msi 790 fx-gd70 thread

[TiTON]

Well I set to 2.62 PLL and 1.51 vcore.... if this fails, I may start from scratch and run my ram/NB stability testing all over again.

I knew going in that I was probably going to get an ok chip not a great clocker, my last CPU I bought that overclocked really well was my opteron 165 and that was only because I knew what stepping to look for and paid a little extra to get it

*edit* made it through prime blend tests 1-9, started test 10 and a bit later puked out a BSOD. Sigh.

Hey Sparky.. just curious.. have you watched the temp level, and voltages during the prime? The BSOD might be caused by a droop on vCore.

- Ton

[Sparky]

Temps were 50 or below the whole time (I'm on water). Volts in CPU-Z were around 1.5 under load. Actual volts, I'm not sure, I haven't poked around the board with my DMM.

With the CPU at stock, ram at stock, and the NB and HT at 2600 prime ran 5 hours last night, then failed. No BSOD or reboot, just a failure. "Round was .498733611, expected less than .4" or something like that. Then a little bit later the program stopped responding according to Windows. Time to fiddle with some NB/HT volts and stuff. I want this perfectly solid darnit

[chew*]

I can confirm that NB vid overvolts or at least was on 13B6....Measure at the read point near debug LED

[Sparky]

CPU_NB on 1.3 is about right. CPU vcore is actually low by 0.011. The other three points are under the heatpipe - nice design

[chew*]

CPU_NB on 1.3 is about right. CPU vcore is actually low by 0.011. The other three points are under the heatpipe - nice design

LOL I can see it now.

User - I would like to request RMA?
MSI - What appears to be the problem?
User - My vdimm is undervolting very bad
MSI - How did you determine this?
USER - I pulled the heatpipe assembly and measured with DMM
MSI - Can I have your serial #?
User - Sure XXXXXXXXXXXXX
MSI - Thank you but we can not process your RMA request due to you voiding warranty by removing heatpipe assembly

[bingo13]

I can confirm that NB vid overvolts or at least was on 13B6....Measure at the read point near debug LED

But it should drop to actual BIOS setting under load, if not, I need to speak with them again. My board does that on 137/1.3 and the latest beta the last time I measured it, but it is on purpose, just like Vdroop on Core VID is really our friend.

[chew*]

Yah I haven't tested it on the latest bios yet bingo, when I get a chance I will load it up with the DMM on.....

Vdroop doesn't bother me really, as long as its consistent throught the entire range of voltages....

Poor voltage regulation like on the asus ddr II is what bothers me.....Droop all the time is a hell of alot better than voltages running away on you.

[Sparky]

At stock, vcore in bios set to 1.35, I get 1.339 measured, vdroop brings it down to 1.333 under load. I'll keep checking that when I start upping volts but if that stays the same then when I was set at 1.51 that actual vcore under load was approximately 1.495 or so... interesting.

I'm getting a NB voltage reading of 1.245 under load as well. Set it to 1.21 in the bios IIRC. So that overvolts by 0.035 when under load, at idle it is almost spot on. Very interesting...

Wonder what the HT voltage is doing? I have that set at 1.21 in the bios right now. Too bad that is under the heatpipe

[MadDias]

sparkyjjo?

u didn´t answer me if u enabled acc...
i´m asking this because i get bsods when it´s disabled.
1.5v is definetly too much for 3.8ghz.

[Sparky]

sparkyjjo?

u didn´t answer me if u enabled acc...
i´m asking this because i get bsods when it´s disabled.
1.5v is definetly too much for 3.8ghz.

Oh sorry, yes I tried it and it actually ended up slightly more unstable. Instant crash as soon as I started prime95 instead of waiting a little bit.

[TurboDiv]

Guys i got my UD70 + CPU but my ram will come Monday . The batch number of CPU is CACYC AC 0911EPDW. Any of you heard something about that batch # ?

[Sparky]

Mine is similar, CACYC AC 0913FPMW. I don't know anything about what is good and what's not though

This is my 3.914GHz suicide shot when I had no idea what I was doing now that I'm starting over on the OC maybe I'll get close to this stable... Or at least full 4GHz+ suicide.


In other news, I got my NB and HT 6 hours prime stable at 2600. I needed a slight bump on the HT voltage Maybe that's why everything was willy nilly unstable earlier

Next up, RAM or CPU.... Hm..

[TiTON]

Hey Guys.. finally want to share some results with everyone on this board. Earlier this week, I basically went back to the drawing board. I am starting my tweaking from scratch to see what I can do to improve my performance with this board. Now understanding more the impact of NB Speed and Keeping the IMC Happy, i started my adventure all over again

It was really funny, I had my ram speed really fast and tight. It was able to run through 3d Benches without a problem at all. So I figured i was golden. I then try to run Spi32m at those speeds, and Super Pi would just not run, after hitting "GO". I mean.. no errors.. just hang. WIndows worked just fine. I have never seen this type of problem before. With SuperPi, I use to either getting an error like "Not Exact Round" or BSOD. I really thought my SuperPi was corrupt at first. But after tweaking a bit more.. it was all RAM related. Crazy stuff.

So I spent the last two nights tweaking the ram, with the CPU on phase. So here is the best SPi32 I was able to get with my memory:

860Mhz 6-5-5-15-22-1T


I also wanted to see what the max speed I was able to achieve period. So relaxed the memory and cranked up the HT.

952Mhz 9-11-11-29-40-1T


Please ignore the times in SuperPi. I was not trying to complete it fast, but wanted to make sure the memory was stable enough to pass those benches. I believe I was running between 4.2 and 4.3Ghz.

Now Memory is done.. time to move on to NB

- Ton

[massman]

952 32M stable is just AMAZING

[chew*]

Uncle bingo knows best Guess my tips helped you out Titon.

Be carefull with those d9's I kept running 32m and they degraded and would no longer pull 6-5-5 thus my 6-6-5 timings I use now.

[TiTON]

952 32M stable is just AMAZING

Massman - Thank you I can probalby push higher, but wasn't really trying. I think I started upping the the HT for the heck of it.. and it finally crashed at like 960 ~ 970. Didn't try to run Spi32m beyond 952.. I just wanted that whole number of 1900Mhz.. versus 1898.

Uncle bingo knows best Guess my tips helped you out Titon.

Be carefull with those d9's I kept running 32m and they degraded and would no longer pull 6-5-5 thus my 6-6-5 timings I use now.

Hey Chew.. i'm done running SPi 32m.. I can't stand it.. I get bored out of my mind :P So there will be no more SPi32m runs for a while. I made sure to stock up on the memory. I have 5 other sticks Lost 1 stick already to condensation :P

Going to play w/ NB on phase, but don't think i will get much due to temps..

But I do have somethign else to dial in on water

Hopefully LN2 on Sat and on Sunday.

- Ton

[massman]

Higher than 960MHz? Or crashing right at 960?

[chew*]

Yep, running 32mPI is about as exciting as watching paint dry....Unfortunately I have found it a necessary evil to dial the ram in perfect....

What you were experiencing was exactly what I have been talking about.....The IMC was not happy...and would most certainly hold you back in the high clocks department

[hellcamino]

My new pump and ram showed up today! Once this thing is done bubbling I'll see what the new ram can do, it's Crucial 2x1gb PC3 14400 CAS 8.

[TiTON]

Higher than 960MHz? Or crashing right at 960?

Hey Massman - actually i really didn't pay attention. It was the last thing I did before going to bed :P

I will still be on phase tonight, while I work on something else. I will see what my max Spi32M is at that timing.

What is the highest you seen that is SPI32m stable on AM3?

Yep, running 32mPI is about as exciting as watching paint dry....Unfortunately I have found it a necessary evil to dial the ram in perfect....

What you were experiencing was exactly what I have been talking about.....The IMC was not happy...and would most certainly hold you back in the high clocks department

Hey Chew - Cool. Great to hear. there maybe more in this CPU

I do have a question.. Will upping my NB speed and volts TOO much have the same effect of not keeping the IMC Happy? I did notice on the GB mobo when I up the vNB too much, SPI32m didn't run. Even at Low NB Speeds. (p.s. I hate bios that say +.01v, and don't tell you what the starting voltage is)

Well.. hopefully keeping everything spi32m happy, i will be able to hit 6Ghz bench.. if I can.. there will be some interesting results by Monday

- Ton

[massman]

Hey Massman - actually i really didn't pay attention. It was the last thing I did before going to bed :P

I will still be on phase tonight, while I work on something else. I will see what my max Spi32M is at that timing.

What is the highest you seen that is SPI32m stable on AM3?

952MHz actually ... but the number 960 is pretty interesting as there are some rumours.

How much tweaking was involved in getting 952 stable? Just relax timings, phase and some volts?

[TiTON]

952MHz actually ... but the number 960 is pretty interesting as there are some rumours.

How much tweaking was involved in getting 952 stable? Just relax timings, phase and some volts?

Hey Massman. Well.. for tweaking.. i wouldn't say "much" tweaking.. just need to not go crazy and change everything. No shotgun approach.

I will give a shot to see if I can get higher the 960. As for the timing I used for 952.. that was spd at whatever I booted at.

Lets see if over 960 will be possible.

- Ton

[Smartidiot89]

Should an "Athlon II X2 250" or "Phenom II x2 550" work on this motherboard without a BIOS update? If not I am kinda in trouble

EDIT: even is a CPU is "incompatible" due to old BIOS, could I still flash the motherboard or would I need an older AM3 CPU to get it working?

[TiTON]

Hey Guys..

Wanted to share my results with my memory testing with this board. Massman.. do you mind PM'ing me and sharing your insight on that 960 Mem thing? It apparently appears to be real.

I was not able to run Spi32m at 960. It was VERY flakey. Sometime I hit 960 and it locked.. some time it didn't. I am sure it would not be able to run 32m w/o some tweaking, which I didn't do. I was able to run Spi1m at that speed. At 964 it locked up no matter what.

People might say that the 960 is the limitation of my ram, which I can say is definitely not. If I can runs the speeds I listed at 952 / 956.. a 960+ run should have been possible. 960+ may be possible w/ more / heavy tweaking and more voltage.. but I was getting bored watching SPi32 m. :P

So ain't crap w/o screen shots.. so here goes Side note - All ram was at 2.2 ~ 2.3 volts. I thought adding voltage would help, but it really didn't. I think I ran higher volts on one, and lower on the other. I am sure it would have all been stable at the lower volts, but wanted to factor out volts as being the limiter.

Spi 32 Cas 9 @ 956Mhz (2.2v ~ 2.3v)


Spi 32 Cas 7 @ 952Mhz (2.2v ~ 2.3v)


Spi 1m Cas9 @ 960Mhz (2.2v ~ 2.3v)


Spi 1m Cas7 @ 960Mhz (2.2v ~ 2.3v)


I really have a feeling I might be able to do like 900+Mhz w/ Cas 6-7-7. Might have to give that a try.

Let me know if you guys have any questions.

- Ton

PS - its a pain to try to crank up NB Speed on phase :P Should be going to LN2 tonight, to see if I can any increase stability on my CPU/NB for 3d Benching.

[Sparky]

This G.Skill isn't too great it seems, it fails to POST at 1600, even with a voltage bump or relaxing the timings a little. Seems that 2 of 3 G.Skill kits I've had haven't overclocked worth a crap (the 3rd one, I'm not sure how much it does since I was CPU limited). I wish that OCZ reaper hadn't been OOS. Oh well. I do know I can at least tighten timings some.

At 1.38 set in the bios I get 1.36 under load measured with the DMM. 3.6GHz clock speed set by multi x18. Running prime now, just started test 7 of blend 1024k. If this is stable, I'll work my way towards 3.8GHz+. If not, I'll see what the vcore is at idle. Either vdroop is more or the offset from bios setting to actual voltage is more.

I think the main thing I don't like about overclocking is how much waiting there is, I can be impatient at times