Numerous Bugs Reported In Intel Prescott Pentium 4
A day after Intel released five LGA 775 chips, the firm reported a list of bugs in its Pentium 4 Prescott produced at 90 nanometres.
Source: The Inquirer
Go ahead Thrax, give it your all on this one. -KFThe report, dated 22nd of June 2004, show some interesting examples of erratanotbugs. For example, R9 is described as "System bus interrupt messages without data which receive a hard failure response may hang the processor". So far no workaround is available for this bug. R16 describes a system hanging as a result of a fatal cache error, while R19 describes a parity error in the L1 cache which could cause the processor to hang. There's no workaround for this one yet, in this stepping.
Source: The Inquirer
0
Comments
KF
1) They had to recall the P3 1.13.
2) The i850 chipset quadrupled the AGP delay with the presence of a PCI card.
3) The wiliamette was a total failure, in fact, the time between the Wiliamette and the Northwood was the fastest revision in Intel history.
4) The Northwood never could do the boolean OR process. Props to Intel's ****ty FPUs.
5) The LGA775 sockets fail after 10-12 CPU insertions
6) They just started recalling the ICH6RW do to a fab glitch
7) The prescott is slower than its predecessor (Newer = faster? Hello? Intel?)
8) Admitting flaws in their abysmal engineering, Jayhawk and Tejas are sacked at the expense of several hundred employees.
9) Failing in their drive to avoid killing the Itanium, Intel slates Nocona for release with x86-64 extensions.
10) Failing to make a CPU perform as its clockspeed is rated, they switch to performance ratings that don't even make sense.
11) Preliminary performance reports state that EMT64 technology is slower than anybody could have ever imagined. Leave it to Intel to make x86-64 suck.
12) And now, bugs in the prescott!
If we thought 20 stages in a pipeline was deep, jesus, the prescott is the Marianas trench of pipelines.
If that weren't enough, they added hyperthreading back on the Northwood <i>not</i> because it was a good performance addition.. But because the pipeline was so deep and unwieldy that Intel was scrabbling for anything it possibly could to keep that pipeline filled.
See, when you have 20-30 stages in a pipeline, it's harder than hell for a computer to keep it filled and operating at nominal efficiency. And a glitch just means you lost even MORE cycles than you already are by virtue of it being a damn long pipeline.
So rather than admit defeat and go all out on Banias-based offerings, they decided to add hyperthreading which would keep their pipeline filled by executing multiple threads.
It wasn't for additional performance. It was so their CPU could have a chance of performing at all.
Bastards.
<b>BRING ON THE DOTHANS</b>
KingFish