Zap's "wire trick" vcore mods for socket 754

[Zap]

Here's my gift to the Anandtech community who have socket 754 boards that allow raising HTT but doesn't give vcore (or not enough)...

Zap's AMD Socket 754 vcore guide

Just a little something I whipped up today. Had better things to do, but wasn't being productive because I've been pissed at some rebate issues, so I needed a distraction.

If anyone tries it out with success, LMK. Also, if anyone thinks I made any errors, LMK.

Enjoy.

 

[Corporate Thug]

well done! THANX

 

[Assimilator1]

Excellent! ,thanks for that

Now maybe I can get my stupid Asus K8N to put more than ~1.46v into the Semp 3100.
1.5v will do me.

 

[Zap]

:thumbsup:

Nice to know it is of use to someone.

 

[AkumaX]

hehe, awesome!

 

[chinkgai]

nice! too bad you didnt come up with this during its prime...but still...nice!

 

[Zap]

What "prime" is that? Back when 130nm Clawhammers weren't clocking too high and socket 754 motherboards didn't have working AGP locks, or now when Palermos start at $65 and some people are getting a whole GHz overclock out of them on new PCIe motehrboards?

 

[Blooz]

Nice job on the guide, Zap!

 

[chinkgai]

Originally posted by: Zap
What "prime" is that? Back when 130nm Clawhammers weren't clocking too high and socket 754 motherboards didn't have working AGP locks, or now when Palermos start at $65 and some people are getting a whole GHz overclock out of them on new PCIe motehrboards?

oh c'mon...i didnt say 754 was dead...i still use one for my downloading rig...lol

im sure you know what i meant by prime (meaning no 939, etc)

 

[Zap]

Heh, well, moot issue as with all 130nm chips the maximum would be to get 0.05v over default. With 90nm chips, can get 0.15v over default. IMO socket 754 overclocking is in its prime NOW with the Palermo cores and choice of AGP or PCIe boards. Heck, even the budget ECS boards are overclockable!

Hmmm, strangely looks like my .sig got corrupted... gonna go fix that... probably something to do with rebuilding the forum database... heh, maybe I'll lose my postcount like some others around here and I'll be an Elite n00b.

 

[DAPUNISHER]

^

 

[Assimilator1]

Hey zap ,I don't suppose you'd know which way this vcore mod will be used by my Asus K8N do you?

 

[Zap]

I do not know... how much vcore does your motherboard give you? I'm thinking that boards which does not max out at exactly 1.55v are a good candidate but really you won't know until you try it, unless your board shows vcore as percentage in which case it'll probably work fine in conjunction. Of course if your board doesn't give you vcore at all (such as Chaintech VNF3 with newest BIOS for 90nm cores) then this gets by the limitation.

 

[Assimilator1]

My mbrd supposedly gives vcore in increments of +0.025v upto 1.55v.
But for some screwed up reason the vcore maxes out at 1.46v (bios HW mon reading) & won't go higher regardless of the actual chosen voltage.

I might give it a go tomorrow.

 

[Zap]

Be sure to report back here if you have success. Good luck. :thumbsup:

 

[Assimilator1]

Well I modded the cpu to 1.45 vcore in the hope that'll give me a higher range.
But strange results indeed ,probably due to the mbrd (Asus K8N)

Anyway ,the def vcore did change from 1.4 to 1.45 ,well 1.46 actually.
But choosing a different vcore in the bios gave these wierd results.
Any setting from 1.45-1.55 gave 1.46
1.575 & 1.6 gave 1.54 ,I didn't fancy trying higher
1.35-1.425 gave 1.39

I wanted to try a def of 1.5v but I was running out of time & was unable to break the paint join I had done! ,even after several attempts.
So atm its stuck at 1.45 def setting ,until I try again to return it to normal.
Maybe tomorrow ,& then hopefully I can mod it to 1.5v def vcore

 

[BW86]

awesome guide :thumbsup:

 

[Zap]

Wlel, ultimately you are able to go to 1.54v where before it "maxed" at 1.46v, right? Strange behavior... any BIOS updates for that board?

 

[Assimilator1]

Only Beta ,otherwise I'm on the latest ,1010.
Another odd thing is that if I left the vcore on auto it would crash in windows ,as if it were too low ,but setting it to 1.45v cured it ,even though in both cases the bios h/w monitor was showing 1.46v ,admitidly that's after I tried to break the paint join ,so maybe it gave an unstable connection? (didn't try it to windows before hand)

1.54 is too much for my liking ,I'll give it another try for 1.5v latter in the week

 

[Assimilator1]

Well this is bizarre ,I've removed the paint & it looks pretty clean around the pins now ,yet it still defaults to 1.45v setting ,but the vcore range available to me is now back to what to was before the mod ,as if it were on def vcore of 1.4v & the mod was removed

Anyway I can manually set the vcore to 1.42v which is its ideal voltage for 2.5GHz.
I'm not sure I want to try a def vcore of 1.5v now incase the 2 add up & I end with 1.55v ,that wouldn't be a problem if I can lower the vcore below def ,but I've no way of knowing before hand if it'll do that.

Anyone else tried?

[edit]btw I did try clearing the CMOS ,no difference

 

[Avalon]

Zap, you are the man.
:thumbsup:

 

[Assimilator1]

My CPU now needs 1.45v ,or maybe I didn't fully test it on 1.42v? though I thought I did...
Anyway ,the upshot of that is I'm very unlikely to gain much speed by increasing the vcore to 1.5v ,assuming I could actually get 1.5v ,which I now can't be certain of.Pitty.
If the def vcore ever returns to 1.4 I'll give 1.5 a try

Anyway ,good info & I'm sure it'll work for others

 

[nyker96]

Zap my compliments on a nice guide on CPU modding. But I do want to issue a caution about putting up modding guides. Saw several people doing so and many ends up getting lots of trouble from people who damaged their equipments because they didn't follow procedures etc. However, people here might be a different crowd. Nice work.

 

[Zap]

Hence my disclaimer at the top of my web page. Yes, fully possible to kill a CPU (or more) with these mods. I did a "wire trick" mod to a socket A system for a friend and it was running great until he decided to mess with stuff - he killed the CPU. Don't know if it was from wires in the wrong places or from chipping the core after repeatedly mounting/unmounting the HSF. He's killed about five socket A CPUs to date that I know of. :roll:

 

[Assimilator1]

lol ,he shouldn't be messing with the inside of a PC then

About 7yrs ago I killed one, an old S7 Cyrix I o/ced & it died within minutes,lol ,what a POS that was! (it deserved it).
Though I came damn close to killing another one recently,an Opteron 146!:Q ,I was removing the heatsink & it yanked out the CPU with it! I couldn't believe it!,the HS compound had such a good seal on it that it overcame the grip the pins had! ,I was damn lucky I pulled it straight up & that I didn't bend or break any pins!.