P4 3.0C L2 cache not working, broken pin also...

X

Xeon22

Bad Trader
Joined
Jan 28, 2003
Messages
1,163
Hi,

I recently bought a 3.0C p4 that had a broken pin. The guy had good references, and says he has fixed it with a wire in the socket. He only mentioned instability, but never mentioned anything about the L2 cache being busted.. Some cache tests completely fail to recognize it, and in Sandra, the cache & memory test there is only testing memory bandwidth, and not cache, as it almost flatlines on the graph. 8kb L1 data cache does show up and seems to work.

When running the CPU math tests in Sandra, while the FPU is on par, the ALU Dhrystone tests are incredibly low. The P4 running at 2ghz scores about 750 ALU, while a stock P4 @ 2ghz should score about 5350. Changing the memory bandwidth changes this number too. Increasing it to 400MHz increases the score to about 1100, as it is not using cache in this test either. To load up explorer windows, it can take me 20 sec to do because of this. I asked the guy about it, and he says he can't remembewr any performance issues of this type.

Now, about the pin, when looking at the pin side, with the corner with the 2 pins not there facing at the top right, the actual broke pin is on the very bottom row, 2nd one from the left. Does anybody have ANY idea what this pin does? I don't have alot of money, and thought I'd purchase it for a good deal, but I feel that the guy was not very honest about it.

Any help that you can provide would be very helpful, especialy the function of this pin, whether or not it'd cause the L2 cache to not work. I'm going to have to spend a little money to have it professionally fixed, and want to verify the function of this pin before I drop anymore money into the chip :(

Please help if you can.this sucks.. :(
 
The L2 couldn't have been busted when I used it as it folded at least 3 work units in 2 days. Without an L2 cache it wouldn't have folded for crap.

Anyone else have a clue? I sold him this CPU with a missing pin. I used a wire in the socket and the CPU worked fine on an Abit VT7 and ASUS P4P800DLX. Without the wire it would restart every few minutes.

As for the pin's use, it helps control the VID voltage to the CPU. I am not smart enough to actually understand what the pins do but my for sale thread linked you to the pin identification page.

Here's a pic of where the pin was missing from:
missing_pin.jpg


Here's a link to the Intel PDF file for the datasheet on P4s with 512kb cache.
ftp://download.intel.com/design/Pentium4/datashts/29864312.pdf
 
if the l2 cache was truly not working, it would take 4 hours to log into windows. my ic7 max3 modded bios allows me to turn off the l2, and the whole system becomes terribly slow, not just at specific parts.
 
That's what I'm dealing with. It took me 3 hours to put on a fresh installation of windows XP. Everything is terribly slow, as you describe.

Now, if anybody can turn off their L2 cache and run sandra, could you do so and see how it affects your ALU performance? Mine dropped down to 700 something as I'd described. This would help me determine if it's exactly the cache. Running the cache latency on CPU-Z, it only detects the L1 8KB data cache, and does not detect any L2 cache.
 
I've read about L2 in the past. I can agree with cookie that when you disable the L2 cache that programs will practically become unworkable. There's no way the L2 wasn't working for me.

edit: Then what the hell coulda killed the cache? It worked when i pulled it from the motherboard. From there I placed it right in the box. What happened with the CPU when you FIRST POSTed it. Did you go right into installing windows or try it on a rig that already had an OS?
 
well here are some facts that are pretty well known. Most p4 dont like high vcore, not becuase the core cant handle it, but because the cache cant handle a high amount of voltage to it. This is why many people sought out for 30 cappers, supposedly the cache there is more resilient to high volting. And you described that missing pin to be the vid controller to the cpu core. What a vid voltage controller does is protect the cpu from getting overvolted or undervolted.

This is probably where the problem started from. Perhaps when xeon put in the chip, the new pin was not making contact with the socket, thus allowing the cpu cache to get trashed by unstable voltages.

But this shoudlnt really happen unless the vcore on the chip was raised while xeon had it. But maybe its possible.

Thats just my take btw. Maybe you are still able to rma the chip back to intel.
 
My complaint to the RMA thing, I doubt Intel will rma a chip with a missing pin. If I was sure they would, I woulda done it. I just didn't have the time or patience. I looked at their RMA section online but they just ask you to call in for RMA, you can't do it online. So once I saw that I said screw it, as I work 8:00-4:30 everyday. Not much time to deal with them on the phone :-/
 
DriveEuro said:
My complaint to the RMA thing, I doubt Intel will rma a chip with a missing pin. If I was sure they would, I woulda done it. I just didn't have the time or patience. I looked at their RMA section online but they just ask you to call in for RMA, you can't do it online. So once I saw that I said screw it, as I work 8:00-4:30 everyday. Not much time to deal with them on the phone :-/
Click to expand...

i rmad a chip a while ago, and the intel rep on the phone was very shocked that i was calling for an rma for a processor as he mentioned it was very unusual for a cpu to get sent back. However, that chip was truely messed up. It was retail, so i had to ship back the chip as well as the heatsink that came with it, odd huh. I do remember that they said they dont cover physical damage. But im sure you can play the "I have no idea whats wrong with this thing" game and they will probably go through with the rma anyways.
 
Well this shines a little bit of light on this. I hope Xeon can get it RMAed or fixed.
 
I had this problem the very first time I put the thing in and loaded windows. I never upped the vcore in anyway. I
 
I only upped the vcore cuz it was rebooting on me so I figured that might have been it. I never went higher than 1.6v though.
 
Well I get another new board today. The one I was using to test it only supports 533fsb. Don't know if that'd cause this problem, but I will be testing on a new board today.
 
Well I got my new mobo today, and the cache thing isn't the case now. Performance wise, it's where it should be. Don't know why it wouldn't work correctly with the other mobo, however, a 533 chip worked fine.

Anyway, I've not tried with the wire mod in place...so it is unstable, just as described. I will be attempting to wire it tonight, but anyway, it's now as described when you sold it to me.
 
You must log in or register to reply here.
Back
Top