Sign in with
Sign up | Sign in
Your question

AMD's response -- it's not a CPU bug

Tags:
Last response: in CPUs
Share
January 24, 2002 8:59:44 PM

<A HREF="http://www.geocrawler.com/lists/3/Linux/35/175/7626960/" target="_new">http://www.geocrawler.com/lists/3/Linux/35/175/7626960/...;/A>

Apparently the "Athlon AGP bug" is not in the Athlon CPUs--it's a bug tied to the VM in Win2K and Linux and the way they handle caching of AGP memory. The only reason this doesn't b0rk on Intel CPUs is because Intel CPUs aren't capable of speculative writes, so the software fallacy has been allowed to slide for a long time.

<i>If a server crashes in a server farm and no one pings it, does it still cost four figures to fix?

More about : amd response cpu bug

January 24, 2002 11:36:37 PM

Quote:

The only reason this doesn't b0rk on Intel CPUs is because Intel CPUs aren't capable of speculative writes, so the software fallacy has been allowed to slide for a long time.

Interesting, no flame bait intended, but would this be considered an Intel bug or a feature Intel processsors lack? What exactly are speculative writes? Sounds like something to do with the cache. Reminds me of hardware data prefetch. I could of course be completely wrong as I am guessing and I'm not 100% familiar with all the features in Intel and AMD processors.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
Related resources
January 25, 2002 12:33:02 AM

Quote:
Interesting, no flame bait intended, but would this be considered an Intel bug or a feature Intel processsors lack?

Just a feature Intel lacks, not an Intel bug. Speculative writing is a performance enhancer.

Quote:
What exactly are speculative writes? Sounds like something to do with the cache.

It <i>is</i> something to do with the cache, but I don't know exactly what (yet). Probably a feature to predict what memory areas might be written soon and keep the cache up to speed on that. I'm guessing it promotes more efficient cache usage.

<i>If a server crashes in a server farm and no one pings it, does it still cost four figures to fix?
January 25, 2002 12:44:27 AM

so who's fault is it that it causing a problem

I spilled coffee all over my wife's nighty... ...serves me right for wearing it?!?
January 25, 2002 12:57:13 AM

The Linux kernel developers probably take most of the blame. Same goes for M$ developers. Somehow, these developers thought they could treat AGP memory space like any other memory space as far as paging went.

AMD's fault in this is that they did not notice the developers' troubles and inform them of the real problem. Considering that the problem was purely software, outside of AMD's scope, that's not much of a fault.

<i>If a server crashes in a server farm and no one pings it, does it still cost four figures to fix?
January 25, 2002 5:34:47 AM

::Sighs::

I bet meltdown is in his room crying now....poor guy.

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
January 25, 2002 6:10:25 AM

he'll be ok. he probably wants the new 2.2Ghz. i guess the guy has to be happy about something.

i just remembered a while back i sold a geforce 2ultra to a lady thinking that was the problem. she came back and told me about this problem. geee its been over a year.

<font color=purple><b>Let us respect each other, so we can all be friends.</b> :wink: </font color=purple>
January 25, 2002 1:57:01 PM

"::Sighs::

I bet meltdown is in his room crying now....poor guy."

nope,...if it were'nt for me you mongrels would not know about this! and if you did you surely wouldn't post it, would you?

I posted a link to slashdot's article, that's all I did and you mongrel started crying, LOL!

also slashdot has their own forum with even more AMD ppl complaining about AMD platform issues.

"<b>AMD/VIA!</b>...you are <i>still</i> the weakest link, good bye!"
January 25, 2002 2:49:43 PM

::signs at AMDMeltdown and Matisaro::

Ahh, when are you guys going to grow up?

Matisaro: I know you were just joking around but trolls don't have a very good sense of humour and will attack whoever attacks them, ignoring them is the better solution, I learnt that the hard way.

AMDMeltdown: If you don't have anything good to say, then don't say a thing. And please stop calling HUMAN BEINGS mongrels!

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
January 26, 2002 3:38:24 AM

Quote:
Matisaro: I know you were just joking around but trolls don't have a very good sense of humour and will attack whoever attacks them, ignoring them is the better solution, I learnt that the hard way.



I ignore him most of the time, joking about him once in a while is good, same with fugger.

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
Anonymous
a b à CPUs
January 26, 2002 4:10:55 AM

So how does one actually go about disabling the 4MB pages? Can someone walk me through it? I'm using Windows 2000. Thanks.

<P ID="edit"><FONT SIZE=-1><EM>Edited by killian on 01/26/02 01:12 AM.</EM></FONT></P>
January 26, 2002 4:49:34 AM

It is a patch, It probably is on the ms update site(unless its already installed.)

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
January 26, 2002 11:23:27 AM

It's a reghack on AMD's site. It's been out for over a year now...it just recently got discovered that it affects Linux as well.

For Linux, you just have to pass "mem=nopentium" to the kernel via your boot loader. Linux will probably have a <i>real</i> fix for the problem soon enough.

<i>If a server crashes in a server farm and no one pings it, does it still cost four figures to fix?
January 26, 2002 12:05:35 PM

From the articles being released today and yesterday, it seems it is deffinatly NOT a cpu bug, but more an os bug because of inproper agp memory usage coding.

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
!