Intel i850 chipset bug/flaw?

A place to talk about whatever Scope music/gear related stuff you want.

Moderators: valis, garyb

Post Reply
sinix
Posts: 198
Joined: Wed Sep 05, 2001 4:00 pm

Post by sinix »

I've been reading a very interesting thread on the recording.org board about a flaw in the Intel i850 boards.

Full thread (you need to register 1st):

http://www.recording.org/cgi-local/ubb/ ... =000221&p=


To quote some:

"5. Sustained PCI Bandwidth
Problem: During a memory read multiple operation, a PCI master will read more than one complete cache
line from memory. In this situation, the MCH pre-fetches information from memory in order to
provide optimal performance. However, the MCH cannot provide information to the PCI master
fast enough. Therefore, the Intel® 82801BA terminates the read cycle early to free up the PCI bus
for other PCI masters to claim.
Implication: The early termination limits the maximum bandwidth to ~90 MB/s.
Workaround: None
Status: There are no plans to fix this erratum.
6. RDRAM* Device Interface Initiate Initialization Operation (IIO) Bit
Problem: The Initiate Initialization Operation (IIO) bit in the RDRAM device initialization control
management (RICM) register may be cleared by the MCH too early.
Implication: If the MCH clears the bit too early and BIOS immediately issues a new initialization op code
(IOP), the MCH may incorrectly control the RDRAM device CMD signal causing an invalid cycle
to write to the memory subsystem, and the IIO bit may not be cleared. This results in a system hang
during the memory initialization process.
Workaround: BIOS must allow at least 1 µs delay between executing initialization opcodes (IOP).
Status: There are no plans to fix this erratum.

Note that in each case the last line reads "There are no plans to fix this erratum." The similar Intel doc for the i845D chipset has NO ISSUES which are not fixed with latest BIOS update. This was enough to steer me away from i850 chipset for an audio machine - if I ever DO manage to bang on that limit, I want it GONE..."

-----

Now I was completely sold on the Asus p4-t boards, it seems like everyone has really good success with these. How does this chipset flaw effect those of us using large amounts of PCI bandwidth for dsp cards?

I wasn't clear if this bug relates more for large disk transfers or just shuffling large amounts of data across the PCI bus in general?

Can anyone shed some light on this?

- sinix
subhuman
Posts: 2573
Joined: Thu Mar 29, 2001 4:00 pm
Location: Galaxy Inside

Post by subhuman »

You still get plenty of PCI traffic. Yes, the i815, and i845 chipsets are slightly better (12 masterVerbs vs. 10) but the 'problem' hardly makes any difference in actual use. It's still lightyears beyond even the latest Via KT266A, which I have also tried.
Post Reply