Then they told everyone it really wasn't a major issue.
Finally they admitted it was, and a recall was issued.
Intel then went on to actually do one of the better and well organised recalls - so OK they couldn't contact buyers direct, but when you contacted Intel be it with an OEM or Retail CPU they helped you get a replacement.

Intel first of all got a lot of negative press and then, once the recall had been actioned got much better press.
Because of all of that if ever Intel do make the mistake of rolling out a CPU with that level of problem again they will recall as the money lost in such action is tiny compared to the negative press they otherwise receive.

That document might sound horrible, however I think it's safe to say there really isn't anything too serious on that list or else we'd have already heard about the recall.

 

[Noto]

is this the reason for the new stepping in Conroe chips?

 

[fornowagain]

http://www.theinquirer.net/default.aspx?article=33942

 

[Justintime]

What you need to remember with Intel is that they were stung once before and it won't happen again.
If anybody remembers the infamous Intel P90 floating point problem where a number divided by itself was giving a result of 0 rather than 1.

Initially Intel denied the error existed.
Then they told everyone it really wasn't a major issue.
Finally they admitted it was, and a recall was issued.
Intel then went on to actually do one of the better and well organised recalls - so OK they couldn't contact buyers direct, but when you contacted Intel be it with an OEM or Retail CPU they helped you get a replacement.

Intel first of all got a lot of negative press and then, once the recall had been actioned got much better press.
Because of all of that if ever Intel do make the mistake of rolling out a CPU with that level of problem again they will recall as the money lost in such action is tiny compared to the negative press they otherwise receive.

That document might sound horrible, however I think it's safe to say there really isn't anything too serious on that list or else we'd have already heard about the recall.

Intel has had some shady goings on, the original floating point bug was an errata that got too much publicity, but remember the 1.13GHz PIII and the i820 fiasco? those got out of hand and did them some fair damage, VIA got more market share and so did AMD.

 

[stoofa]

Agreed but again Intel recalled the 820 chipset & SDRAM combination.
They also went above and beyond - they were taking back any make of 820 based motherboard with up to 256mb of SDRAM and in return you were sent an Intel branded 820 motherboard and 256mb RDRAM.

I'm not saying Intel are faultless - far from it.
However they are not scared to recall anything serious and unfixable and do their very best to solve the problems.

 

[FrankJH]

What you need to remember with Intel is that they were stung once before and it won't happen again.
If anybody remembers the infamous Intel P90 floating point problem where a number divided by itself was giving a result of 0 rather than 1.

Initially Intel denied the error existed.
Then they told everyone it really wasn't a major issue.
Finally they admitted it was, and a recall was issued.
Intel then went on to actually do one of the better and well organised recalls - so OK they couldn't contact buyers direct, but when you contacted Intel be it with an OEM or Retail CPU they helped you get a replacement.

Intel first of all got a lot of negative press and then, once the recall had been actioned got much better press.
Because of all of that if ever Intel do make the mistake of rolling out a CPU with that level of problem again they will recall as the money lost in such action is tiny compared to the negative press they otherwise receive.

That document might sound horrible, however I think it's safe to say there really isn't anything too serious on that list or else we'd have already heard about the recall.

You are showing your age

I remember this on the BBC news

 

[Corasik]

Wasnt it P60/P66. I actually had one of these, and got a letter from Intel Swindon requesting I returned it for free RMA.

P60/66 were 5V socket 4 processors. Later P4's were 3.3V die shrink chips. Pretty sure they had fixed the FDIV bug by then, although apparently there was another bug that popped up, but didnt get as widely publicised as there was a bios/microcode update that masked the problem.

 

[whammer]

Corasik, you are correct it was the P60/66 chip that had the FDIV error. I remember working on a system for my brother many years ago, and I enquired about a replacement CPU for a laugh. They were prepared to swap it out no questions asked!
God, I'm starting to feel old now.

Regards,

Whammer

 

[Justintime]

Pentium 60, 66, 75, 90 or 100MHz can be affected, i'm 99% sure i saw that on Intels site at one point over the years.

 

[Corasik]

Perhaps they had already fixed the problem by the time I upgraded my recalled P66. I went straight from a 66mhz Socket 4 system, to a 120mhz Socket 7.

Just been looking at Wikipedia and it seems there was a socket 5 version of Pentium for the 75-120mhz chips. So perhaps only socket 4 and 5 Pentiums were affected.

Either way, errata in processors is common, there is a almost equally long list of errata in the AMD64 processors. The majority of problems can be masked out with microcode updates stored in bios, but occasionally like the poor old Pentium, the only option is a full scale recall.

Sometimes the worse errata are 'fixed' between steppings.

 

[NathanE]

I had a Pentium 90MHz and it definately didn't have that bug. Only the first few batches were affected IIRC.

IIRC the P60's also suffered from overheating issues.

 

[stoofa]

It just goes to show that the issue wasn't actually that big as it wasn't until the release of the P90 that anybody started demanding replacement CPU's!
The P60/P66 were the very first Pentium CPU's - ran at 5v rather than 3.3v

It was a while ago however as far as I remember all P60/P66 CPU's were initially effected (obviously once you'd swapped yours over then you were added a non-effected one to the field).
First release P90s and P100s also had the issue and that was when all the fuss kicked off.
As far as I was aware the problem never stretched any further than the P100.

Anyway - we're going nicely off topic
Intel may still have errata in their CPU's and chipsets, however they have now learnt that it is better to come clean on major issues and get a recall issued.

 

[GAMEfreak]

i guess powerPC chips must be less erroneous? after all isnt that what they use for the number crunching part of scientific research? i know they use it for consoles now too like the gekko in the gamecube and the Broadway ( Wii processor) xenon (x360) and cell BE ( ps3) , and as such they can't really issue patches for such errors unless the user has a connection to x-box live or the likes, which isnt exactly common in comparison to the size of the userbase.

also, has a console every been recalled because of CPU level errors?