Phenom 9500 w/ MSI K9A2 Platinum

[bgbdbstrd]

I just tested the official 1.4v on a 9500, but there are no nb / pcie multipliers, no disable tlb fix option, no autoxpress. when i ran winrar benchmark, it's showing the tlb fix is enabled, so i guess we'd have to disable it manually with this bios.

anyone else tried this bios on a phenom?

one other thing, on some chipsets, increasing the pci-e frequency helps in the cpu oc, what's the recommended pci-e freq for the 9500+k9a2?

[Brother Esau]

Well finally got me a Phenom 9600 BE ordered $189.00 "FREE" Shipping at New Egg with Promo code was too hard to pass up!

[bgbdbstrd]

Does MSI change an already uploaded BIOS? i requested for a BIOS that has the PCI-E multiplier and NB multiplier unlocked, and the CS advised me to get the 103 BIOS. When I checked the FTP site, I saw 2 BIOSES uploaded just today. the 140 BIOS i believe is the official one which madfaze posted last 22nd, but it was uploaded only on the 24th? anybody tried 13T?



A7376AMS.13T 1024 KB 3/24/2008 1:55:00 PM
A7376AMS.140 1024 KB 3/24/2008 3:28:00 PM

[batmang]

Nothing too great to report here. But I finally started messing with my Phenom setup. 3DMark06 shows 2.3GHz, but its really at 2.5GHz on all four cores. I'm not sure why 3DMark reports the wrong clock, anyone know? 2.5GHz seems to be my max stable. 2.6GHz is impossible, for me at least. Anyway, here are my results:

Screenshot of everything:
http://jmbat.com/media/phenomoc/2poi...rk06_10981.jpg

3DMark 06 result:


-OS is Vista 64-Bit
-Video drivers are Catalyst 8.3
-AMD OverDrive is version 2.0.14 (2.0.17 reports it cant find my an AMD CPU.)
-BTW, I'm using AMD OverDrive to overclock

I'll up the video card some more and I'm sure I'll hit 11k. Woot.

[batmang]

Another screenshot for you guys. Nothing crazy. Upped the 3870 speeds, broke 11K in 3DMark06. Still stable at 2.5GHz.

http://www.jmbat.com/media/phenomoc/...4_3DMark06.jpg

[Grian]

Vista 64 bit is the culprit for your bad Overclock, I couldn't get 2.5 stable on my phenom BE in vista, I went to xp pro 32 bit and now I've hit 2.9 stable

Nothing too great to report here. But I finally started messing with my Phenom setup. 3DMark06 shows 2.3GHz, but its really at 2.5GHz on all four cores. I'm not sure why 3DMark reports the wrong clock, anyone know? 2.5GHz seems to be my max stable. 2.6GHz is impossible, for me at least. Anyway, here are my results:

Screenshot of everything:
http://jmbat.com/media/phenomoc/2poi...rk06_10981.jpg

3DMark 06 result:


-OS is Vista 64-Bit
-Video drivers are Catalyst 8.3
-AMD OverDrive is version 2.0.14 (2.0.17 reports it cant find my an AMD CPU.)
-BTW, I'm using AMD OverDrive to overclock

I'll up the video card some more and I'm sure I'll hit 11k. Woot.

[Mathos]

Man it's been quiet around here, did KTE give up on phenom completely or just take a break?

[batmang]

Vista 64 bit is the culprit for your bad Overclock, I couldn't get 2.5 stable on my phenom BE in vista, I went to xp pro 32 bit and now I've hit 2.9 stable

Wow... Vista is that much of a overclock killer? LAMMMMME.

[Extelleron]

Wow... Vista is that much of a overclock killer? LAMMMMME.

It's not Vista, it's the 64-bit. Phenoms do not overclock at all in a 64-bit OS; they're really unstable.

[Kayin]

Speaking of 64 bit, what BIOS do I need to run with a 3600X2 and 4 gigs of Geil Esoteria on x64? Just put my order in for half my new RAM (yeah, 8 GB, but I render on this machine) and I NEED 64 bit, but I also want to run this board. If I need to run stock, I can do it, but at the same time, I'd like to get my overclock out of it...

This machine works just as hard as it plays, so I need to stuff it full. Kerkythea renders and giant Sketchup files, as well as Terragen worlds-it needs to just run, and not bog down and take 8 hours for a single render, or skip skip skip when moving a model-all functions of enough RAM, which for me necessitates Vista 64...

[KTE]

19th time on XS losing a lengthy post in 4 days!!

[Daveburt714]

19th time on XS losing a lengthy post in 4 days!!

I HATE when that happens!!

Using the GBT 780G board? Try lowering the HT Link speed to 1400Mhz...

[karbonkid]

I NEED 64 bit, but I also want to run this board. If I need to run stock, I can do it, but at the same time, I'd like to get my overclock out of it...

I'm in the same predicament - are ALL 780 boards unstable OC'd with Phenom on 64-bit? I, again, need 64-bit for rendering (Kerkythea, Blender, Yafray, Indigo), but I use Linux (BlueWhite64) as opposed to Windows.

P.S., nice to see another Kerkythea user here

[Kayin]

I don't have a Phenom yet, but I want to be able to drop one in. So far, in 32 bit Windows, it's stable, but RAM starved for my apps.

[madfaze]

hopefully there'll be an update on bios that we could monitor NB, SB, PWM temps...

[bgbdbstrd]

been talking with the msi support about the including a nb multiplier option in the bios, after going back and forth for almost a week, a "bios engineer" supposedly told the customer service rep that the HT multiplier is the same as the NB multiplier?!? dont know if they will ever include it in a future bios revision...

what's the best bios for k9a2 in terms of a stable oc for the phenom?

[madfaze]

been talking with the msi support about the including a nb multiplier option in the bios, after going back and forth for almost a week, a "bios engineer" supposedly told the customer service rep that the HT multiplier is the same as the NB multiplier?!? dont know if they will ever include it in a future bios revision...

what's the best bios for k9a2 in terms of a stable oc for the phenom?

read few pages back KTE mentioned POJ and 113...btw what batch you got on your phenom?what is your highest stable OC?

[KTE]

Sorry fella's, been extremely busy with many things too lengthy to mention here and whenever I did have some time to post or get on, either I logged in and was called away never to return, I lost my post with system/Fx crash, or the time was fully spent fighting 891 Virus/Worms/Trojans/Spyware/Adware/Rootkits! Since I move around daily and I'm given a different system to use, the system I was given around 4 days ago was buttload infected and it's took me till yesterday late to clean it out.

Only thing I've done since my last post here with PCs is test a new pair of Crucial Ballistix Tracer 2x1GB 1066. Managed 1.968v / 2.00v 1080 5-5-5-10 maximum stable [Test 1/3 - Test 2/3] and so far just finished testing 1080 5-5-5-5 2.065v / 2.096v perfectly stable, Memtest and in-Windows [Test 1/2] including 100% load 22min Kerkythea rendering.

So, hows MSI been treatin' you fellas?
Any feedback on these:
Any better BIOS for oc? [best before was 113]
Any better BIOS for stability? [best before was P0J]
Any better BIOS for hardware support? [best before was P0J]

MSI told me the new features are coming but today they told me there is stlll no BIOS for those options we asked for.

On a side note, this is just some extremely ignorant drivel. Really, this is poor even for a super moron.

Get ready for some length now...

Tried that linpack+specview thing:

linpack+specview 242W DC
prime95+3dmark06 241W DC
linpack 235W DC
prime95 200-210W DC

And finaly the winner
linpack+3dmark06 260W DC

Everything measured with Odin, P5K-E + qx6850, so it's a little OT.

Hmm.. you know, max I get with a 450FSB 3.6G 1.36v Q6600 G0 is 224W DC momentary and 208W DC constant using P95 SFTT. So that wattage is looking quite a lot.

Max possible I get by running Intel TAT + ATi Tool.

But I'm mainly interested in Phenom for now, so did you test it?

Have you worked out if you have a bad rail reading, which rail is to which connector and how much difference there is with volts in P-Tuner from DMM volts?

I spent quite long on these, had to take my system to 4W TDP to check out that the 8-pin CPU +12V was 12V1 and 12V2 but even at 8W DC and 15W DC, they read 0A each, which meant my system was pulling wattage from even the 12V3 (the ATX 24-pin).

hmmmm 3 new bios's have been posted to the MSI FTP site, all within the last couple days.

A7376AMS.P0G
A7376AMS.141
A7376ACI.103

Has anyone tried any of them yet?

I've tried P0G, it was released a while back to me but 103/141 I've not seen before. Anyone test 141?
Thanks for informing and linkin' them BTW

AMD7376ACI.103 includes a PCIe frequency setting aned TLB disable. It does not include northbridge multiplier settings. That's the only thing missing now... I have a rig on stock coolers running FSB 230 CPU 2530 Mem 920 4-5-5-15-2T Vcpu 1.376 Vnb 1.275 (seems to go to 1.25) Vht 1.225 (seems to go to 1.25) Vmem 2.2 4x 1gb OCZ Pc6400 FlexXLC CL3 PCIe 115 HTMult 9x = 2070MHz just passed 20 hours Prime95 w/ affinity & 900MB ea. BIOS CAI.103seems pretty good, just missing the northbridge multiplier...

Thanks for the feedback. Have you check if TLB Patch is disabled on all cores now?

agreed, P0G and 141 neather boot for me (BSOD still, new install)
running stock on ACI.103 right now..

64-bit, right?

Accoring to TLB-disable program (v.1.04) the TLB fix is set to disable in the bios, and the MSR's are still set, this is most likly SP1 in vista. though,

Yeah, SP1 overrides BIOS set TLB Patch status when windows loads.. something definitely in one of the driver files...

lost 200 k/sec on winrar from 133, WTF? why does it keep getting slower? LOL

edit : after a reboot, it's only lost 100 k/sec went from ~1200, to ~1100

So... P0E is 100KB/s slower than 113, 133 is 100KB/s slower than P0E and now 103 is 100KB/s slower than 133? Not good

just an update, and a happy one... this is autotune, fresh install, working on getting back up and running-and a warning to others, as well...

My issue was with the heatpipe-it hit PCI-E0, which made the sinks lift and resulted in crashing. Fixed, as you can see...

Thanks for the feedback details, glad you have it working.

official 1.4v is out

click me here
so far so good...

Thanks for the info and link, how is it, anything new or different? Tested its perf?

I just tested the official 1.4v on a 9500, but there are no nb / pcie multipliers, no disable tlb fix option, no autoxpress. when i ran winrar benchmark, it's showing the tlb fix is enabled, so i guess we'd have to disable it manually with this bios.

anyone else tried this bios on a phenom?

Thanks for the feedback... looks like another one to avoid then.

one other thing, on some chipsets, increasing the pci-e frequency helps in the cpu oc, what's the recommended pci-e freq for the 9500+k9a2?

Not sure here, it only helped me when I set high (past 237) on HT with 9500/9600, not on 96BE and I recall Sami used it when setting high HT. For me it was 108-110MHz maximum which stabilized with IDE drives. I'm not sure why it would help unless like earlier CPUs, K10h worked its base clock frequency (HT Ref) from the PCI frequency using divisors.

Well finally got me a Phenom 9600 BE ordered $189.00 "FREE" Shipping at New Egg with Promo code was too hard to pass up!

Eh, camon Bro, wheres 3G and 2.8G benches we're waiting for

Any testing yet=> max NB, max HT, max MHz, max MEM, max bench, max stable, max blow up... max fire

Does MSI change an already uploaded BIOS?

Not usually, nope.

i requested for a BIOS that has the PCI-E multiplier and NB multiplier unlocked, and the CS advised me to get the 103 BIOS. When I checked the FTP site, I saw 2 BIOSES uploaded just today. the 140 BIOS i believe is the official one which madfaze posted last 22nd, but it was uploaded only on the 24th? anybody tried 13T?

A7376AMS.13T 1024 KB 3/24/2008 1:55:00 PM
A7376AMS.140 1024 KB 3/24/2008 3:28:00 PM

13T is new, looks like a tester release. Can you upload and link that here?
Also, can anyone pass me the FTP for MSI please, I don't have my own system with me till at least a week and so don't have any site links saved here to check.

Did you test 13T BTW?

Another screenshot for you guys. Nothing crazy. Upped the 3870 speeds, broke 11K in 3DMark06. Still stable at 2.5GHz.

http://www.jmbat.com/media/phenomoc/...4_3DMark06.jpg

Good going, you can ignore FM detection, we don't look to it for any authentication since it's very unpredictable and buggy. AMD OverDrive/AMD Power Mon and CPU-Z are authoritative enough (incl. Memset).

Man it's been quiet around here, did KTE give up on phenom completely or just take a break?

Still around, just very busy honestly. Not even had time to pick up another Phenom or ask uncle if 4850e/B3 are in. Will try this weekend.

Vista 64 bit is the culprit for your bad Overclock, I couldn't get 2.5 stable on my phenom BE in vista, I went to xp pro 32 bit and now I've hit 2.9 stable

Wow... Vista is that much of a overclock killer? LAMMMMME.

It's not Vista, it's the 64-bit. Phenoms do not overclock at all in a 64-bit OS; they're really unstable.

According to my experience and that of Lightman (IIRC), it's just Vista 64b and not XP 64b that is extremely problematic. I mean, Achim had less oc/stability on Linux 64b than XP 32b but still a higher oc than others and I had 2.622G max stable on Ubuntu 64b while my max stable on XP 32b was 2.68G, so not much difference (and that was reached entirely by in-windows tweaks available on the XP platform).

Speaking of 64 bit, what BIOS do I need to run with a 3600X2 and 4 gigs of Geil Esoteria on x64? Just put my order in for half my new RAM (yeah, 8 GB, but I render on this machine) and I NEED 64 bit, but I also want to run this board. If I need to run stock, I can do it, but at the same time, I'd like to get my overclock out of it...

Have you tried P0J and P0G? How were they? P0J should work with them AFAIK.

I'm in the same predicament - are ALL 780 boards unstable OC'd with Phenom on 64-bit?

Not all, my 790FX, 770 and 780G ran Phenom B2 200x12 2.4G and 225x11 2.475G on XP 32b, Ubuntu 64b, Fedora Core 64b and Vista Ultimate 64b perfectly fine. I didn't push it more because of the poor BIOS I was on and because I had to sell 'n' deliver it off that day without killing it... you know.

I think its pot luck whether you have one that works upto around 2.8G on 64b or not, but there is a trend of having major problems on Vista 64b. There are guys who have 2.8G on Vista 64b running perfectly fine yet others who can't even get 2.4G running on it. I would love to have their CPUs shipped to test what and why.. but you know they will never do that

I use Kerkythea, Blender, 3DS Max 9, Apophysis, SPECViewperf 10.0, Maya 6.5, POVRay 3.7, Cinebench 10, Cinebench 9, Mandelbrot and Fract to test my system rendering performances, comparing between many speeds, settings and AMD vs,. Intel systems. One problem I had with Kerkythea is I couldn't make (didn't spend time learning it) a decent model to render so I was using the example image file WinOSi and rendering that with 4 threads at 1280x1024 with stock settings, which took around 2m30s on the Q6600 450x8 3600MHz / 1080 5-5-5-5 RAM. Didn't look much of a test though. Between Kayin and karbonkid (and anyone else capable) are you guys able to provide me a settings file to render which is reflective of your usage better for performance comparisons (heavier is better - 2GB limit though)?

Would be great if you could, even in Sketchup and Blender if you want, its appreciated.

===========EXTRA 45nm DISCUSSION=============

AMD had plenty of time with Intels 45nm process variation ramp problems, (causing long delays/poor availability) but AFAIK, Intel has sorted the problems out and is now going to release Xeon L5420 2.5G and Xeon L5410 2.33G low-voltage quad-cores, at only 50W TDP with a 7-year life cycle. You know whats best.. they have a 40W TDP 3G low-voltage dual-core coming out too.
Yup, depending on price, AMD will now lose out quite a lot in that market... massivley from their only "energy efficient" products. Their process leakage problems caused them to fall very behind.

TBH, naturally, without anything but HK/MG integration, over SiON/Poly-Si, you are bound to make collosal energy/leakage gains, so this is no biggy, but Intel also has high frequency and strong perf. at those wattages, that is a MPU biggy. IIRC (memory is weak here) Intel is using TiN gate with HfO2+MgO high-k dielectric for NMOS and HfO2+Al2O3 high-k dielectric for PMOS in their released 45nm chips. What that does is enhances electron mobilities and reduces charge-trapping (PBTI) also lowering threshold voltage. Using older Jan '08 DATA since I'm not exact of the updates:

Intel 65nm 2005:
NMOS Idsat 1210 μA/nm @ Ioff 100 nA/μm
PMOS Idsat 710 μA/nm @ Ioff 100 nA/μm

Intel 45nm 2007:
NMOS Idsat 1360 μA/nm @ Ioff 100 nA/μm
PMOS Idsat 1070 μA/nm @ Ioff 100 nA/μm

45nm+HK/MG Idsat Gain:
NMOS +12.4%
PMOS +50.7%

Intel very strangely skipped presenting any details or turning up at IEEE IEDM 2006, so they only allow us to compare 2005 vs 2007 data, we'll do now for IBM/AMD (ITSA). Not sure but those values look AC as typically given, but, either way, this data is even less critical than ever before for 45nm HK/MG - Actual CPU transistor Idsat depends entirely on your fabrication processing stage, many problems there, especially with plasma deposition techniques, chemical vapor desposition and the stages where substrate is subjected to high temperatures, which can ruin a transistors performance immensely resulting in lower actual perf. and sometimes, far lower (this is why the words "best on paper" were used at IEDM '07). Intel sticking with dry lithography meant they were bound to face much more problems in this department than those using immersion.

ITSA 65nm 2005:
NMOS Idsat 1259 μA/nm @ Ioff 200 nA/μm
PMOS Idsat 735 μA/nm @ Ioff 200 nA/μm

ITSA 45nm 2007:
NMOS Idsat 1364 μA/nm @ Ioff 200 nA/μm
PMOS Idsat N/A μA/nm @ Ioff N/A nA/μm

45nm Idsat Gain:
NMOS +8.34%
PMOS +??%

Now that shows nothing really. What can you expect from ITSA 45nm?
First of all, far lower wiring delay and power leakage due to Ultralow-k use, more than that of Intel 45nm, and secondly, IBM/AMD already presented their experimental findings of the highest PMOS (110) transistor performance with only using conventional SiON + compressive liner and eSiGe stressors with optimized Rext. at Lgate=35nm, Vdd=1.0V, 250 nm poly-pitch: Ion over 1 mA/μm at Ioff 100 nA/μm. AFAIK the concentration of pFET Germanium was <30% and SiC was used for nFET performance advancements. This was back in September 2007, so quite obviously with HK/MG, even with metal gate-first approach, they are not going to be much behind Intel 45nm on pure transistor perf., if not quite ahead judging by their research experimental data we have from them.

Overall, AMD expects to achieve a 20% performance improvement at the 45 nm generation from strained silicon and ultralow-k dielectrics, with high-k/metal gate providing the ability for further enhancements.

SOI was best for soft error immunity, low voltage characteristics and low junction capacitance but suffered parisitic effects, poly-gate scaling, and especially, FBE. Well, on top of the above developments by IBM/AMD, IBM in the paper “Record RF performance of 45-nm SOI CMOS Technology" showed 'peak transition frequencies (ƒT) of 485 GHz and 345 GHz for floating-body p-FET and n-FET devices, by employing a notched-body contact layout which significantly reduced parasitic capacitance and gate leakage current, ergo improving RF performance'.

Put the basics together and ITSA could have something very special, it certainly looks so for current data. I know IBM is going to produce a 45nm CPU too, can't wait to see what the heck that would be looking at Power6 65nm. However, while this does more or less guarantee boosted clock speeds for IBM as it did with Intel, not so for AMD; that's design and microarchitecture dependent. If you read 2001-2003 research papers, all the high clock speeds were even then possible just by shrinking process node, but the most significant paramount problem was power density of the MPU rising with FO4 delays coming to their lowest for stability, i.e it would go above 200W/cm², which is a total design failure. Since then they planned, the only way forward to continue scaling is by "decreasing delectric" and "focusing on multi-processor parallelism", and here we are today. Have a nice read

[justapost]

Hmm.. you know, max I get with a 450FSB 3.6G 1.36v Q6600 G0 is 224W DC momentary and 208W DC constant using P95 SFTT. So that wattage is looking quite a lot.

Max possible I get by running Intel TAT + ATi Tool.

But I'm mainly interested in Phenom for now, so did you test it?

Unfortunately the intel linpack binaries do not work on the phenom system, so i could not yet try it, must build my own binaries first and i'm very short in spare time atm.
I used an qx6850 GO ES did not play with the FSB, due to lack of time. With an 11x multi and 333MHz FSB wattage is ~230W DC during prime95 test.

Have you worked out if you have a bad rail reading, which rail is to which connector and how much difference there is with volts in P-Tuner from DMM volts?

I spent quite long on these, had to take my system to 4W TDP to check out that the 8-pin CPU +12V was 12V1 and 12V2 but even at 8W DC and 15W DC, they read 0A each, which meant my system was pulling wattage from even the 12V3 (the ATX 24-pin).

Still on the todo list. If i plug in the USB Stick 12V2 raises about 3-4W.

===========EXTRA 45nm DISCUSSION=============
AMD had plenty of time with Intels 45nm process variation ramp problems, (causing long delays/poor availability) but AFAIK, Intel has sorted the problems out and is now going to release Xeon L5420 2.5G and Xeon L5410 2.33G low-voltage quad-cores, at only 50W TDP with a 7-year life cycle. You know whats best.. they have a 40W TDP 3G low-voltage dual-core coming out too.
Yup, depending on price, AMD will now lose out quite a lot in that market... massivley from their only "energy efficient" products. Their process leakage problems caused them to fall very behind.

Saw that news. Those xeon's are for dual socket systems, and they still use fbdimm's, that might level things out abit. Does 7-year life cycle mean 7-years warranty?

[dr_drache]

@KTE yea 64bit, talked to MSI about drivers/bios and they told me to talk to promise, that's their problem. (didn't matter than i told them that some bios work, and others don't)

here is the 13T Bios : A7376AMS.13T.zip


havn't had a chance to test it myself yet.

[Mathos]

Yeah if you get a b3 phenom you won't even need to worry about the TLB fixes in the BIOS's. You might actually be able to test out the performance on the actual release bios's properly. I think as far as the release bios's go maybe 1.2 would be best, since that one still has the custom p-states. I'm just concerned that the SP1 vista thing may be a processor driver issue that may effect all phenoms regardless of stepping.

If I can get the money saved up I'd be half tempted to just save up again and buy a new 9850, just to get around the b2 stepping stuff.

and yeah, that MSI forum thing about the IHS I actually replied to and told em the other way around.

[bgbdbstrd]

read few pages back KTE mentioned POJ and 113...btw what batch you got on your phenom?what is your highest stable OC?

Thanks! finally got in here


I'm using P0J, I'm only getting around 2.5Ghz stable (230 Mhz HTT/FSB) with a minor vcore bump. I can get it stable at around 2.7Ghz (can do benches, like 3dmark06 and over an hour with prime95 running on all 4 cores, didnt error out, i just stopped it after an hour, not really that patient)but im not comfortable nor do i feel it necessary with the vcore increase ( a little less than 1.4v, dont have the guts to go higher) for a 24x7 rig


is there a way for me to find the batch without taking off the hsf? taking off the geminii and moving my case to mess with the innards (the old steel version) would be a full time job

[KTE]

Guys, the AMD Overdrive 2.0.17, Vista SP1 Patch and TLB Patch issues have all been fixed in the latest release, follow these instructions by Sami: http://www.xtremesystems.org/forums/...d.php?t=181197

Post#40 onwards

Now if you click the Turbo mode (yellow), it should disable the TLB Fix on all cores.

Also, I'm seeing many of you guys actually go after 2/4x2GB setups, current favorites being G.Skill F2-8500CL5D-4GBPK, G.Skill F2-8000CL5D-4GBPQ and OCZ OCZ2RPR10664GK. I can test these on this board with a Phenom/X2 if you guys want, see what works, how and try working around them if problematic?
Let me know.

I'll get back to you guys individually later, uber busy once again.

[dr_drache]

KTE i would LOVE if you could test G.Skill F2-8500CL5D-4GBPK VS.OCZ OCZ2RPR10664GK

i am still debating on which of these 2 kits is best for me...
i like the idea of the OCZ head pipes, but at the same time i can buy thermalrights kits if they make a differnce...

so, if you could, it'll be awsome....

[Kayin]

As above, and if you can GEiL Esoteria 6400 4-4-4-12 as well.