|
So, if some of you don't already know, AMD has recently released a series of 90nm chips to the 754 line, all with Sempron badging. Looking at previous roadmaps, this was to be expected as they phase out the SKT A line, however, these CPU's come with a catch...no 64-bit, no CnQ, and a small SMALL L2 cache(128kb). This week I got a 2600+ Sempron plopped on my desk, so it shall end up dutifully tortured and tweaked.
The processors marking are as follows:
SDA2600AIO2BA
CBBID 0504APEW
On the bottom corners are the standard gold triangle for placement, and on the other side is a small gold "N".
On the underside of the CPU is a 5-digit number 27973.
The included heatsink is the newer model aluminium type with added fan bracket on top, with 60MM fan, and what has been tested and confirmed as Shin-Itzu thermal paste. The heatsink features the newer clip-type as well, which can be difficult to secure properly, over last year's clip.
The chip's pcb is slightly different than the older newcastle substrate; it has a higher matte finish.
I recently changed the flow in my set-up to go pump-GPU-rad-CPU-rad-resevoir-pump, which has greatly helped in isolating the heat from each source. I put in fresh paste on the 3400+ yesterday and saw a immediate 4c drop from the previous installation, which had the thermal paste thoroughly broken in. Not a standard lopp, but it seems to be the best i can get with what i have. I'm going to give the thermal paste a good week to break in, and then I'll start clocking up again. I had the CPU @ -30 last night, and managed to get it higher than I ever expected, and was able to confirm that the 2600mhz wall i had had before was entirely due to heat. 
My board had maxed out @ 265mhz HTT before lastnight (ABIT KV8PRO), but upon removing the northbridge heatsink, i found a blob of paste smaller than a dime...it actually only covered the "VIA" branding on the chipset!  This may have been part of my problem breaking 2600mhz as well, but we shall find out.
My board does not have the voltage options available for lower than 1.5v, at least with the 3400+ in there, so it remains to be seen exactly what will happen when I drop the 2600+ in, but I have updated the bios to support these newer chips.
I also picked up some Kingston ValueRam Samsung TCCC sticks; got them far cheaper than you would think, really. These chips can be found in numerous pc3700 and pc4000 sticks, and although the ValueRam line is tested, I was hoping for those pc4000 speeds. I missed out tho. My Winbond-based vdata performs much better, although performance at the low-end is very similar. 2-3-2-6-1@ 200mhz; 3-3-3-8-1 @ 217; 3-4-4-8-1@ 233mhz. Any higher just does not work without errors, although i think i can tweak a few more mhz out of them once there are sufficiently burned.
DISCUSS this HERE: |
