**Official DFI LanParty UT 790FX-M2R Review/Overclock/Guide Thread**

[KTE]

I pointed what I mean out about two weeks ago with my last 9600 BE. Look here=> http://www.xtremesystems.org/forums/...&postcount=736



Froze straight after that.

lemme understand this better... u talking about 2.6GHz overclock right? and at idle a core might speed to 2.7GHz????

At any core clock, even just idling at 2.35G.

Hmmm thats interesting infor KTE. I had my 9600 BE running at 2.7Ghz stable at 1.262v. Would boot fine, ran prime 95 stable for 24 hours. Stable in all benchies, like 3dmark06 and pcmark05. But, every once in a while after restarting for unknown reasons when I would go to start up Coretemp, or Everest to monitor temps it would lock those programs and refuse to load them. Granted that was also on 1.2 Bios on the k9a2 plat, so any gain from being at 2.7 was wiped out by the TLB fix. Rolled back to 1.13 to get away from TLB fix, and ended up having to knock the OC down to 2.5 at 1.25v for full stability.

I'm on 1.13 BETA, it gives me best results TBH. But I'll try for some high bandwidth dependent numbers by rolling over to P0E next. Sami showed 211 x 12.5 CPU, 211 x 12 IMC/HT and 211 x 2.66 RAM and I know this chip does 216 x 12.5 CPU, 216 x 12 NB/HT and 216 x 2.66 RAM easy. Imagine CR1 at 1148 with low volts.. would be nice don't you think?

Someone on Toms hardware has a theory that part of the Phenom's issue is a possible hotspot in the die, revolving around either the Northbridge or L3 cache. I do believe that the individual core readings that coretemp shows are correct. Under full load at 2.7 my cores were staying below 31c, but CPU temp according to everest was running and staying below 40c full load on all 4 cores.. So that temp difference between the cores and CPU has to be coming from somewhere.

The monitoring apps read from the sub-IHS temp probes AFAIK but the highest internal core temp reading is what EVEREST labels as "CPU" and that has to stay <70C.
Just know for sure, that idling, on air/water (not chilled), the CPU cannot be ambient temps, as an IC will current running through and resistance. On just air cooling, 95W TDP, even 45W TDP, with CnQ enabled (which IIRC throws TDP down to 60W on Phenom) the temps cannot be less than ~+8C above heatsink ambient (not room ambient, but heatsink ambient which is hotter than case/room). These are IC limitations once there is TDP. The RD790 chip has a 10W TDP and at 22C ambient it will be idling at plus 28C, so you can judge next to that for accurate/inaccurate values.

Not sure what is happening with temp values nowadays, we know Intel has big problems with its Penryn line probes mostly faulty but some are sounding ludicrous. I saw someone boasting about 1.65V 4.5G full load on normal air at 40C. I mean, yeah, 5W TDP chip, fair enough, but at >180W TDP? That honestly made me wonder is wrong here.

can you uninstall your phenom and cnq driver? hope that fixed freezing issue.

Why should this cause the problem if CnQ is BIOS disabled tictac?

I installed the driver after my freezing actually.

oh if you open more tham one program that do i/o operation like coretemp and everest that no good. just use one program ok?

100% true.

Opening CPUZ/EVEREST/CoreTemp along with AOD will and can cause this problem. That's why I only try and open one. Worst is if you have any of them open and try to open AOD. That'll usually throw the PLL haywire up fast.

[tictac]

ok .. lets see how high your phenom can clock stable in windows safe mode.

[KTE]

ok .. lets see how high your phenom can clock stable in windows safe mode.

Well, for me, we have problems as Safe Mode puts 2-3 minutes of 70-100&#37; processor load whilst booting and doesn't let any driver/app work to show frequency.

However, I tried it just now. Same limit as before -> 2756 and below stable and sub-2840 into Windows capturable. Booted this pretty easy first time.

[Rammsteiner]

Yep, it can increase oc/stability. Just have a high MaxAsyncLatency value and you should be good. If drive strength is incorrect, the system should tell you at bootup when it performs the calculations.

Already tried high MAL but well, about 10ns. That booted up fine however stability was a complete other thing not showing its face.

Well, the DFI BIOS at least has all those Drive Strength options like in the AOD. I saw a post of ozzimark somewhere (as he had the 4x 1GB record in the DDR2 database for AMD). He used some sort of program to change Drive Strength however Ive no clue what the calculations are behind those settings.

And how would the system tell me at boot up if calculations are incorrect? post codes? Like I got to reset my CMOS? Or more as Im about to start Windows, detecting IDE drives etc? Is there anywhere some sort of guide for it?

And as I said, I heard D9 can get killed pretty fast when Drive Strength is way too high/strong, so Ive really no clue but I'd like to learn about it. The fact I got it to run 293Mhz and do 1M SuperPI calculations and a few Everest tests tell me it's somehow possible to keep that speed, but I'd like to keep the V's a little bit reasonable though.

Also, as far as I understood, Drive Strength etc doesnt decrease performance but is like PCI-e clocks right? It can increase stability but if the frequency is too high/wrong it can kill the cards/DIMMs in it?

[Rammsteiner]

Ok, I tried setting Drive Strength to weak mode and the 3 strengths of 1.50x to 1.0x. Posted however didnt go beyond 'FF'. Set the changed strengths to 1.25x instead and it booted and ran 32M SuperPI with 2.3V . DIMM's weren't very hot, warm though, I guess about 30~40C. Gonna play some more later on but Drive Strengths are awesome obviously.

Just tried 2.25V but that wasn't doing it, got the SQR error in SuperPI. However maybe a little more Drive Strength tweaking will solve it. At least I know one of the changed settings doesnt like to be at 1.0x. So Ill trial and error wise try to find the one(s). Also the original 1.0x Ill try to start at 0.75x.

Less Voltage and less Drive Strengths seem to be a sweet spot, at least for D9GMH. Voltage isnt much of a problem for me though, more worried about Drive Strengths as they can kill D9's if they're way too tight.

FYI, Ive a 80CFM 92mm fan pointed towards the DIMM's from ~10cm and an unknown CFM 120mm fan 10cm above the DIMM's sucking hot air out of the case. So Ill try and get better results.

I dont feel like changing the Ohm settings though as that seems a little bit dangerous.

AFAIK lowering Drive Strengths doesnt hurt performance, not sure about it really. Does anyone know?

[Brother Esau]

No actually in some instances it increases it due to the way the memory interact s and the the way the data is transmitted to the Memory controller in the cpu.

It also could be that Transend Memory does not completely play well with DFI Motherboards that just may be the case as DFI Boards are very picky with Memory!

[prostoman]

Ok, I tried setting Drive Strength to weak mode and the 3 strengths of 1.50x to 1.0x. Posted however didnt go beyond 'FF'. Set the changed strengths to 1.25x instead and it booted and ran 32M SuperPI with 2.3V . DIMM's weren't very hot, warm though, I guess about 30~40C. Gonna play some more later on but Drive Strengths are awesome obviously.

Just tried 2.25V but that wasn't doing it, got the SQR error in SuperPI. However maybe a little more Drive Strength tweaking will solve it. At least I know one of the changed settings doesnt like to be at 1.0x. So Ill trial and error wise try to find the one(s). Also the original 1.0x Ill try to start at 0.75x.

Less Voltage and less Drive Strengths seem to be a sweet spot, at least for D9GMH. Voltage isnt much of a problem for me though, more worried about Drive Strengths as they can kill D9's if they're way too tight.

FYI, Ive a 80CFM 92mm fan pointed towards the DIMM's from ~10cm and an unknown CFM 120mm fan 10cm above the DIMM's sucking hot air out of the case. So Ill try and get better results.

I dont feel like changing the Ohm settings though as that seems a little bit dangerous.

AFAIK lowering Drive Strengths doesnt hurt performance, not sure about it really. Does anyone know?

Not sure if this will help anyone, but my experience from playing with Drive strenghts on my Lanparty nf590SLi board, was that by changing the Drive Strength values, my memory was more stable overclocked further in MHz but worse timings, or that my memory would be more stable with tight timings, but much lower overcklock in MHz (same Voltage). Long time since I played with those values, so I dont remember exactly what made the difference for what.

I hope somebody got the point

[Rammsteiner]

No actually in some instances it increases it due to the way the memory interact s and the the way the data is transmitted to the Memory controller in the cpu.

You mean the Drive Strength or the performance? Cause increasing the Drive Strength is the thing Im just trying not to do. If it's about performance, yeah. Certain lower timings dont automatically translate to better performance. When I had the A64 3400+ Clawhammer with OCZ PC3700EB I remember tRAS lower than 10 and higher than 10 gave worse performance than tRAS of 10.

It also could be that Transend Memory does not completely play well with DFI Motherboards that just may be the case as DFI Boards are very picky with Memory!

Haha, yeah, DFI is quite picky, especially combined with AMD CPU's. But with some tweaking things gets less worse

Im now running back at 2.3V again and set all Strengths to 1.0x apart from one, was set to 1.25x (can't remember which one anymore though) however it looks like I picked the right one as it booted up. Will post later on which one I changed.

However, Im thinking right if I set all those settings lower the Drive Strength becomes lower too right? I mean, it looks like a multiplier so lower multiplier would give lower... yeah lower... wtf Ive no clue.

Maybe Drive Strength is more about the Ampere trough the slots? Higher Volt with same Ampere gives more Watt which might be the thing killing D9's. If Ampere is lower Watt's will be reduced too. But Ive no clue on that. I just would like to know whether Im actually decreasing or increasing Drive Strengths the way Im changing things now. At least I set the DRAM Drivers Weak Mode to Weak, that's for sure weak then

[Iketh]

Hmmm thats interesting infor KTE. I had my 9600 BE running at 2.7Ghz stable at 1.262v. Would boot fine, ran prime 95 stable for 24 hours. Stable in all benchies, like 3dmark06 and pcmark05. But, every once in a while after restarting for unknown reasons when I would go to start up Coretemp, or Everest to monitor temps it would lock those programs and refuse to load them. Granted that was also on 1.2 Bios on the k9a2 plat, so any gain from being at 2.7 was wiped out by the TLB fix. Rolled back to 1.13 to get away from TLB fix, and ended up having to knock the OC down to 2.5 at 1.25v for full stability.

Someone on Toms hardware has a theory that part of the Phenom's issue is a possible hotspot in the die, revolving around either the Northbridge or L3 cache. I do believe that the individual core readings that coretemp shows are correct. Under full load at 2.7 my cores were staying below 31c, but CPU temp according to everest was running and staying below 40c full load on all 4 cores.. So that temp difference between the cores and CPU has to be coming from somewhere.

phenom does NOT have individual core temp capability, ty i know... one diode for the entire package

[KTE]

phenom does NOT have individual core temp capability, ty i know... one diode for the entire package

It has 14 temp probes actually but for some reason, only one beneath the IHS and one internal core is gauged by software. That's why all Phenom core temps are always in sync.

[Brother Esau]

You mean the Drive Strength or the performance? Cause increasing the Drive Strength is the thing Im just trying not to do. If it's about performance, yeah. Certain lower timings dont automatically translate to better performance. When I had the A64 3400+ Clawhammer with OCZ PC3700EB I remember tRAS lower than 10 and higher than 10 gave worse performance than tRAS of 10.


Haha, yeah, DFI is quite picky, especially combined with AMD CPU's. But with some tweaking things gets less worse

Im now running back at 2.3V again and set all Strengths to 1.0x apart from one, was set to 1.25x (can't remember which one anymore though) however it looks like I picked the right one as it booted up. Will post later on which one I changed.

However, Im thinking right if I set all those settings lower the Drive Strength becomes lower too right? I mean, it looks like a multiplier so lower multiplier would give lower... yeah lower... wtf Ive no clue.

Maybe Drive Strength is more about the Ampere trough the slots? Higher Volt with same Ampere gives more Watt which might be the thing killing D9's. If Ampere is lower Watt's will be reduced too. But Ive no clue on that. I just would like to know whether Im actually decreasing or increasing Drive Strengths the way Im changing things now. At least I set the DRAM Drivers Weak Mode to Weak, that's for sure weak then

Well its like this .....If you feed High Voltage to you're Ram in return you must give your CPU more Voltage as well. So in essence AMD processors "Prefer" Lower voltage memory thats just the way it works perhaps a thorough reading and understanding of the Architecture would be a wise move for you

YOU Must learn How the Memory Communicates with the Memory Controller if you want any success on AMD Platform without damaging you're components!


Balance is the solution Wood Pecker , not Voltage!

[Rammsteiner]

I found a PDF of AM2 CPU's with a lot of info. Also how RAM manufacturers should program the modules to work with AM2. However the fun part is it's all written in strange program language and not making any sense to me.

[KTE]

Are you talking about the BKDG?

[Iketh]

drive strengths increase and decrease performance most of the time... i used winrar's built-in benchmark when playing with drive strengths on my asus m2n-e, it's possible for one setting adjustment to give +30MB/s and vice-versa

[Rammsteiner]

Are you talking about the BKDG?

No idea, I mean this one:
http://www.amd.com/us-en/assets/cont...docs/32559.pdf

And it makes no sense. I lost any clue right after Cas Latency lol.

At this point I hate Intel. Normally there was a guide for every AMD fart after 3 minutes of a released product and now there isn't even a slight little explanation of those Drive Settings. I know the 790FX isn't that old however those timings excisted already on the nForce 590SLI, and that one has been out for a long time.

[KTE]

Yeah it's the BKDG but not written for you but for BIOS and software developers, so it makes perfect sense to them.

[Rammsteiner]

Well, it's actually quite interesting as far as I can understand it.

Also found this, looked at the wrong chapter previously.

Field Descriptions
CKE Drive Strength (CkeDrvStren)—Bits 1–0. This field is used to select the drive strength of the
CKE pins.
00b = 1.0x.
01b = 1.25x
10b = 1.5x (Default)
11b = 2.0x
CS/ODT Drive Strength (CsOdtDrvStren)—Bits 5–4. This field is used to select the drive strength
of the CS and ODT pins.
00b = 1.0x.
01b = 1.25x
10b = 1.5x (Default)
11b = 2.0x
Address/Command Drive Strength (AddrCmdDrvStren)—Bits 9–8. This field is used to select the
drive strength of the address, RAS, CAS, WE, bank and parity pins.
00b = 1.0x.
01b = 1.25x
10b = 1.5x (Default)
11b = 2.0x
MEMCLK Drive Strength (ClkDrvStren)—Bits 13–12. This field is used to select the drive
strength of the MEMCLK pins.
00b = 0.75x.
01b = 1.0x (Default)
10b = 1.25x
11b = 1.5x
Data Drive Strength (DataDrvStren)—Bits 17–16. This field is used to select the drive strength of
the Data pins.
00b = 0.75x.
01b = 1.0x (Default)
10b = 1.25x
11b = 1.5x
DQS Drive Strength (DqsDrvStren)—Bits 21–20. This field is used to select the drive strength of
the DQS pins.
00b = 0.75x.
01b = 1.0x (Default)
10b = 1.25x
11b = 1.5x
Note: The DM[8:0] and DQS[17:9] functions share pins on the DIMM connector. The function
selection is applied based on whether the DIMM is populated with by-4 (x4) DRAM devices,
in which case the DQS[17:9] function is applied, or not (x8 or x16 DRAM devices), in which
case the DM[8:0] function is applied. However, the DM function is associated with the data
pin group and should therefore be controlled DataDrvStren. While the processor supports
concurrent population of x4 and non-x4 DIMMs, the determination as to which field controls
the drive strength of these pins is applied statically based on these rules:
- If all DIMMs of a channel are populated with non-x4 devices, DataDrvStren is applied.
- If any DIMMs of a channel is populated with x4 devices, DqsDrvStren is applied.
Processor On-die Termination (ProcOdt)—Bits 29–28. This field is used to select the resistance of
the on-die termination resistors.
00b = 300 ohms +/- 20&#37;
01b = 150 ohms +/- 20%.
10b = 75 ohms +/- 20%.
11b = Reserved.

So AFAIK reducing On-Die Termination reduces resistance of some sort of power. I guess that power goes to/is available for the DIMMs
Adress/Command Drive Strength if you want tighter timings, if you increase the multiplier you might get just that little bit better stability with timings
MEMCLK Drive Strength increasing might give you a little more stability by increasing the multiplier for the last few Mhz
etc
Or am I on the wrong side and should it be the other way around? Decrease the multi's?

[Rammsteiner]

Thus to go short for the little I understand, Addr/Cmd Drive Strength could be set lower than default 1.50x, MEMCLK might be tried a little higher to 2.0x to increase stability in high Mhz. Data Drive Strength has maybe something to do with 1T/2T timings, could be set lower eventually.

CS/ODT Drive Strength and CKE are still a big questionmark for me but Im sure Ill find it tomorrow somewhere.

And On-Die Termination... meh, still dont feel like playing with that really.

And after those 6 Drive Strength settings, why is there still the DRAM Drivers Weak Mode

[Rammsteiner]

Allright, I searched for the D9GMH B6-3 datasheet from Micron and found out the following

CS/ODT Drive Strength is about the resistance for all sorts of DQS parameters, varying from 50, 75 and 150Ohm. However the BIOS setting on this motherboard is about the CPU ODT. Also I dont know if the multipliers effect the resistance or actually the Ampere so resistance gets reduced.

CKE Drive Strength is some sort of control thing on the DDR2 for about everything which the Memory Controller will receive.

[Brother Esau]

Announcement:....D.F.I Recalls / Removes...... 1/03/2008 Official Bios

Here's the Post from RGone .....@DFIClub

[sonikbuzz]

So... i tested the 1/03/2008 bios and finally i get a nice 385 Mhz bus speed




But i can't get something like 8.5x385 perhaps with the next bios...

[KTE]

Thus to go short for the little I understand, Addr/Cmd Drive Strength could be set lower than default 1.50x, MEMCLK might be tried a little higher to 2.0x to increase stability in high Mhz. Data Drive Strength has maybe something to do with 1T/2T timings, could be set lower eventually.

This is a useful read into its technical aspects: http://www.altera.com/literature/an/an408.pdf

Mess about with MaxAsyncLatency instead of Drive Strength, you're wasting your time.
DRAM Drive Strength is maximum set by default on most boards but traditionally the DFI sets them very low (weak) which was good for Samsung TCCD and nothing else that I know recall. This RD790 frm DFI sets it pretty high up. It's for DRAM stability when more than one DIMM exists especially under loading. Most RAMs will like strong DDS which let's you attain higher clock speeds usually. This is DRAM Drive Strength, different to Data Drive Strength, which loves high signal strength and weaker values.

[Rammsteiner]

ffs, Read the whole CPU PDF and Micron PDF and then this. I decided this afternoon already to quit it and set everything back as they were, Normal Drive Strength and 1.00x to the top three ones and 1.50x to the lower three ones. Especially after I booted into Windows but everything flipped and I got a warning about HDD corruption and possible registries being deleted.

I decided to stay with the current settings, 4-4-4-10 with 2.25V @ 502Mhz. Does the work fine and does after all even less than a 500MB/s transfer less than the 586Mhz 5-5-5-15. So Im fine with this now I guess.

Thx though KTE for linking that, makes a lot more sense with less pages!

And new DFI BIOS... 1/30... w0000t, curious what that will bring us.

[Brother Esau]

Yes I am very Glad that DFI decided to listen to all that has been posted here
Remeber Kids...... Church Mice Get nothing thats why I encourage all of you to get involved and give you're feedback on the DFI Bios for this board as it takes more than one person to Rock the Boat Catch my Drift? (No Pun Intended)

[Brother Esau]

Announcement: .......""NEW" ...1/30/2008 ..Beta Bios Test the Bios at you're own Risk! Sorry guys but I am getting ready next week to move back East and there is no Physical way I can be there to help in the event of a Bad Flash due to the move!

Campbell

EDIT:.....OK...So far no C1 Error , No Problems with NIC Latency . No Problems with Ram Latency with Internal Values set incorrectly in DRAM Table , Different Bios Layout ...Me Likes so far , and nothing ATM seems bad

GO AHEAD BOYS Flash to new BETA