First nForce3 Socket 939 Board Out!

[pkrew]

I like them both charlie. Both are well made and overclock well. I'll have to play with some of the memory settings, but the Asus seems a little faster in spi. I'm pretty sure I got 29s at 2.9G on the Asus. The Gigabyte got 30s, but is probably not a full second behind. My guess is that the Asus will be quicker in 2001se as well, but I'll have to check that out. The main advantage of the gigabyte is the ease of the mods and the higher vdimm that can be reached. The vcore mod on the Asus is also abit risky. It would seem to me that if you got a bad connection or pulled a wire loose your vcore could go very high and fry the cpu. Let me get the mods done of the gigabyte and I'll run some 3d tests this weekend.

[charlie]

:thumbsup:

I would hope that there's some sort of protection circuit for the Vcore......... I was real careful with mine, using a 100 ohm VR set at 20 ohm initially and working my way UP...

C

[k|ngp|n]

It's nice to see more boards coming our for 939!!
Now cmon....lets see some Xtreme 3d benching for crying out loud I'm so anxious to see what they can do with a high clocked pro or XT.

[blinky]

nice finds guys

in s754 the via chipsets were always a bit faster, but they couldnt quite clock as high as nf3, maybe this will be true for s939, altho on s940 the sk8v was king, so i dunno

[Overclock Kid]

via was faster becasue of the HT bus speed 800mhz vs. 600mhz

now we have 1ghz vs. 1ghz.

but yeah sk8v+fx53 was/still is deadly

[longshot]

im starting to get excited.....

[pkrew]

Just to give an update on this board. I did both the vdimm and vcore mod. I'm not too impressed atm. Took vcore to 2v and best I could get was spi at 2.934. Also the memory seems to clock worse with the vdimm mod. The results could be that the bh-5 having trouble with the higher volts or that you have to mod both vcore chips because of the dps. I'm going to try first removing the vdimm mod and see if that gives me a better oc. If not then later maybe trying to mod both vcore chips. The Asus overvolting to 1.82 gave me spi at 2.95. Charlie, next time you run your board could you run the voltage up a little and see if there's overvoltage protection at 2.1v?

[k|ngp|n]

Just to give an update on this board. I did both the vdimm and vcore mod. I'm not too impressed atm.

Are you talking about the gigabyte pkrew?[

[Overclock Kid]

P> did you do the "CAP" ddr mod?

3.3v direct to cap or the vdimm/vdimm mod as in the winbond chip??

that other guy that did the vcore also he gained almost nothing so there must be something thats not going right dual chips maybe like you said

[pkrew]

Yes, I'm talking about the gigabyte and yes I did the cap mod. I have my 3.3v rail set at 3.3v. Wish I had some ch-5 atm. Unfortunately the only place to get them here would be Compusa, Best Buy or Circuit City. Anyone know what kingston value ram has ch-5 chips. I've never looked for that chip.

Also I don't have another 50k vr atm. would have to wait until monday. I'll measure the resitance of what I have right now and see if I have anything that'll match it for the other chip

[saaya]

any kingston value ram with winbond chips is ch5, some are remarked as kingston chips, but the two dots make it easy to spot chips.

the kingston part name doesnt say anything about the chips they used. the best would be to go to a hardwarestore and check the memory yourself.

[StormPC]

Originally posted by k|ngp|n
It's nice to see more boards coming our for 939!!
Now cmon....lets see some Xtreme 3d benching for crying out loud I'm so anxious to see what they can do with a high clocked pro or XT.

I don't think the Newcastle 939's will ever break 35k on 2k1. The FX-939 will. I know it's early, but I have to say I'm surprised that the best S939 showing so far is this FX-53 939, which is currently #11 (see below).

http://service.futuremark.com/compare?2k1=7915393

The 3800+ and 3500+ have not made the first page and probably won't. Dual channel looks cool in Sandra but it doesn't make up for the cache difference between the ClawHammer S754 and the Newcastle S939 in 2k1 and most apps.

Compare the above FX-53 with my stock 3700+ non-vmod air with much less MHz on the CPU.

http://service.futuremark.com/compare?2k1=7911005

Yes, I have an XT, but my entire system is unmodded and 100% aircooled. These new cards are very limited by CPUs, so his FX-53 much higher clocks, along with his Mach II Prommie should do better I think. The XT doesn't make that big a difference in 2k1. Case in point:

http://service.futuremark.com/compare?2k1=7935263

This is a stock X800 Pro with ghetto clocks on a very fast stock aircooled 3700+.

If you want the best bang for the buck it's S754 Clawhammer, not S939. If you need to have the fastest setup on earth, S939 is it, but ONLY if you pair it up with an FX-53 and some very nice RAM and VC. And if you are not vmodding even the FX-53 is a waste, as the 3700+ performs very close to the FX-53 in most applications stock.

[pkrew]

Thanks Saaya, I'll look for some tomorrow.

As far as the FX-53 score. Hard to say for sure but looking at the individual scores I'd say that that's an air oc on the card. He's on winXP and dx9. With win2k and dx8 the best I got on air at 2.87 on a 3400+ with an unmodded x800pro on air was 31400. He's also using cats 4.6 which are slower. When I put the card on a moded mach I and ran it at 690/636 I gained 2300 points. That would put him at 34.6k. Not to shabby

[StormPC]

But still nothing that couldn't be done on a 3700+ in all probability. You're not using the fastest ClawHammer available, but he is. It is pretty obvious to me that a 130nm ClawHammer is a 130nm ClawHammer regardless of pin count or memory bandwidth. 90nm should bring new wonderfulness, but S939 is not it. To me it's mostly AMD marketing. Don't get me wrong. I'm glad that AMD is finally getting the hang of the marketing game. Maybe they'll be around for a long time, which is good for all of us. I'd hate to see the crap Intel would force feed us if they had no competition.

[xgman]

Originally posted by charlie
:thumbsup:

I would hope that there's some sort of protection circuit for the Vcore......... I was real careful with mine, using a 100 ohm VR set at 20 ohm initially and working my way UP...

C

I thought it was start at 100 and work down for safety. no?

[xgman]

Well it doesn't seem like there is going to be that "939 superboard" waiting in the wings. So more or less the same performance as say a 940/FX53 give or take. Via pro and NF3-250-939 seems very close but I guess it's still a little early.

Pkrew, does everything (locks, multis, etc) at least work right in the latest bios for the gigabyte (F3?)?

[pkrew]

All the locks and multi's work fine. I also think that the vcore mod for the board isn't working right, but only time will tell.

[xgman]

I think I'll try the vdim using the 1K VR method to the chip leg.

Well when all is said and done, you gonna pick which board you like better, via or NF3 nased 939?

[pkrew]

Just picked up some value ram that appears to be ch-5. Wouldn't boot at 2.5-2-2-8 and 3.34v. Tried it at 2.5-3-2-8 and any oc on the fsb resulted in error. at 3-3-3-8 I couldn't get past 208 fsb. Probably just crappy ram.

[xgman]

ouch. I'm going to be using 3700EB, so I hope this board is better than the Abit AV8. At least I will have mutipliers. Do the locks appear to be working though?

[pkrew]

Ok, now we're cooking with gas. Got frustrated and threw in my lvl2 and decided to try the second set of vdimm slots. Got into windows and it errored in spi at 208fsb. I said to hell with it and set it at 230. It ran fine. Here's what I ended up with at 3.34v at 2225. In case you can't read it its 256 x 10.5. Time to get the vcore mod figured out.

[xgman]

what does the stock vcore 1.7? actually produce when monitored?

You still on the same vdim mod or did you switch?

[misteroadster]

Originally posted by pkrew
Ok, now we're cooking with gas. Got frustrated and threw in my lvl2 and decided to try the second set of vdimm slots. Got into windows and it errored in spi at 208fsb. I said to hell with it and set it at 230. It ran fine. Here's what I ended up with at 3.34v at 2225. In case you can't read it its 256 x 10.5. Time to get the vcore mod figured out.

yes , i had the same problem , only the purples dimms works fine.
but you had to know that your ram are not at 256 mhz but 256X10,5/11=244,3 mhz , it's very good , the best i could have with LL 1.2 was 240 @ 5/2/2/2 .
Any question : LVL2 are bh-5 sticks ?

[xgman]

Originally posted by misteroadster
but you had to know that your ram are not at 256 mhz but 256X10,5/11=244,3 mhz

? I thought he was running sync ddr400? Not sure what you are saying. Why are you dividing by 11?

[misteroadster]

Originally posted by xgman
? I thought he was running sync ddr400? Not sure what you are saying. Why are you dividing by 11?

my english is too bad to explain you correctly that , but the mem controller set multiplier to the superior value , because it must be synchronized with HTT , and coef X,5 should be too complex , that explain you can go higher at coeff x.5