Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: a question for you Vincent

Author: Aaron Gordon

Date: 15:44:01 07/02/03

Go up one level in this thread


On July 02, 2003 at 18:25:22, Keith Evans wrote:

>On July 02, 2003 at 17:58:58, Aaron Gordon wrote:
>
>
>>Hard for me to explain, but the extra heat BurnK7 generates would push your chip
>>'over the edge' if it was that unstable. Running your 'important' code for
>>example would result in a cpu temperature a few degrees cooler. Read my post for
>>proof how the cpu would be stable in those few degrees. Also, I don't run the
>>chips on the very edge. Don't knock my pretested chips before you try them.
>>I've taken a lot of time/effort to make sure they work flawlessly.
>
>This is only true if the code in question is exercising the critical timing
>path(s). Designers typically do a static timing analysis to find the worst case
>paths in our chips - I'm sure that AMD does this. What you are doing is a
>dynamic timing verification with no knowledge of the worst case paths. You may
>be running enough different tests that you will catch it, but unless you have
>verified it then you don't really know.
>
>We typically use an oven to heat parts up for qualification testing. BurnK7
>seems more like something that you use to test that your system level design
>(airflow,...) is OK, rather than to qualify a chip across temperature.
>
>AMD themselves missed the JPEG bug. I'm sure that they spent a lot of time and
>effort developing their tests too.
>
>If I were buying your chips for a home gaming PC them I would feel comfortable.
>I would not buy them for a machine which was doing something like producing IC
>masks. There I would spend the extra couple of thousand for a machine which is
>operating within the manufacturer's specifications. (Yes and even then I could
>get burned.)
>
>Are the packages hermetically sealed before and after modification? (I don't
>know if they are hermetically sealed before or not, and I don't know what goes
>on during the modification.)

For the most part, the chips are tested in a temperature controlled area with
"average" cooling. That way you don't need some god-like cooler to run stable.
Any old cooler rated for near what the chip is rated at will do fine.

I understand your concerns about testing, however, temperature does make a
massive difference. If the chips are that on edge in the first place, BurnK7
and/or Prime95 will produce errors, crash/lock/reboot the system, etc.
Overclocking doesn't mean you have to run on edge. If you check out my post
here:

http://www.talkchess.com/forums/1/message.html?304354

you'll see I go the extra mile to ensure stability. Well below the chips actual
maximum clock speed.

When you consider some P4-3.06GHz chips won't run 3.2GHz stable (%4.37
difference in clock speed), thats extremely close to the edge as is. My chips
usually are 15% or more below the 'edge' of stability.




This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.