K7D-Master Success!!! (So far...)

[cyclone6]

Progress has been made! I took 2 XP1700+ T-bred Bs (1 - 0310 xpmw and 1 - 0310 wpmw, both from Excaliberpc.com for $72 each) and tested them in a single CPU board (Epox 8rga+). Each ran stabily at 192 mhz X 12.0 multi = 2304 mhz.

Now the scary part (which is why I've been posting about blowing bridges...). With a max multi of 12.5 in the K7D, the best I could hope for was 12.5*150 = 1875 mhz Time to blow some bridges. Using the chip guide given by cmcquistion (thanks for your advice through this effort), and the power-supply method, I blew 3 bridges on both chips. (They aren't kidding when they say it sparks - does it ever!) But when I threw the first one in the K7D and booted up (at 100 mhz first), I got a 100 mhz x 16.0 multi for 1600 mhz. It worked So did the second one.

I've now got one CPU running by itself at 16.0 x 144 mhz = 2304 mhz. I have not modded the chips for SMP yet...that comes tonight. Does anyone have any idea why this might not work from this point? Haven't I already made it?

 

[verbatim]

keep it up hope u got good cooling.......

 

[eli]

oh.. man i'm excited. If this works I'll get me 2 of those and hit them up with 17 multiplier. (I can't do fsb... i think it's my crappy ram.)

Let us know, and you will need some nice cooling.

 

[cmcquistion]

Did you run Prime95 on each of those chips at 2300 MHz, to test for stability and make sure they're not making bad calculations?

If so, WOW is that a great overclock!

 

[cyclone6]

Did not run Prime 95

I run air quality models for a living (which is why I want and need the fastest CPUs around) - and I ran a series of 3 of those over a 2 day period. So no, I have not run Prime 95. Are there really cases where an O/Cd computer will stay up - run normally, and do bad math?

I guess I could run an air quality model on a non-O/Cd machine and compare the output files. That would at least let me sleep knowing that the numbers I used in my last report are OK...

I've got the rear-window defogger drying as we speak. I will stick these babies in the K7D tomorrow and see what happens.

 

[cmcquistion]

Re: Did not run Prime 95

I run air quality models for a living (which is why I want and need the fastest CPUs around) - and I ran a series of 3 of those over a 2 day period. So no, I have not run Prime 95. Are there really cases where an O/Cd computer will stay up - run normally, and do bad math?

Yes. That is why it's so important to run Prime95's Torture Test. It will test the CPU and it will stop if it encounters errors.

 

[FunkDaMonkMan]

Your rig sounds amazing. As long as you do the SMP mod correctly, I don't see how it won't work. Good Luck!

 

[cyclone6]

My rig isn't amazing at all...

Its just those wonderful XP1700+ t-breds. I bought 3 - and 2 of them do over 2.3 ghz. The third one I'm not sure about... its been a bit flakey. But I'm not giving up on it yet.

MY main concern with running the CPUs in a dual setup is power and chipset voltage. I've got an Antec true power 430w supply. And when I was running the chips in my single board, I upped the vdd voltage to 1.8v from the default 1.6v. I also had the vcore set at 1.675v, but that is not a problem with the K7D-master. Being able to run a single CPU in the K7D at 2.3 ghz does give me great hope that I will be able to run 2 at the same speed.

I've got 2 coolermaster HHC-001 Heat Pipes - and I'll have to watch the temps closely. (I always try to test the system with the MB out of the case - it makes resetting jumpers so much easier.)

I will stick these chips in later tonight and let you guys know what components I manage to fry

 

[cyclone6]

One quick word about the bridge blowing...

For the most part, the bridge blowing went well. I tried it first on the XP1700 thats been giving me some problems. Worked as advertised. Same with chip #2. But on the 3rd chip, when I dropped it into the K7D, it did not post. Just made some horrific clicking sounds. So back to the workbench. I tried each of the 3 bridges again, and after a few minutes one of the 3 went "pop". Turns out that my hands were not steady enough the first time through, and I had not blown the bridge. Just touched the two needles together and got a big spark. After the second try, the chip immediately posted at 16x.

I wonder what the post without the succesful blown bridge #4 was trying to give me. Maybe a multi of 100? Lets see...that would give me 100*133mhz = 13300 mhz

Too bad it didn't work...

 

[CPFitz14]

Must...resist.....temptation....to...blow...bridges. I want to do this so bad, but I just can't gather up the courage to do it. I've got no money if I lose a processor, but i'm tired of being stuck at 1875mhz.

-CPFitz-

 

[cyclone6]

No guarantees, but it really works...

Must...resist.....temptation....to...blow...bridges. I want to do this so bad, but I just can't gather up the courage to do it. I've got no money if I lose a processor, but i'm tired of being stuck at 1875mhz.

-CPFitz-

I understand your reluctance to try the bridge blowing. It scared the living shi*t out of me. But after reading the posts here, I decided I wanted the power of the xp1700s running in SMP

All I can tell you to make you feel better about the bridge blowing is that I had never done it before, and first tried it just yesterday. I ended up modding 3 chips (15x on the first one - in case I can't make it run above 2 ghz - and 16x on the second 2). They all now work as advertised. It is a bit scary.

The funny part is that one of my friends at work came into my office only minutes before I started the procedure. He didn't leave until I was done. I think he believes I am clinically insane. Applying voltage across a cpu to blow a bridge. (He really got a kick out of it when CPU #3 would not post...uh oh...there goes a CPU!) But when the CPU posted after the second attempt, he was amazed. He now refers to my office as "Dexter's laboratory".
Its really not that bad...

I am now preparing to insert the two CPUs into my dual board. I must admit that I am more than a bit nervous. I really want those 2 CPUs running at 2.3 ghz. What a machine that would be!

I will let you guys all know how it goes. And for cmcquistion, I did run the air quality models at non-OC/d speeds, and the results were identical. But I will abide by OC etiquite, and run Prime 95 when I am done. (If it works!) Sure hope it does...

 

[cyclone6]

One more point...

I'm about to shut this baby down and try the two chips at the same time. Just a quick thought about AMD vs Intel. As stated above, I run air quality models for a living. Floating point intensive to the max. It was a few years ago when I had my first experience with AMD vs Intel. I bought a P4-1.7 ghz chip thinking it would be much faster than our best chip at the time - a P3-1.0 ghz. Bought the board, bought the VERY EXPENSIVE rambus memory - and put the machine together. Then ran benchmarks.

It sucked. Can't sugar coat this one.

I was so sure that I had done something wrong that I brought a small sample run down to CompUsa to see how it would run on their p4-1.7.

Same thing.

I have never built another intel based machine since. They LIED to me when they released those P4 speeds. (Remember - this is all about speeds running our models - not gaming and some of those things where FSB speed is so important - and where it looks like Intel has a real advantage.)

Anyway, many benchmarks later (and all completed in the last 2 weeks), our 8 AMD athlon t-bird 1.4s run the models at the same speed as a P4-2.4 ghz chip. I figure an xp1700 t-bred clocked at 2.3 ghz is equal to a P4-3.9 ghz chip. If I can get two of those working in an SMP setup, well...

Too bad Intel doesn't make a 3.9 ghz chip. And won't for awhile.

I just hope AMD sticks in there and sticks it to Intel.

 

[cyclone6]

Update #1

OK - so the 2 chips are in, and booted. (I started at 100 mhz). So I have 2 1700 t-breds running at 16x and 100 mhz = 1600 mhz each. Their is one strange thing, though. When I run WCPUID it show CPU #1 as an XP, and CPU #2 as an MP

But it booted and I am running. Next step...change the fsd to 133. Update to come...

 

[cmcquistion]

If it recognizes one as an XP, then one of your L5 mods is squirrely. I suggest fixing it, before it causes problems. It will probably cause boot problems, otherwise.

Once you get them both in there and working, run two instances of Prime95 overnight and make sure they are stable. No offense, but I'm willing to bet that they won't pass Prime95 at 2300 MHz. If they don't pass, then they are making bad calculations and your system is VERY likely to experience random crashes and reboots.

 

[cyclone6]

If it recognizes one as an XP, then one of your L5 mods is squirrely. I suggest fixing it, before it causes problems. It will probably cause boot problems, otherwise.

Once you get them both in there and working, run two instances of Prime95 overnight and make sure they are stable. No offense, but I'm willing to bet that they won't pass Prime95 at 2300 MHz. If they don't pass, then they are making bad calculations and your system is VERY likely to experience random crashes and reboots.

You are right. I was not able (yet) to get this machine up and stable. I wonder if the XP/MP issue has something to do with it. It is running again as I type this, but only at 100 mhz. I guess its time to go back and test each processor individually in the K7D-master instead of the Epox board. Worst-case: I just go back to my modified XP2100s in the dual board, and run a single CPU at 2.3 ghz in the Epox board.

No offense taken...I know the one chip (the 0310 xpmw) will do 2.3 ghz. The other did it too for awhile - but I didn't run it over several days. I still have hope though that I can get this to work...

 

[JimmyK]

running 2 x XP 1700s modded to 15 multiplier at 150 fsb stable at 1.625v (2250)

u have the correct/best stepping Im using 2 DLT3C JIUHB 0310 XPMW's

my friend is using a water block and is running stable at 2.400 gighaertz on his iwll mpx2

scratch off the L5 bridge and do the mod again worked for me without a hitch just have good case cooling, Im personally using 8 dollar heatsink/fan combos but my case is very well laid out and cooled and quiet as well (use a multimeter to make sure u have scratched off that transparent thin layer of crap ontop of the L5 bridges)

and yes both my system and my friends 2.400 gig system run prime 95 stable and cool dont be disheartened it will work and it will be stable, just check the voltage and cooling, I try to keep my procs at 53 and under with full load.

 

[JimmyK]

and I used a razor blade and a mag glass to cut bridges worked without a hitch just have a multimeter present to make sure u killed them.

 

[plizzo]

any updates cyclone? im tossing around the idea of blowing some bridges since my gigabyte board doesnt have much movement for fsb. i think i need to hear a few more success stories before i risk losing 2 tbreds though.

 

[cmcquistion]

I've blown bridges on about 8 CPU's so far, with no trouble at all. I've done Palomino's, T-breds, and Morgan Durons.

 

[Xtc4u]

Ok so you blew some bridges to max out the chips on the "limited" K7D. If it doesnt run stable at the 16Xxxx couldn't you just paint over one of the bridges to drop it back to 15? But daaaaymm if it works you got me droolin.