Athlon XP - It's The Pits

Add Your Comments

First look at AMD’s newest CPU. — Joe and Ed.

SUMMARY: Put away your pencils – AMD’s “Golden Pits” on the XP made the multiplier fix a lot harder.

Ed picked up a retail box version of the XP last weekend at a local computer show. This is a 1400 MHz part, now marked by AMD as “1600+” on the box. Nothing on the box identifies the CPU as a 1400 MHz part. It puts out 62.8 watts at spec 1.75 volts.

AMD XP

This is it in all its glory – lots of bridges compared to T-Birds.

L3 Pits

The L3s show the diabolical way AMD is fixing the XP – burn, baby, burn!

L1 L3

This shows the L1/L3 “pits” that are dug by laser into the CPU’s ceramic carrier. Normally the pits are dark, as shown by the two L3 pits. If you carefully scrape the dark area, the gold contacts are uncovered.

Detail

This is a shot with the flash set at a low angle to show the pit’s depth.

L1

This shot shows only the L1 pits – note that it’s not only “connect the dots”, but now it’s “fill the pits and connect the dots”.

We are literally playing with this now. We’re using rear window defogger paint to do this and it’s not easy. We’re seeing posts in the Forum by others that indicate it may take 2 or 3 applications to get the dots connected.

Our first shot at it got it up to 11X133 – others are having the same experience, so we’re in good company!

Ed had first crack, found that it would not boot at all using an A7V133 PCB rev. 1.04 board and the 1007 BIOS. Asus is stating that
only PCB rev. 1.05 or better will work with this. A few have reported that they’ve gotten a rev. 1.04 board working, but Ed isn’t one of them.

He then tried the chip on the MSI K7Master with BIOS revision 1.3, but this board now just functions for a little while before dying, which it did with this processor, too.

We did run it on Iwill’s KK266+ – it booted up at spec speed, although the BIOS did not identify it as an XP. We could vary the voltage no problem, but not the multipliers. For the first run, we varied FSBs up to 166 MHz. No stability testing yet, just playing.

We have tried closing the L1s, and the CPU now boots at 11 rather than 10.5 but that’s it. We suspect this is a BIOS issue – we’re going to play with this thing some more and see what we can do with it.

Email Joe

I ran some quick benchmarks and found them interesting when compared to a T-Bird at the same speed. Please note, though, that this was on a motherboard that was not recognizing
the XP as an XP, and thus no SSE enhancements, etc. (We’re looking to buy the Epox 8KHA+ for some serious testing).

First I ran the XP at 1650 (11×150) on Iwill’s KK266+, 1.93 volts (spec 63 watts, O/C at 90 watts), air cooled with a Glaciator. RAM was set to CAS2, fastest memory settings, with a Leadtek GeForce2 Pro, 32 MB:

Benchmark

Result

SiSandra 2001 CPU

4620/2281

SiSandra 2001 MultiMedia

9208/10595

SiSandra 2001 MEM

685/715

Quake Demo 1 & 2

185.1/182.8

3DMark 2000

10207

3DMark 2001

FAIL

At this setting, 3DMark 2001 would freeze. I then ran the same benchmarks with EXACTLY the same settings for both the XP and T-Bird at 1460 (11×133) – the only thing I did was swap CPUs:

Benchmark

XP

T-BIRD

SiSandra 2001 CPU

4082/2016

4085/2000

SiSandra 2001 MultiMedia

8141/9368

7956/10037

SiSandra 2001 MEM

617/632

570/650

Quake Demo 1 & 2

164.1/162.2

154.4/151.7

3DMark 2000

10207

10163

3DMark 2001

4131

3998

Looks like the XP is a definite improvement over the T-Bird, but let’s face it – for most of us, we’re not going to see the change on screen. For those of you into image processing, the XP will show performance gains.

Cooling is a tad easier with the XP – at 1400, the T-Bird radiates 72 watts compared to the XP at 63 watts. A little easier, but not even close to Intel’s ice cubes.

We’ll continue to play around with it and see what we can do.

Email Joe

Leave a Reply

Your email address will not be published. Required fields are marked *