Phenom 9500 w/ MSI K9A2 Platinum

[justapost]

Idle+Load 24/7 Stability
2.8G was not stable - it froze after 16 hours during WISE. I jumped MHz.

froze after 2hrs here 1.325V is what i use atm.

2.9G was no go at 1.4v real, fails WISE stability after 8 hours although can bench easily.

Behaved unstable here at 1.4V. Prime ran fine for a few minutes but sandra prozessor arithmetic benchmark froze the machine even at 1.425.

1.4v 2.9G did not even boot with CnQ turned on. Disabled it then.
.

Hmm on the DFI mobo I can set CnQ and voltages to Auto and use a cpu multi suitable for stock voltage and the given ref HT.
Voltages and he multi can be modified via msr registers then.
The PC must be in Desktop mode however if ref HT x stock CPU multi exceeds the max the CPU can handle at stock volts.

3G 1.552v idle is 97W DC for CPU, which is what I'm at now. I have tried enabling CnQ and am successfully running it right now at 400MHz 0.7v with 3000MHz 1.552v being the load setting. CnQ gives me 12W DC idle CPU power.

For the record you use the wattages from everest odin readings?

Some [totally unoptimized] 3G benches FYI:-

Always good for a platform comparison.

[Oldguy932]

Well, I just put my 9600 into my crosshair again. I found the nb voltage option that I wasn't using before. Its running 260x11.5 1.3v cpu, 1.35v nb. And to my liking, nb voltage maxes at 1.575. I'll try and get a 3.1ghz shot if I can, but I dunno because of how this cpu hates extra vcore.

[KTE]

Firstly just to clarify mine is not a cherry, its actually a poor one compared to others I know of and have seen. Its more like the average, good ones will do what you see Tony's doing and more. I've just asked about the 3 others at uncles apart from this, all 3 are better than mine.

Stability
3.00GHz failed WISE stability after +7 hours at 1.552v idle
Currently testing this at 1.512v idle / 1.488v load.

Yes DAMMIT 9850 definitely still has the perfectly stable load tested-unstable idle and WISE tested problem at the high end!

EDIT 1/2/3/4/5/6: system just froze during WISE testing for 3055 [35mins], 3000 [432mins], 2990 [86mins], 2975 [25mins], 2915 [14mins], 2860 [9mins] and 2780 [11mins]. Somethings wrong here, I've had 2.9G 1.45v WISE and load stable earlier.. only thing higher now is HT/NB... back to stock HT/NB testings

EDIT 7: No, 2800 didn't fail with WISE and load testing at 1.376v load. That certainly points to HT or NB causing the freezing - this is one good thing about unlocked multi, you have many ways to approach a problem and thus, figure it out

Also I'm currently testing a setting just to check but.. Phenom B2 and B3 has a problem with 1066 mode oc. Stability in this mode is very hard past stock 1066 [this looks like the cause of instability to me - my perfectly fine setting I tested in everything but when I changed from 800 mode to 1066, it crashed during WISE although it was far under bench/MEM limit]

-Max CPU P95/load stable is 3055MHz [furthest tested - 3100 is definitely not] but it will fail WISE stability.
-Max MEM benchable is 638MHz but max stable is 613MHz so far. 9600BE had a 626MHz limit = all on stock MEM volts of 2.2v.
-Max NB stable at stock volts is 2380MHz, nothing more. Same with HT, 2380MHz.

Safe Temperatures
Tested HTC and THERMTRIP temperatures. People are confused about these values but the temp. values you see in EVEREST/AOD/CoreTemp/HWMon are all coming from a CPU register and the maximum Tcontrol_max value is calibrated to the same value as the real Tcase_max for this CPU - which is independently monitored and calibrated within the Thermal Control Circuitry (TCC). Thus, when Tctl_max is reached, Tcase_max is reached, TCC detects this and activates HTC/THERMTRIP to either throttle CPU or shut it down.

Through tests I found Tctl_max is a value in between 84 to 90C in software. Anything above 84C starts erroring but 86C does not shut down. A degrees value very close to it, soon after but before 90C is the shut down max temp. Thus keep it below 75C full loads and you're fine and within CPU stability range

Performance and Tweaks
Just messing around, B3 scored 2590 Single CPU in CB10 at 2990 with tweaks [beat the 3G score above].

WinRAR is 2478KB/s at 3000CPU 240HT 2400NB 638MEM. MEM/Cache b/w and other benches at similar settings can be seen here=>



















Sandra Multi-core b/w has always been very weak for Phenom

2k6 3000 vs 2990 tweaked
vs

Updated EVEREST scores

















I'm not going to push it yet, validator has a bug that its rejecting many different legitimate CPUs of users outright. Once its fixed by Sam/Franck, I'll go into it like a train, I know it wants it, I can hear it scream.... give me 1.7v....

This might help.
It's not Tony's "water park workshop", but should keep the Phenom reasonably cool.
http://www.vr-zone.com/?i=5142

Thanks man but I already have wc if I wanted it - I didn't want to test Phenom for wc, but air, since I'm not keeping it but just playing with it for others to see and help them work it out and these are definitely just budget chips for the masses

froze after 2hrs here 1.325V is what i use atm.

I'm going back to basics to sort the problems out. Testing 200x14 2800 at 1.4v idle / 1.376v load ATM. You have an advantage, your BIOS are far better for bootup oc with all volts and multis, we don't have that on this board so things are a lot harder to work with and oc.

Behaved unstable here at 1.4V. Prime ran fine for a few minutes but sandra prozessor arithmetic benchmark froze the machine even at 1.425.

Last time, 1.45v was what I needed for 200x14.5 2900 stability. My main aim isn't bench stability though, load stability has been a lot easier on my Phenoms than WISE stability

Hmm on the DFI mobo I can set CnQ and voltages to Auto and use a cpu multi suitable for stock voltage and the given ref HT.
Voltages and he multi can be modified via msr registers then.
The PC must be in Desktop mode however if ref HT x stock CPU multi exceeds the max the CPU can handle at stock volts.

By the looks of it, your board changes VIDs where voltages should be changed. The MSI can do CnQ at any voltage/MHz [as long as its stable] and yep, thanks to Sams tool, tweaking is far far far easier - I just apply everything through it

For the record you use the wattages from everest odin readings?

Calibrated Odin current readings with clamp ammeter, so check those, then AC watt meter and DMM for volts, system VAC and then work out power. Have traced each rail back, so I know which rail is powering which component to distinguish between all.

Always good for a platform comparison.

Yup, been cross-checking with yours to see if any of my settings were unstable [when unstable, there will be a performance drop]

Have you tried the 2200 800 comparison we did before yet?
Since we already have B2 perf. known, would be good to add B3 perf. to it too
I'll add Abit 770 and GBT 780G scores to it soon aswell, need to flash a new BIOS on them before that or they don't boot Phenom B3.

Well, I just put my 9600 into my crosshair again. I found the nb voltage option that I wasn't using before. Its running 260x11.5 1.3v cpu, 1.35v nb. And to my liking, nb voltage maxes at 1.575. I'll try and get a 3.1ghz shot if I can, but I dunno because of how this cpu hates extra vcore.

So you can run this setting daily without problems or is it just bench/load stable?
Very good for a 9600 though.

[Oldguy932]

This cpu never had a problem with being idle stable. The BE I have is the only one to ever show problems with that, but it was because it had started to deteriorate. Max boot so far was 268x11.5, not at all stable though. Right now I'm doing everest stability at 262x11.5 with 1.34v cpu and 1.35v nb. I need a 790fx board or 780a board with the tweakability of my crosshair, I miss the memory bandwidth.

Edit: And the 9600 was previously bench stable at 266x11.5 for wprime32m and superpi1m, 264 was where I did other benches I think.

[justapost]

I'm not going to push it yet, validator has a bug that its rejecting many different legitimate CPUs of users outright. Once its fixed by Sam/Franck, I'll go into it like a train, I know it wants it, I can hear it scream.... give me 1.7v....

I want my steak medium please.

I'm going back to basics to sort the problems out. Testing 200x14 2800 at 1.4v idle / 1.376v load ATM. You have an advantage, your BIOS are far better for bootup oc with all volts and multis, we don't have that on this board so things are a lot harder to work with and oc.
Last time, 1.45v was what I needed for 200x14.5 2900 stability. My main aim isn't bench stability though, load stability has been a lot easier on my Phenoms than WISE stability

CPU starts to behave odd for a while after too much heat and voltage lovin.
Benched 2.8GHz/2.0GHz 1066-5-5-5-16-18 today. I get better results with 5-5-5-16-18 than with 5-5-5-15-11 btw also the ram needs ~2.05V now.
Stability testing 2.8GHz/2.0GHz at 1.3125V/1.285V over night, will bench that tomorrow and go for 2.8/2.4 then.
Will post a summary of benchmarks results for 2.8 at different NB speeds then.
After that ill go for 3GHz ref HT wise.

By the looks of it, your board changes VIDs where voltages should be changed. The MSI can do CnQ at any voltage/MHz [as long as its stable] and yep, thanks to Sams tool, tweaking is far far far easier - I just apply everything through it

Slammed mode, VID's are not used only p-state register modification works. Whithout that VID's are reported as stock.

Calibrated Odin current readings with clamp ammeter, so check those, then AC watt meter and DMM for volts, system VAC and then work out power. Have traced each rail back, so I know which rail is powering which component to distinguish between all.

I'm new to those mesurements. Saw that 12V1 is for the cpu only. I guess i can verify at least the volts with a DMM on the rails?

Yup, been cross-checking with yours to see if any of my settings were unstable [when unstable, there will be a performance drop]

Those where all ran in unganged mode.

Have you tried the 2200 800 comparison we did before yet?
Since we already have B2 perf. known, would be good to add B3 perf. to it too
I'll add Abit 770 and GBT 780G scores to it soon aswell, need to flash a new BIOS on them before that or they don't boot Phenom B3.

Expecting the GBT 780G here also. Have not yet tried the 2200 800 comparison but i expect a slight performance drop due to the hardware tlb workaround.

[BrowncoatGR]

Expecting the GBT 780G here also. Have not yet tried the 2200 800 comparison but i expect a slight performance drop due to the hardware tlb workaround.

ΚΤΕ i suppose you are going to OC the Phenom on the 780G? Im really curious how accurate this Anandtech article is. It seems really strange to me. BTW how do you manipulate CnQ settings in windows. Is there a special tool that allows it?

[Brother Esau]

Guys does the latest version of CoreTemp work accurately with Phenom or no please?

[Oldguy932]

So far the max htt has been 268. I think the HT might need more voltage, but I haven't tried anything with it yet because I put the xp-90c on to get the load temps back down. I should really get around to buying a quieter cooler because the vantec tornado is just deafening...

[KTE]

Issues.. again

ID: idle
LD: load

2900MHz... I cannot get it WISE stable from 1.519v ID to 1.45v ID - anymore. It will freeze after 3 to 18 hours uptime. It is fully bench/load stable at 1.45v ID though, make no doubt, just like you saw with B2 - but for WISE, same problem as B2 The way I see it since the beginning is the boards with weak PWM area are the ones giving this problem. We need to approach this with facts, appropriate knowledge and using a grounded mind.

The ASUS/DFI/Sapphire/GBT DQ6 790FX have the best PWM area out of these boards, which can handle larger power draw with far more stability and thus, naturally, more higher watt oc is possible this way - on all boards, you have to cool the PWM area active if you want to oc and throw in +150W power through them to the CPU, which you can easily draw with an oc'd Phenom.

Another thing, you have to verify your PSU has enough juice on the rails (Amps), holding good loadline voltage under these loads, or it'll fail outright at POST.

Compare the Intel boards' PWM area with AMD boards' PWM area; most Intel P35 budget and mainstream boards I've come across have much better PWM area than the AMD 790FX/790X budget boards have been given, and far better than most 770/780 boards have. Why?

The R25 inductors on the MSI board are very weak for oc 125W CPUs - R80 is what you want for good oc. The limit of this 4+1 phase will be very close to 200W, thus you're restricted to oc. All the SMD Inductors are rated for an inductance at a certain frequency, ambient conditions and maximum Adc (amps). The ones on the MSI board seem very likely to be rated at 40Adc max at 25C with 0.18uH inductance at 100kHz. The inductors also have a max amperage beyond which they will begin degrading and break down. As you increase temperature above 25C, the maximum amps they ca handle will decrease and so will their efficiency.

All lower and cheaper chipset boards, i.e. 790X/790GX/780G/770 will on the majority have a weaker PWM design than their 790FX counterparts, so they will:-
a) overheat much with 125W CPU if they support them
b) will not be able to handle much more than 150W even if that with 3+1 phase, any lower will have major trouble not burning if not v.high quality inductors
c) will be under maximum stress and degrade quickly
d) have a high risk of catching fire and burning. The key to limit this is volts, the current with Phenoms is set after a cold reset at bootup by the CPU and you can access that value looking through PVI codes and MSRs and hence work out what power your CPU can pull at max to stay below the threshold.

Volts
Don't freak out at my volts, this board has huge vDroop at high volts. It is much less at lower volts. Like I said, I'm stuck on a bad BIOS for oc, volts are limited. This is how the VCore works out on this board:-

1.519v idle is 1.456v load
1.480v idle is 1.424v load [CPU pulls 203W DC at 2.9G this setting]
1.464v idle is 1.416v load
1.456v idle is 1.416v load

I could not get anything above 2.8G WISE stable yet. Load/bench stable I have had upto 3055MHz. I don't know how to tackle this but will first have to try another stronger PWM board to check since its reacting like the B2s

CPU does not like voltage for 100% WISE stability, loves it for benching though. Around 1.424v idle / 1.376v load seems to be the WISE full stability limit here. ATM I'm not touching RAM or NB, just CPU and I'm testing how stable 2.8G is.

BTW, I've heard conjectures everywhere for a while about Phenom being lower 3D Mark'06 perf. than X2 6400 and so on, well no, it's not and has never been. Even a 4500MHz Wolfdale cannot beat a 2990MHz Phenom in 2k6 CPU. Compare the two: 2990 CPU vs 4500 CPU

This cpu never had a problem with being idle stable. The BE I have is the only one to ever show problems with that, but it was because it had started to deteriorate. Max boot so far was 268x11.5, not at all stable though. Right now I'm doing everest stability at 262x11.5 with 1.34v cpu and 1.35v nb. I need a 790fx board or 780a board with the tweakability of my crosshair, I miss the memory bandwidth.

Edit: And the 9600 was previously bench stable at 266x11.5 for wprime32m and superpi1m, 264 was where I did other benches I think.

Problem why I can't increase HT Ref is due to being stuck at 10x NB multi with no NB multi or NB volts option in BIOS. Over 238MHz HT / 2380HTL / 2380NB is not possible before crashing at stock volts boot, so I'm limited until I change BIOS.

CPU starts to behave odd for a while after too much heat and voltage lovin.

Which PSU are you testing it with? Check your rails, at 2.9G 1.424v full load, I was at 16.1A draw from the CPU 12V1 rail, which is not far from the maximum the rail can handle before OCP makes it shutdown

Benched 2.8GHz/2.0GHz 1066-5-5-5-16-18 today. I get better results with 5-5-5-16-18 than with 5-5-5-15-11 btw also the ram needs ~2.05V now.

Yeah, same as my both sets then. Need 2.05v for 1080 5-5-5-15 Memtest stability. I don't run anything not Memtest stable though, while in windows it seems fine, it will start giving worse problems like OS corruption sooner or later which I don't want to deal with knowing how much-a pain it is. For benching, 638 5-5-5-15-11 2.2v is stable enough.

Stability testing 2.8GHz/2.0GHz at 1.3125V/1.285V over night, will bench that tomorrow and go for 2.8/2.4 then.

That's what I've decided, to test lowest volts 2.8G/2.0G stability today and then move on.

Will post a summary of benchmarks results for 2.8 at different NB speeds then.
After that ill go for 3GHz ref HT wise.

Thanks. If B2 could hit 2.6G NB, B3 should at least do this much, mine has already benched 2380NB at stock volts, the max possible on B2 1.3VID was 2200NB for me before it degraded, it was then 2150NB (not stable but benchable). The minute I flash a better BIOS, I'll experiment on NB alone.

I'm new to those mesurements. Saw that 12V1 is for the cpu only. I guess i can verify at least the volts with a DMM on the rails?

Verify which rail increases current and decreases volts by which component load. The source rails my not be the same as the labeled rails, i.e 12V1 and 12V2 labeled on the PSU and separated outside may only actually be 12V1 combined, so it can get tricky

Yep, for volts DMM will be far more accurate.

ΚΤΕ i suppose you are going to OC the Phenom on the 780G? Im really curious how accurate this Anandtech article is. It seems really strange to me.

Not sure, 780G I bought for X2 4850e and 9100e or 9150e only. The article may be journalism but it has some good points.

First of all, all of those problems in there are true, some to more of an extent than others. My 780i board and replacement 790i board both killed MEM, HDD and MEM perf. like with others.
The 780G/790X/790GX/770 boards with cheap, weak and lower rated PWM section will have problems with high TDP CPUs, without a doubt. I have never considered 9850 etc being feasible for 780G boards for HTPC and such similar low watt uses, it is more an enthusiast CPU with high power, SMD and circuitry requirements. Secondly, the BIOS on such boards won't allow much oc on these CPUs. Thirdly, and most importantly, if you go beyond the inductor max.current limits, you'll have a fried board very quickly and at the least, a very degraded board. The 3-phase designs are not for Phenom above 95W in any way, you're asking for trouble. The 3+1 designs can handle higher, you just have to cool the damn PWM section to try and stay as near 25C as possible.

Some problems with that article is the common problem with geeks: they start passing off assumptions as facts too often when either their own knowledge, resources or experience is weak in a field. Those guys haven't worked with these boards long enough to figure out why. The problem with most of the boards Gary lists including GBT 780G is an overheating PWM area problem rather than a non-supporting problem, all he needs is to keep a fan on them like any experienced overclocker will do second nature when overclocking a CPU or with high wattage CPUs running on lower quality inductors and/or less than 3+1 phase motherboard PWM designs and many will handle 125W CPUs fine.

IMO, even owning a 9850/9750, I will never use it inside that board unless I can find a very good, low volt overclock. That will make it a very powerful offering, I "think" 2.7G 1.25v is a very real possibility for 100% stability on such 780G/790GX boards and that setting will give the 9850 BE a 125.5W maximum rating, which is still well within the board working limit and stock power draw. If you can then underclock to 0.7v 600MHz idle, then power draw/temps would be excellent.

That all said, I'll test the 9850 upto 2.8G 1.4v on the 770 and 780G boards to see how it copes, and if I get "death within seconds" like they're showing.

Another note is, the guys saying they've had "dead" boards which start but keep restarting before POST is the same problem I was getting on 3 boards. It's a cold reset problem rather than dead CPU/MB, like I found after 6 chips.

BTW how do you manipulate CnQ settings in windows. Is there a special tool that allows it?

Using Sams Phenom MSR tool, or worse, MSR Editor in EVEREST/Crystal CUPID

Guys does the latest version of CoreTemp work accurately with Phenom or no please?

Works fine for its intended purpose, read my last post starting section above.
Accurate? That's a subjective term only AMD or such microprocessor MFGs can finalize really, or a professional EE in research. You need to know the Tcase_max for your CPU, when it's reached and what that software reads when it is reached to stay below it, and that we already know.

So far the max htt has been 268. I think the HT might need more voltage, but I haven't tried anything with it yet because I put the xp-90c on to get the load temps back down. I should really get around to buying a quieter cooler because the vantec tornado is just deafening...

I'd just get a Xigmatek S1283 rather than that, very cheap too

[aGeoM]

KTE It is possible to activate AOD Turbo mode from Sam's tool?
Tried all AOD versions, all of them gave me troubles, and I can't use none.
I'm following your work on the 9850 with close attention, as always an 5* work.

[aGeoM]

This might help.
It's not Tony's "water park workshop", but should keep the Phenom reasonably cool.
http://www.vr-zone.com/?i=5142

Check this out.

Review

[SocketMan]

KTE It is possible to activate AOD Turbo mode from Sam's tool?
Tried all AOD versions, all of them gave me troubles, and I can't use none.
I'm following your work on the 9850 with close attention, as always an 5* work.

Once you run the "tlbdisable" tool then aod starts up with the circle "red"
already,on my board anyway.
Like the green light on that Cooler Master kit

[KTE]

KTE It is possible to activate AOD Turbo mode from Sam's tool?
Tried all AOD versions, all of them gave me troubles, and I can't use none.

Which problems?
Which stepping Phenom?

Sam's tool can change any MSR, as long as you know it and edit the correct MSR. When you say "Turbo mode", do you mean AOD Yellow or AOD Red?

Once you run the "tlbdisable" tool then aod starts up with the circle "red"already,on my board anyway.

Is this with the latest AOD version?

With mine, AOD "button color" was always yellow and the patch was still disabled.

[aGeoM]

Once you run the "tlbdisable" tool then aod starts up with the circle "red"
already,on my board anyway.
Like the green light on that Cooler Master kit

Mine,no color in AOD's circle (Vista SP1 x64/Server 2008 Ent).

Seams to be a Watter kit.

[aGeoM]

Which problems?
Which stepping Phenom?

Screen stripes all over, restart system...,just don't work at all...
Still on B2, I'm expecting B3 by Friday (It's cames by German Post, no tracking system, it's the first time I shop in this e-tailer)

Sam's tool can change any MSR, as long as you know it and edit the correct MSR. When you say "Turbo mode", do you mean AOD Yellow or AOD Red?

AOD's Red mode, and I don't know witch MSR register to edit.

Is this with the latest AOD version?

With mine, AOD "button color" was always yellow and the patch was still disabled.

Versions: 2.0.7; 2.0.9; 2.0.12; 2.0.14b; and 2.0.17, tried all.

After Sam's tool applied, I have Yellow circle.

Thanks.

[jonspd]

Kte did you ever try 1.13 with the new chip? or are you still on POJ

[KTE]

I don't know which MSRs to edit to have AOD Red mode, I've not tracked them down or seen them tracked down yet, sorry.

64b definitely still has more problems than 32b. My B2 had the same problem everyone else had... Vista is actually easier than Ubuntu 64b is. Just tried B3...

I'm stuck with no Gfx drivers, so 800x600 which is a PITA, so quickly=>

Stock



2800MHz 100% load (see panel icon) - same settings as XP 32b



I have issues above that but I have those in 32b too.

Jon, still on P0J, have to pick someone up and go tracks 5 minutes ago and on way back I'm going back to other flat to pick up Xig cooler and some TIM to then install 1.13 and try oc. I've gotta be off somewhere now

[jonspd]

let us know about 1.13 please so I know if I need to flash before install of a new chip

and if you get a chance to post up some smp client F@H frame rates around 2.8ghz or better that would be great. If I recall you had ran it before.

[aGeoM]

Thanks KTE, no need for sorry.

[karbonkid]

Can you try running mprime for a little while in 64-bit ubuntu? I'm really considering this board, but I don't use Windows and there's no point getting a board that has stability or compatibility issues with my OS of choice. Also, when you say you 'don't have graphics drivers' do they not work, or do you actually not have them in order to try?

[justapost]

Thank you for the updates KTE,
Nearly finished 2.8GHz benching. 2.4GHz for the NB require lots o volts to be stable. Benched 2.8/2.4 at 1.4/1.4 finaly and even there it froze one time during sandra latency test. Had no issues so far during application usage. Are you testing in CnQ tweaked mode? Tony reported an 7x ht multi helped him alot with stablilty above 3GHz, have you tried that setting?
About linux, i had no problems booting into 64bit debian testing with 32bit winxp stable settings up to 2.6GHz. During a linpack run watching the syslog i saw that some cores stopped responding in time above 2.4GHz.
Tried last week released Xen 4.1 here and it has issues detecting more than one core. The have full virtualisation support on amd platform now so i'm really curious how it will perform. Received the GTB780G mobo today. Equipped it with the 9600BE and 4x1GB G.Skills and will try that mobo now for xen testing. Seems to detect all four cores proper (at least during install right now) and without not responding timeouts. I suspect problems with apic/acpi implementations on the DFI.
Have you noticed lower power requirements from the mem? I think 2.12V was the lowest possible with the B2 at 533-5-5-5-16-18 now they run stable at ~2.0V.

[Oldguy932]

I just upped the HT voltage to 1.5v, so 268 works fine now. Nothing higher is bootable, probably just limitation of the bios cause I'm still using 904 which was the first to support phenoms on the board. Currently running 268x11@1.325v because this board has some vdroop issues and it doesn't like high voltage to get the 1.3v at load that it needs to run 3+ghz. Not sure if that's a board issue or a psu issue, I have the pc power and cooling 510 watt, but I could move the 1kw one over to check.

2 more days till I get my 9850 BE...

[aGeoM]

This is what happens most of the time with AOD.

[KTE]

Well, like I said, I had no cooler or clip to fasten HSF down, nor any spare TIM... stock HSF resting on top was giving me 45C idle and Zalman 9700 fullspeed resting on top was giving me 35C idle but very high load.

So here comes my fav. cooler so far, at 22C Ta, just one HS fan=> Xigmatek HDT-S1283

Idle



5min. Stabilized Load



Remember, my last 9600BE stock was 87W load, much lower TDP. I've measured high power draw for the 9850BE stock alone. It will be very hot like the QX6700 and QX9770/9775 are, so these coolers will have to handle high heat loads.

I'm pretty good with getting temps. down on coolers with different mountings, TIMs and stock fans... but that just beat my Tunic Tower 120 by 5C ID and 4C LD, thermal resistance of around 0.12 C/W => very impressive cooler

And to add, you cannot hear the fan over ambient noise or my GFX fan noise, I have to keep checking EVEREST to see if its even running because I keep thinking its stopped or powered down to 4V

Beware! Damn if you use this cooler, the space on the MB between heatsink right bottom and RAM is extremely minute - you will have to push one fin up to get one RAM module in - laboriously tight fit and removal will also be a pain. Because of this, RAM will run hot as there is no airflow down there, you have the lowest aluminium heatsink fin which heats up, touching the first RAM module by covering its top heat escape route [gap], and hanging over the second RAM module but blocking the fan airflow from going down there; the heat from the RAM HS cannot escape anywhere [tshh tshh.. D9s run very hot and die very quickly by running hot] but the heat is recycled within both RAM modules, heating things up even more, so definitely get a little fan near it. I've just added a temperature controlled 60mm fan there which is at 800RPM and you can't even hear it above the ambient noise; it is running the RAM IC temp at 35C load, which is 39C better than stock, so good for me.

Anyway, for air, this cooler, 1/3 the price of TRUE here, is definitely recommended! Now, finally, the air oc starts

Here's stock Sandra [512MB access] Phenom 9850 BE bandwidth compared - they are our own results, upper two are mine and lower two are Achim's;
Phenom 9850 BE 2.5G 1066 Unganged vs Phenom 9850 BE 2.8G 800 Ganged vs Phenom 9600 BE 2.3G 1066 vs QX6850 2.33G 1066



Here's x264 benchmark [the older one we used before] comparing 9850BE 2.5G with 9600BE 2.6G =>



Check EVEREST PhotoWorxx stock 800 Ganged vs 1066 Unganged

let us know about 1.13 please so I know if I need to flash before install of a new chip

and if you get a chance to post up some smp client F@H frame rates around 2.8ghz or better that would be great. If I recall you had ran it before.

Hmmm... don't have the client installed but I have the bench CD lying around somewhere, HD is in a mess right now because I don't have free time for PCs. I'll check it out but I can't yet as I'm stability testing, apologies.

Can you try running mprime for a little while in 64-bit ubuntu? I'm really considering this board, but I don't use Windows and there's no point getting a board that has stability or compatibility issues with my OS of choice. Also, when you say you 'don't have graphics drivers' do they not work, or do you actually not have them in order to try?

They are provided by Linux, but the install procedure never works.
And it keeps asking me for the install CD when I don't have it anymore. It puts me off booting into it again.

mprime? I'd run it but it's a pain to run anything at those screen settings, it defaults to 65Hz which is way too bad to look at with a 150Hz CRT.

Thank you for the updates KTE,
Nearly finished 2.8GHz benching. 2.4GHz for the NB require lots o volts to be stable. Benched 2.8/2.4 at 1.4/1.4 finaly and even there it froze one time during sandra latency test. Had no issues so far during application usage.

You're welcome - I have benched 2.4GNB, required 1.325VID

My 9600BE required 1.4VID for 2.4G NB stable, but soon, it degraded... I have a strong suspicion that its not the cores that degrade but the IMC LOL

Are you testing in CnQ tweaked mode?

Have done, but not always. If I want to bench, I just click the power icon and switch CnQ off by choosing Max Battery:

Tony reported an 7x ht multi helped him alot with stablilty above 3GHz, have you tried that setting?

Nope, thanks for the tip, I'll try it for sure.

About linux, i had no problems booting into 64bit debian testing with 32bit winxp stable settings up to 2.6GHz. During a linpack run watching the syslog i saw that some cores stopped responding in time above 2.4GHz.
Tried last week released Xen 4.1 here and it has issues detecting more than one core. The have full virtualisation support on amd platform now so i'm really curious how it will perform. Received the GTB780G mobo today. Equipped it with the 9600BE and 4x1GB G.Skills and will try that mobo now for xen testing. Seems to detect all four cores proper (at least during install right now) and without not responding timeouts. I suspect problems with apic/acpi implementations on the DFI.

The cores not responding happened with me too but at anything above 2.6G with the 9850BE. I've not checked for 9850BE, so far it seems fine. I just need to GFX damn drivers working and I'll be happy to test in it

Have you noticed lower power requirements from the mem? I think 2.12V was the lowest possible with the B2 at 533-5-5-5-16-18 now they run stable at ~2.0V.

Hmm.. you know, I always Memtest any setting and never rely on what happens in-windows because windows thinks too much is stable, so it was actually the opposite for me

9600BE did 626 5-5-5-15-11 2.15v stable but 9850BE did not do that setting stable even at 2.3v. 9850BE did not run even 560 that setting stable at 2.2v, Memtest failed at Test#6, second run.

In-Windows, I've had more stability than B2 though, benched 638 5-5-5-15-11 2.2v which was not possible before. To check if it is better, I need higher HT and to test 500 4-4-4-4 1T 2.2v. B2 never had this stable but could easily bench it. Still it isn't really stable setting for me, just not crashing yet, Memtest shows it with many errors.

My this set of D9 is good, I can drop to 1.8v real 1066 5-5-5-15-13 and bench it in-windows perfectly fine - but it won't be Memtest stable, it needs 2.05v for that.

I just upped the HT voltage to 1.5v, so 268 works fine now. Nothing higher is bootable, probably just limitation of the bios cause I'm still using 904 which was the first to support phenoms on the board.

1.5v HT voltage is needed? That's a first that I've seen, I have benched 2596MHz HT at stock 1.2v and even 260-274HT at that voltage across quite a few Phenoms

Why don't you change to a better BIOS if BIOS is the limitation?

Currently running 268x11@1.325v because this board has some vdroop issues and it doesn't like high voltage to get the 1.3v at load that it needs to run 3+ghz. Not sure if that's a board issue or a psu issue, I have the pc power and cooling 510 watt, but I could move the 1kw one over to check.

I might test this chip in the ASUS 790FX again very soon, just to see if max stable, max bench, max screen is worth the 2.5x price it has over the MSI. For volts, just use a DMM and check idle-load rails voltages.

This is what happens most of the time with AOD.

Wow, fireworks?

Seems you have a compatibility issue with it. I get that if I start RMClock, computer shuts down instantly, right that second!

I have no idea why that could be, quite frankly. check C:\Program Files\AMD\OverDrive folder for the logs. They should have errors in them. If they do, post those errors back, we can look over them and Sami may look over it and be able to advise you better

Hint:-
If you want AOD to start quickly and with bare minimum needed to oc, then go into C:\Program Files\AMD\OverDrive and cut/paste the file PageSettings.xml to a safe folder outside of this folder [as I've shown earlier].

Now go back into C:\Program Files\AMD\OverDrive and move the file PageSettings.xml I'm attaching with this post in there.

Now tell me how quick and trouble free AOD opens

Vista, just find the directory for the same file and do the same procedure.

[KTE]

I just noticed....

AOD has NB VID option working!

Great work, finally, in-windows oc is now possible if you know what you're doing.

BTW 9850BE stock 1066 Unganged, WinRAR is around 2100KiB/s.

EDIT: not sure I've said it before but there is also 8450, 8650, 8750 to be releasing anytime this month. 2.1G, 2.3G, 2.4G B3 X3s.