High cache latencies-compromised CPU

[Mr_Nuke]

Hi there! I had some trouble with my Athlon XP 1600+. It was running as hot as 80C on full load. Then I remembered that when i bought my PC, due to some compatibility issues, I had to change the mobo. The guy that did this( at that time I was less thean a newbie) did not change the thermal pad. So i went looking and got a free tube of Silver Grease that came with a Titan cooler(I got just the paste, not the cooler)from a PC shop. I cleaned the die and HS and applied the paste, got everything in place. I also installed a sleeve bearing 80mm intake fan beneath the HDD and reversed the fan beneath the PSU to push out the air. It's much more silent than b4 andw/ the Prime95 torture test I only get 52C MAX!!!ON A PALOMINO CORE!!!.With stock HSF.

The problem is that CPU-Z identifies some nasty cache latencies:
L1:4 cycles
L2:25 cycles(21 cycles for this only level)
I know these are high and I've lost a lot of performance since i first turned it on with the new mobo(k7T Turbo2 by the way).

Are these latencies normal?
What are the normal latencies for an Athlon XP 1600+?
Have these latencies risen from running @ 80C for more then 2 years?

 

[TheRod]

Have these latencies risen from running @ 80C for more then 2 years?

Impossible, digital equipment can't "slow down". They work or they don't. Layer 2 cache are always slower than Layer 1 cache. You should not bother with these numbers. If you run benchmarks and get good results compared to other system similar to yours. You don't need to worry with these cache latencies numbers.

By the way, you can't determine cache speed by checking latencies numbers.

--
Lookin' to fill that <font color=blue>GOD</font color=blue> shape hole!

 

[P4Man]

>Have these latencies risen from running @ 80C for more then
>2 years?

Yes, quite likely. Due to the heat, the silicon probably expanded, which leads to slower timings (longer distance). The upside is that due to this expanding, your caches are likely quite a bit bigger than before, so it shouldnt matter much.


Oh, and if you believe that, I've got a bridge to sell you.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[ChipDeath]

LOL

---
Epox 8RDA+ rev1.1 w/ Custom NB HS
XP1700+ @205x11 (~2.26Ghz), 1.575Vcore
2x256Mb Corsair PC3200LL 2-2-2-4
Sapphire 9800Pro 420/744

 

[endyen]

The line2 is wrong. Sounds like it's doing system memory .

 

[juin]

lol

i need to change useur name.

 

[Mr_Nuke]

The shop I bought the PC from is a dealer of Flamingo Computers(East Europe). So I contacted Flamingo, explained my problem and they asked for some numbers and said I have their full support, so I might get a new CPU, if they see it's not my fault.
Thanks!

 

[P4Man]

System memory in 21 clockcycles, wow that would be AWESOME. Just FYI, even an Opteron with its ODMC has a memory latency of around ~160 clock cycles, which is rather excellent already.

The numbers reported by Mr Nuke are completely normal.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[juin]

System memory in 21 clockcycles

decrease your clock speed to 100 mghz.You should have your 21 cycle lantency.

i need to change useur name.

 

[Mr_Nuke]

The numbers reported by Mr Nuke are completely normal.

Here's a more likely scenario: The CPU operating at the limit temperature some transistors(locally) could have pessed the temperatuire limit, some frying up. So now the cache has to do the same operations, but with less tranzistors available, resulting in these higher timings.

You must know that the first time I checked the latencies with CPU-Z I had 17 clockcycles L2 cache, and the next day it was 21. Also I experienced a gradual performance decrease, and yes, I defragmented the HDD, reinstalled Windows,etc. but the performasnce level didn't rise back. Let's not forget stability issues: the first time I did a boot-up with my K7T Turbo2 It was extremly stable, now it even crashes on idle, with no reason. My belief is still that the CPU is compromised.

Anyway I achieved a performance nobody else on the forum did, or at least I'm not aware of: a Palomino core running at only 52C on full load.