Phenom 9500 w/ MSI K9A2 Platinum

[KTE]

I raised the VID to 1.40 (@ 13x), it just continually ran the Windows boot screen but it didn't crash. It seemed like progress, so I slowly inched it up to 1.45 VID same result each time even after a couple minutes of waiting (HD activity stopped).
I finally gave it one shot @ 1.475 Vid... That resulted in a Blue Screen "Interupt Not Received on Secondary Core"

Yeah, that's a typical sign of the core not being able to physically do the MHz. All of mine will show the same at their limit and the limit comes instantly. My 2nd 9600 had a similar limit, 2602 was max boot but stability was around 2495 max.

Yeah, NB VID was @ 1.35, and it was stable enough to bench as long as it was at the same multi as the cores it was good up to 2409Mhz. The latency times on the L3 cache actually didn't improve from 2000 though, strange.

In which application?
Real-life improves pretty well you know.

Doesn't POH have the TLB patch?

Can disable it in the BIOS.

Actually AOD doesn't seem to work too bad with Vista64, some of the voltage adjustments get squirley, but if you make sure they change and only do 1 at a time it's liveable...

Are you sure the voltages are "squirely" even at 1.25VID?

When you increase the VID, the voltages won't be what they read in AOD/BIOS. The voltages are mapped according to the VID, so 1.25VID gets minimum 1.232V but 1.3VID may show up as 1.232/1.24 selectable, but it will be much higher up by design.

Bad luck landing a bad chip I guess.

OC using 10x, 11x and 12x multi.. see how far you get.

BTW Newegg is expensive for me too...

[jonspd]

I just hope the b3's when they are released hit 3.0ghz out the gate on stock cooler. If that is the case we can hope for around 3.2 which would be a great step up form the 2.6ish we are seeing.


I wish you luck dave as we all know it comes and goes and for sure takes alot more to clock one of these then your ordinary q6600.

I did notice my cpu needs .03 more vcore to run 246 more mhz on the nb at same cpu clock. I guess the reason for the vcore isn't just to control the cpu clockability but to also control the nb clocking which doesn't make much since if you as me the nb vid should be used for that but the more strain you put on the cpu the more power it needs...



I just wish we had a nice bios modder around here helping us even if it was in a different language it would do wonders I'm sure as I remember very much so on a few platforms that just the modded bios could give you around 100mhz on the older cpus...


I think if there was a way to X out core 3 we could see alot better numbers even tho the multi threaded apps would lose on there part.

BTW Newegg is expensive for me too...

so where is cheaper ?

[Daveburt714]

Does POH completly disable the patch though (ie. no performance hit)?

I understand the concept of the VID settup now, I was talking about when you change the v's in AOD, sometimes you have to go up 2 bumps to get 1 etc... And I never try to change 2 different v's at the same time, thats just asking for problems in Vista64...

As for NewEgg being expensive, we don't all have an Uncle in the business...

I'll keep ya posted, I might be able to squeeze 20-30 more Mhz out of this one compared to my old chip.... Geeez, I've never been lucky in the chip lottery (well maybe my 3.0Ghz Opty, but thats it!)...

[jonspd]

I thought everyone had a 3ghz opty

jking

[Daveburt714]

I thought everyone had a 3ghz opty

jking

You think thats funny man... I had to buy a 185 to get a 3Ghz Opty...

I really think I bought the same chip I had before, I've done massive testing on this tonight... Looks like I'm gonna end up with the same specs as my std 9600.

250x10x8x8, So far even 9x on the NB is unreachable and thats taking the NB Vid all the way up to 1.35v's...

Like Rodney Dangerfield, I get no respect/luck...

Sigh.....

[KTE]

I just hope the b3's when they are released hit 3.0ghz out the gate on stock cooler. If that is the case we can hope for around 3.2 which would be a great step up form the 2.6ish we are seeing.

That's a big difference you're asking for there, +400-500MHz... to be frank, I don't have similar expectations at all. I think running 2.7G stable on the unlocked multi B3 for many is best I'm hoping for (for others). Firstly because these speeds are mainly a core design limitation rather than heat/voltage/process, so Intel B3>G0 scenario doesn't apply. Secondly, because with speed increases, the core Voltage/Amps and thus Power and Heat Dissipated will increase quite much. At present, there is no room for anything even plus 2.4GHz to be accepted with the current TDP's, let alone above the 9900 TDP. Thirdly and most importantly, AMD is too stuck with Barcelona and 45nm for much with Phenoms now. I see this as the best option rather than waste time over a core you'll make ultimate losses on to set you a year back in debt. I do expect a 2.8GHz FX release though, most likely late Q2. AMD is known to improve TDP/clock speeds much even with very old and limited cores and process technology compared to what they first launch. Their stepping tweaks traditionally offer good advancement in oc headroom, so who knows. It took Intel 45nm to be able to retail a dual C2 plus 3GHz.. let's see what it takes AMD to hit 3GHz and above on dual, tri and quad with this core arch., if ever.

I'm the bookie, I'm wanting a large handle here... place your bets on a 3GHz K10h CPU (X2/X3/X4) release before August '08 now !!

1000/1?

I did notice my cpu needs .03 more vcore to run 246 more mhz on the nb at same cpu clock. I guess the reason for the vcore isn't just to control the cpu clockability but to also control the nb clocking which doesn't make much since if you as me the nb vid should be used for that but the more strain you put on the cpu the more power it needs...

Make sure of what you're saying here .. because I can tell you officially that NB voltage doesn't affect the CPU voltage/clock speeds or vice versa, just as NB PCIe voltage does not affect them and how NB HT voltage also does not affect any of them.

Try doing some more long stability testing and use 1.13 BETA for it, it'll give you more clear answers. I can run 2300/2400 with 1.184V/1.325VID 100% stable (no freezing etc) just as I can run 2500/2400 with 1.272V/1.325VID 100% stable. Same with PCIe and HT clocking. The voltages fed to the NB are coming from a different source, separate VR's. All of these components being within the CPU still won't allow the voltage at the Vdd pins to affect one another. Would be a weird happening though, don't you think?

I just wish we had a nice bios modder around here helping us even if it was in a different language it would do wonders I'm sure as I remember very much so on a few platforms that just the modded bios could give you around 100mhz on the older cpus...

If you study BIOS modding you'll realize it's no go with MSI/AMI BIOSes so far. You can change some options but they still don't appear after the flash. No usefully successful reverse engineering yet.

I think if there was a way to X out core 3 we could see alot better numbers even tho the multi threaded apps would lose on there part.

This should be MSI's job to do as ASUS has this. Many using ASUS RD790 boards post and bench higher oc's on one/two cores, since they can individually clock cores and leave the others low. If you can have say 3GHz on 2 cores and 2.6GHz on the other two, then I think that's a pretty successful OC outcome and bodes better for AMD. Right now, I would say average stable is;

9500 => 2.5-2.75G
9600 => 2.55G-2.65G
9600 BE => 2.6G

so where is cheaper ?

No where really.

Coming from US and UK/EU being such a v.strong economy and currency, against the USD especially, it's double the price for me then what you pay. So when you have it for $240, I, here, get it cheapest bargain hunted for $305 (yes through my uncle or I'd have to pay $330 cheapest ). AMD's getting away cheaply right now with Intel 45nm facing repeated delays, quad-core bug, volume production problems and price increases since August hype and worst, along with lower FSB than 65nm, lower top voltage limit and thus overall lower clocks for marginal benefits in most day to day situations. Otherwise at these prices, AMD wouldn't be selling much if Intel 45nm quads and even duals could even be available in proper supply at reasonable prices as they had initially advertised for months, and AMD would not be able to sustain this lineup MSRP (~$280) nor the ASP. Moving to 45nm is expensive for a while before you break-even since the tooling costs are extremely high, hence Intel isn't really in any decent situation to sell 45nm at cost or near cost price and nor is AMD with their 65nm. Right now the Q6600 looks near to as cheap as it can be for 65nm yet without cutting into the good gross profit margins but you can see it for yourself; if C2Q 2.4G was selling for $1110-880 in late Jan. '07, and around $530 by May '07, then compare that to $240 for Phenom 9600 BE, the top model. Massive price difference for starters, and yet AMD's 65nm production is much more costly overall, so in effect they needed higher prices not 3/5 lower. It doesn't give AMD's product much headroom or shelf-life either, and as you have seen the depreciation on Phenom's is massive, very quick value dropping. Intel's way is the way to cover any losses in your various lineups and make large profits at the beginning; you have to be early into the market and beat your competitor, then you sit back, price dearly and enjoy. Lack of competition allows this pricing to happen. If AMD had released this core by even May at $350 (gives them far better room for price cuts), Intel wouldn't have been allowed to carry any of that pricing, neither the major profits, and it's common to do worse under extreme pressure and little time and do very well with ample time and little pressure if the company keeps working. As of now I don't see any useful market competition.

Does POH completly disable the patch though (ie. no performance hit)?

Only on the 1st core, the other 3 will have to be disabled manually through using CrystalCPUID or RW-everything.

I understand the concept of the VID settup now, I was talking about when you change the v's in AOD, sometimes you have to go up 2 bumps to get 1 etc...

Yep, this is also what I referred to: each VID has a set of fixed voltages it can allow (something like 5 or 6), nothing else is possible although AOD still shows you the whole range of voltages. These you can check out by setting a VID and choosing a voltage in the BIOS, and then booting up. They will always remain fixed for each VID (tiny fluctuations are VRM/PSU based). So the various VIDs have say for instance 5 different voltages you can choose. By choosing anything in between these voltage ranges within AOD, will make no change unless you're very close to the next actual available voltage according to the VID. Get me?

Experiment and you'll see.

As for NewEgg being expensive, we don't all have an Uncle in the business...

You know I buy personal systems for myself aswell. Two are back at home and one X2 BE system half complete here. Waiting for a 780G for that.

And for me, paying +$200 for a CPU is dear... I scavenge by parts very carefully.

I'll keep ya posted, I might be able to squeeze 20-30 more Mhz out of this one compared to my old chip.... Geeez, I've never been lucky in the chip lottery (well maybe my 3.0Ghz Opty, but thats it!)...

Keep going yet.. a better BIOS will help at least a little.

250x10x8x8, So far even 9x on the NB is unreachable and thats taking the NB Vid all the way up to 1.35v's...

Set 12x CPU multiplier, 1.3625VID on NB, drop HT Ref. to 205 and choose 12x NB multiplier... what do you get?

Can you drop NB multiplier to 8x with a BE?

Both of mine haven't, which was a BIOS bug.

I've been running 2640 now for over 20 hours without fail, from the only failure it will see: the random idling freezing.

[jonspd]

Well I was going on the comparison of our platform's max speed to that of other boards. Seeing some dfi users getting around 2.7-8 then it would only be a 200-300 mhz gain on the new version

[KTE]

I'm topping the Orb any minute right now with a Phenom. Watch those data transfer limits fly out of the window.

Don't want to keep you waiting too long...

1) Check it out: http://www.orb.com/
2) AMD Live On Demand: http://www.amdlive.com/gb-en/about_features.aspx
3) AMD KidRocket: http://kidrocket.org/amd/
4) AMD Streamload: http://amd.streamload.com/

Check the pics of the setup and benches out here=> (hoping they work)
1) http://amd.orb.com/orb/data/download...9ffLDD&l=KTE-T
3) http://amd.orb.com/orb/data/download...HbOINl&l=KTE-T
2)
Benched XP SP2 vs Vista SP1:
http://amd.orb.com/orb/data/download...sbWm36&l=KTE-T
http://amd.orb.com/orb/data/download...jESt5F&l=KTE-T
http://amd.orb.com/orb/data/download...EDxkch&l=KTE-T
4) 25GB

What did you think I was talking about?

Those are some new AMD ventures. Easy way to send/receive/access anything on your devices for free and very hassle free too, logs in auto if setup and works very well with 25GB online free storage. Only AMD processors can run this all by the way and it's free. My uncles little son just played his fav. game using KidRocket. Quite nifty, he loved it. It's many idea's around combined into one.

Only downside from initial trials was... the loading/uploading is a little slower than I would like.

[misteroadster]

Hello all, i'm actually testing any 9600BE, and i'm experiencing a strange Phenomen

My 9600BE overclocks better when it's hot.
With my Thermaright 90 ultra with fan 13x200 is a no go, vista won't boot (cpu temp is around 40° idle).
When i unplug the fan the temp is raising to 70° idle so i can boot @ 2.6ghz
The best frequency i can boot is around 100° idle, and vista boot @ 2.7Ghz.
Anyone can try to increase the temps and see if the overclock is better ?

I assume it's really strange ...

[Oldguy932]

That makes sense cause when mine are on the vapo they don't clock nearly as well as on air.

[jonspd]

tis kinda funny I think I had better results on my stock cooler

not really gonna test it as heating up the whole system may harm other parts....

has anyone tried the new ati chipset drivers msi has on there page for this board?

5 days old what's wrong with you people not keeping us up to day or did I miss the post

http://global.msi.com.tw/index.php?f...1&prod_no=1332

[Daveburt714]

Keep going yet.. a better BIOS will help at least a little.
Set 12x CPU multiplier, 1.3625VID on NB, drop HT Ref. to 205 and choose 12x NB multiplier... what do you get?

Can you drop NB multiplier to 8x with a BE?

Both of mine haven't, which was a BIOS bug.

I'll try that KTE and let ya know. The main reason I seem to get better results (at least in my head) with a 250 HT is it keeps my memory @ spec, maybe it's just old habit from the X2.

As far as the 8x NB, this chip WILL do it, at least AOD and CPU-z says that it does. IIRC, it will go down to 7x as well. I can post some screenies if you want.... Were you trying it on 1.13 Bios? Thats what I'm still running..

has anyone tried the new ati chipset drivers msi has on there page for this board?

5 days old what's wrong with you people not keeping us up to day or did I miss the post

I DL'd them Jon, when they went to load it looked like it was just the Cat 8.2 drivers though so I cancelled the install. If you install them let me know if I missed something, and if it helped...

Is it just me, or does it seem like the NIC drivers have some problems?
If I really want to test stability, seems like opening 4-5 tabs in IE and flipping back and forth or opening attachments/pics will cause a crash if the system is the least bit unstable.

[jonspd]

aggreed on the nic and interwed problem.

was gonna wait to see if KTE would tell what was in the driver pack as I dont like being the guinea pig. I dont think there just the cat drivers tho as they are for the ati chipset. which IMO would be the system drivers not the gpu drivers tho I could be wrong.

[Daveburt714]

Well KTE... That worked man!
I already knew it would OC pretty well as long as I kept it at the same multi as the CPU, don't know why I gave up on that idea so quick...
Thanks for reminding me to play with that

I tried to run a 1M superpi though, and it kept erroring, mabey 1.375 NB?
Or.... Maybe 250x10x10x8, that would be pretty sweet if I could get that to run.

BTW, Looks like your ORB links don't work unless your machine is running. Gives the message "ERROR: Tell your friend to turn their machine on"

[jonspd]

lol @ error IMO 1.375 is kinda high unless your going for 2.6ghz on the nb. I'm at 2200 with 1.288

[KTE]

You guys on Vista will get lower HT clocks than on XP. Vista induces a HT wall itself, well known since Kyosen's tests with Barcelona (I've tried it, I know Lightman has aswell). 234-250 is it's best limit.

Hello all, i'm actually testing any 9600BE, and i'm experiencing a strange Phenomen

My 9600BE overclocks better when it's hot.
With my Thermaright 90 ultra with fan 13x200 is a no go, vista won't boot (cpu temp is around 40° idle).
When i unplug the fan the temp is raising to 70° idle so i can boot @ 2.6ghz
The best frequency i can boot is around 100° idle, and vista boot @ 2.7Ghz.

Nothing against what you observe but physically that doesn't make sense. None of mine even booted 55C. Maybe your sensors are fudged and what it shows you as 70C is actually 20C.

100C real boot => no way. Dead chip over 80C real ran, TCC itself shuts down the chip auto at 70C -ish and it has a separate monitoring diode than the core readouts (hottest part). This looks like a case of fudged sensors, so far.

Tell me what your ambient temps are idling?

Anyone can try to increase the temps and see if the overclock is better ?

I'll try it tonight but there is a very large chance of having a dead chip this way. Basic electrical engineering (physics actually) decides that the chip will have a heat limit beyond which it will malfunction and become damaged. AMD has certified this as 70C, so it will be around that region for all these chips no doubt.

has anyone tried the new ati chipset drivers msi has on there page for this board?

Cat 8.2 I linked earlier

The chipset drivers are in every Catalyst release. Thanks for the link though.

I'll try that KTE and let ya know. The main reason I seem to get better results (at least in my head) with a 250 HT is it keeps my memory @ spec, maybe it's just old habit from the X2.

You can try both.
One setting you can keep for running the 1:2 divider for now and another when the new BIOS with some of these FID/DID options returns, so you can then run 1:2.66 divider with that setting then.

As far as the 8x NB, this chip WILL do it, at least AOD and CPU-z says that it does. IIRC, it will go down to 7x as well. I can post some screenies if you want.... Were you trying it on 1.13 Bios?

Yeah I tried it on both BE's with that BIOS, no POST (stock everything). I had to do a deep clear CMOS routine or no boot each time. Tried it many times over the last 2 months. If yours does POST then obviously it has it working. I'm waiting for a decent BIOS because that 1.13 BIOS still had many bugs which can make your OC confusing and limited (inconsistent), so a new one where all of you can drop from 9x NB multi will show me whether it's my chip which was hardware limited or the BIOS.

That's why I've still not tested the HT ref. singled out. My NB is always 9x or above.

Is it just me, or does it seem like the NIC drivers have some problems?
If I really want to test stability, seems like opening 4-5 tabs in IE and flipping back and forth or opening attachments/pics will cause a crash if the system is the least bit unstable.

I usually run Winbench 2000 Quality test to find the tiny instabilities after real stability testing is passed, and if that passes I first try the Firefox routine and then move on to downloading a video and repeat playing it back using VLC player and going away (6-20 hours). When I come back, that shows me whether the system is undergoing the "idle instability" disorder very well.

My current setup (2640) doesn't have this problems for well over 35 hours testing now. Looks like it's time to move up a notch.

I haven't experienced any NIC problems BTW, in fact this system gives me the best network connection and ping so far hence why I like to use it over the others. My Q6600 system gives me a very poor net connection for a driver related reason pretty obviously and doesn't let me connect using MSN/Trillian (both brand new OS installs); continuous source of frustration to me.

was gonna wait to see if KTE would tell what was in the driver pack as I dont like being the guinea pig. I dont think there just the cat drivers tho as they are for the ati chipset. which IMO would be the system drivers not the gpu drivers tho I could be wrong.

The servers were too slow, most were downloading at 3.5KB/s for a 195MB file. Server 5 is usually the quickest but that was slow till now (345KB/s now). Package traditionally contains the ATi AVIVO drivers, WDM drivers and the SB600 chipset drivers including for RAID. All the following marked in red:

I tried to run a 1M superpi though, and it kept erroring, mabey 1.375 NB?

The BE according to my experience;
-Runs sub 2.4GHz Nb with low volts
-Runs 2.4GHz Nb easily but requires a tad higher volts
-Runs 2.5GHz Nb but with a bigger jump in volts
-Requires high->very high volts and good cooling for above 2.53GHz NB. Sometimes maximum voltage and very good cooling. Phase/DICE/LN2 gets high Nb speeds especially.

Of course, every chip differs but this was with two separate chips.

Super Pi errors are usually memory related at those (low) CPU speeds, unless you've not given it around 1.28V CPU (which it may need).

If your chip can do 250HT stable (any multi), then try these combinations:

250x10x10x9 (IMC will have to be plus 1.4VID)
↑
[and all HT ref. in between]
↓
240x10.5x10x9

Set 1.25VID CPU, 1.416V CPU (1.392V real), 1.35VID Nb and try 240x11x10x9 too. See if it boots and runs.

BTW, Looks like your ORB links don't work unless your machine is running. Gives the message "ERROR: Tell your friend to turn their machine on"

Yep, thanks for the feedback. Orb works through the On Demand software installed on your PC, so you have to have it running to be connected to online. It acts as a server, so when I link anything from my system using On Demand, you're actually downloading directly from my PC.

That's why when I switch the program off it reads computer is off. They should work now... for a while, before I delete both the links and the linked source.

lol @ error IMO 1.375 is kinda high unless your going for 2.6ghz on the nb. I'm at 2200 with 1.288

For over 2.45GHz Nb you need higher volts than those (on air). As you've seen before, mine needs 1.312V for 2400 Nb but after many hours that'll freeze and one notch above at 1.325V gets 2400 Nb fully stable.

[KTE]

Hmm.. ok guys just checked out my RAM.
MSI RD790 seems to set the wrong [various] drive strengths (higher signal strength>more stress>more power>more heat). My RAM is rated for lower [various] drive strengths than this board is setting -> this will overheat RAM and can possibly kill them too.

Booting 1066 mode, this is what it sets Vs. what it's supposed to set:

Vs.
EVEREST Vs. SPDTool

As you can see, all drive strengths at 1066 EPP profile are rated for 1.00x but the board is setting them 1.25x-1.50x. So this means to get them back to the rated, they would need to be dropped. This is much more stress on the RAM and outside of it's power spec although it does usually help much with oc'ing RAM and getting RAM stable at oc's.

To get into a little more detail, each integrated circuit cell designed using V[HSIC]HDL creates large nets which contain specific cells (driving transistor gates) driving many other gates within. Each are designed with a Drive Strength referring to the number of adjacent gates they can drive (drive power) and a 1.00x DS refers to being able to drive two gates, 2.00x is the ability to drive four gates and so on.

BTW, some updated tools, they work good:

EVEREST Ultimate 4.20 Build 1297: http://www.lavalys.com/beta/everestu...dfg3mn0cxz.zip
CPU-Z 1.44.1: http://www.cpuid.com/download/cpuz_144.zip
HWMonitor 1.07: http://www.cpuid.com/download/HWMonitor_107.zip

CPU-Z now detects voltages well and EVEREST is benchmarking Phenom a little higher.

[misteroadster]

You guys on Vista will get lower HT clocks than on XP. Vista induces a HT wall itself, well known since Kyosen's tests with Barcelona (I've tried it, I know Lightman has aswell). 234-250 is it's best limit.



Nothing against what you observe but physically that doesn't make sense. None of mine even booted 55C. Maybe your sensors are fudged and what it shows you as 70C is actually 20C.

100C real boot => no way. Dead chip over 80C real ran, TCC itself shuts down the chip auto at 70C -ish and it has a separate monitoring diode than the core readouts (hottest part). This looks like a case of fudged sensors, so far.

Tell me what your ambient temps are idling?
I'll try it tonight but there is a very large chance of having a dead chip this way. Basic electrical engineering (physics actually) decides that the chip will have a heat limit beyond which it will malfunction and become damaged. AMD has certified this as 70C, so it will be around that region for all these chips no doubt.

I can assure you the temps i'd said are accurate, at 90° the heatsink was really hot, i've even seen 120°c while i was priming, but after the motherboard shutdown. I assume it's really strange but it's for real. I know i'm crazy
P.S : The chip is still alive

[KTE]

I can assure you the temps i'd said are accurate, at 90° the heatsink was really hot, i've even seen 120°c while i was priming, but after the motherboard shutdown. I assume it's really strange but it's for real. I know i'm crazy
P.S : The chip is still alive

You're running the MSI K9A2 Platinum?
Which BIOS are you running?

As I asked earlier, can you post your idle temp at stock speeds/volts with yuor normal fan on that heatsink?

First open up AOD Voltage/Temp monitor section, along with HWMonitor I linked above, with CoreTemp and then an EVEREST Sensor section and capture the screenshot with them all on. Post that back please.

[KTE]

I just had a hang after many hours.

I'll tell you how it happened: I Memtest'ed and then Prime95 tested

2652MHz 1.25VID 1.416V
1836MHz NB/HT stock volts
544MHz RAM @ tRC 11 and 75ns.

Previously I had ran tRC 14 and tRFC 105ns for over 24 hours at that setup. They passed so I dropped the tRFC from 105ns for both DIMMs to 75ns and tRC from 14 to 11. Inside 55minutes it hung whilst just changing tabs in Fx. When I took a cursory look at the power, it had shot up to 201W AC, whilst idling it was 114W AC. Thus it means the cores were actually loaded.

You would categorize this as a hard-lock, a CPU hang rather than freezing.

This is not making any sense to me, this just seems like a core bug to me rather than the core instability itself. I've just never seen this happen in all oc'ing and testing since 1995 and this isn't how a CPU functions unless it's bugged above a MHz range. There are quite a few errata's documented in the revision guide which can cause this behavior though.

I mean, how would AMD check for instability? They can run all the tools we have and still won't catch it unless they tested each chip for at least 7 days, including idling for over 2 days. I'm not sure what do check here.

I've now booted up again with higher CPU voltages and higher tRC and 195ns tRFC. It now hung inside 20 minutes.
My VCore was fluctuating from 1.432V all the way to 1.468V, although I set 1.416V BIOS (which usually gives 1.408V). My +12V rails are stable put, I really wish they had used R80 inductors on these boards.

Now, booted 1.392/1.384V at the above MHz... so far 63 minutes and going without a problem.

I also tried dropping NB FID on BIOS P0H using WPCREDIT now, as it worked on any Phenom which could drop it (even though so did 1.13).. well from 9x to 8x resulted in no POST, same as I found with any BIOS out there. Which means my Phenom BE is locked on NB multi to 9x as the lowest.
I'm curios now to see if the same applies to CPU FID too.

BTW, for any of you using CrystalCPUID, the following MSR let's you change CPU VID using it (and hence CPU voltage)

MSR Number: 0xC0010070

For example,

0x0 0x3000400A = 1.150VID
0x0 0x3000300A = 1.250VID
0x0 0x3000200A = 1.350VID

And so on.

[Oldguy932]

So your saying that I'd get higher clocks if I managed to find a copy of winxp and installed it? Time to go look for my good ole msdn dvd's...

[misteroadster]

You're running the MSI K9A2 Platinum?
Which BIOS are you running?

As I asked earlier, can you post your idle temp at stock speeds/volts with yuor normal fan on that heatsink?

First open up AOD Voltage/Temp monitor section, along with HWMonitor I linked above, with CoreTemp and then an EVEREST Sensor section and capture the screenshot with them all on. Post that back please.

No i'm running an Asus Crosshair.
At 2400Mhz and stock voltage with a TR Ultra 90, temp is 40° idle about

[aGeoM]

And for everyone try this software 144M test to see if anything is initially stable at any MHz. It works to catch errors excellent on Intel CPUs but seems to have problems with AMD CPUs. Please monitor your temps, volts and power draw: http://www.superprime.org/

Someone post a screenshot back of 144M completed with Phenom because I'd like to see if it runs on AMD CPUs. It's a cracking piece of software if it does run, better than Linpack 32b IMO.

My Phenom isn't booting yet, destroyed HDD has thousands of errors now. Using Q6600.

I think is not working, times are not constant..

[jonspd]

KTE so are the drivers any better?

[KTE]

No i'm running an Asus Crosshair.
At 2400Mhz and stock voltage with a TR Ultra 90, temp is 40° idle about

Ah, well your board was known to give inaccurate speeds/readings many times.

40C idle seems about OK to me though.... still further investigation gets this:

Quote from the "Revision Guide for AMD NPT Family 0Fh Processors", Revision 3.19:

128 Inaccurate Temperature Measurement with AMD SI

Description
The AMD SI thermal measurement feature does not meet ±5°C accuracy.

Potential Effect on System
Inaccurate temperature measurements reported through AMD SI Read Processor Control Temperature (Tcontrol) or Thermtrip Status register [CurTmp] (Dev 3xE4h[23:14]).

Suggested Workaround
Systems should be designed with the additional ability to use the analog thermal diode for temperature measurements.

Fix Planned
Yes

Any temps you see going that high above max rated without killing the chip would have to fall into that category... most likely.

I think is not working, times are not constant..

Thank you! It wasn't working on mine and I had to confirm it with another AMD user. It works on Intel systems perfectly fine though.

KTE so are the drivers any better?

Been busy stability tested my chip so I don't know yet. Gaming with something which uses 35%/100% on 4 cores shows it no better so far.

For synthetic benches, my score was actually a tiny bit lower with these than with Cat 8.1. I'll post later on with more bench details.

So far over a day now and 2652MHz is stable (no freezes, idled and ran a video for hours). I need this to last for over 40 hours and then I'll go up a notch.

BTW, as I stated earlier, if I put voltage above 1.42V on any MHz, it will fail and freeze soon after bootup.