Phenom 9500 w/ MSI K9A2 Platinum

Quote:

---

Originally Posted by Swanie

Thanks KTE that good news. My DFI also had "issues" due to BIOS mostly .Seem though MSI is more up to date getting things fixed. Can't wait for BE's to hit the net. Here in South Africa we get things slowly so anyone that knows where to get please let us all know

---

Swanie
Try here
http://sybaritic.co.za/store/
Little expensive but they are quick, also here.
http://www.take2.co.za/electronics
And Dfi here
http://www.prophecy.co.za/

Quote:

---

Originally Posted by KTE

Achim,

Ah OK. The first reading on the meter is watts, the second is the VA and the third the power factor (P(W)/S(VA)=PF i.e., 176.5/185.4 = 0.952 or 95.2%). These meters are incapable of showing PSU efficiency since you need an ammeter or a clamp meter or a SMPS ATE to measure that using values of how much amps each rail is carrying from the PSU and what the voltage for each rail is (multiply the two for the DC wattage) compared to how much VAC is being drawn in. So it's the ratio of how much the PSU pulls and how much it then gives out which is the PSU efficiency. Power factor that is being reported above is very different.

Youre power factor radings are how they should be. APFC is getting 0.95 PF while PFC is getting 0.78 PF. That's how it should be realistically. To measure efficiency you have to look at the manufacturer testing reports, do it yourself if you have the equipment or knowledge (large corporations or industries have engineers with the relevant tools many times), or look towards people wo have reviewed similar items like online websites such as PCPerspective, Anandtech, Toms Hardware, Hardforum, Techgage, Jonnyguru.com and so on. They have the relevant equipment to be able to measure the PSU efficiency at different loads. Inductive or capacitive loads, the efficiency (and power factor) will be fluctuating and never constant, so at 100W DC your PSU might be 74% efficient while at 200W it might be 76% and at 300W it might be 78% (and so on).

---

Thanks mate, i already did some online research and also grabbed my old physics book.
I thought the volt meter tries to detect the efficiency. :rofl:
Also on all my tests i only noted the VA results. :slap:

Quote:

---

Originally Posted by KTE

The Zalman 460W ultra quiet PSU you have came out before the wave of 80%+ efficient started popping up and it didn't have the best efficiency in the non-80+ PSU's either, since the Seasonic PSU's typically held that spot and they topped out at around 81-74% efficiency throughout the loads. At low loads these larger PSU's typically have low efficiency, like 68-74% below 100W DC. Yours 'aint 80% efficient PSU throughout the loads, bare this in mind, it's 75-80% at >220VAC below 40C above 80W DC AFAIK and I don't remember what it's report found the max efficiency at. It would be near the 50% range (near 270-300W DC).

---

Found it,
http://www.pcper.com/article.php?aid...e=expert&pid=4
Will do further tesing with the HD2600XT, to ramp up loads in the 80,5%-81,3% area. 80,8 with +-0,33 (standard deviation) should give me good results.

Thank you for all the very details explanations you gave. And allways keep an eye on your boss. ;)

PS: I ran over your whole thread looking for tlb-errata perfomance comparisons.
Have you ran any benches with/without the patch that showed the impact beside latency and sandra?

Quote:

---

Originally Posted by Emperor

guys, need some info. the silly errata patch simply means at higher clock speeds, it doesnt freeze up and doesnt affect overclocking?

---

It doesn't mean higher clocks but it definitely means very bad perf. Like 1.1GHz ;)

Quote:

---

Originally Posted by justapost

Also on all my tests i only noted the VA results. :slap:

Found it,
http://www.pcper.com/article.php?aid...e=expert&pid=4
Will do further tesing with the HD2600XT, to ramp up loads in the 80,5%-81,3% area. 80,8 with +-0,33 (standard deviation) should give me good results.

---

Very nice. I'd like to know what you find with your testing now (idle/load) for CPU and GFX.

Quote:

---

Thank you for all the very details explanations you gave. And allways keep an eye on your boss. ;)

---

You're welcome. :)

Quote:

---

PS: I ran over your whole thread looking for tlb-errata perfomance comparisons.
Have you ran any benches with/without the patch that showed the impact beside latency and sandra?

---

I tried it for a few hours (?) maybe an hour at max. It was too bad to continue with, severly slow performance and so I never ran anything else. My system was running on some FAH in the background so I couldn't accept having that bad performance.
Right now since 2-3 days its running WCG non-stop for about a week, so I can't really do anything yet.

Quote:

---

Originally Posted by KTE

It doesn't mean higher clocks but it definitely means very bad perf. Like 1.1GHz ;)

---

so it's like saying i still can run at my max 2.XGhz but it feel like 1.xGhz and wouldnt lock up right ?

Your lockups could be anything. Instability also causes lockups.

You can only run whatever is your max stable like this with the patch as far as I or anyone experienced. It doesn't increase your MHz or oc stability.

You get very bad perf, worse than stock. You're talking 30-50% of where it should be. That's what I was getting, extremely bad perf.

It'll only help you in cases where you have lockups at stock. If you don't get lockups at stock, the bug is not affecting you. AMD doesn't cover oc problems since its outside warranty and it's the users problem.

Quote:

---

Originally Posted by KTE

Your lockups could be anything. Instability also causes lockups.

You can only run whatever is your max stable like this with the patch as far as I or anyone experienced. It doesn't increase your MHz or oc stability.

You get very bad perf, worse than stock. You're talking 30-50% of where it should be. That's what I was getting, extremely bad perf.

It'll only help you in cases where you have lockups at stock. If you don't get lockups at stock, the bug is not affecting you. AMD doesn't cover oc problems since its outside warranty and it's the users problem.

---

but errata is something that hits at 2.4Ghz and full load on 4cores isnt it ?

so if i apply this patch at 235x11 it would be worse than 220x11 without patch right ?



does MSI have an option to switch on and off the patch?

Why would AMD cover 2.4GHz errata when that's outside their 2.2GHz/2.3GHz warranty and never guranteed by them? ;)

The errata is present in the silicon. It will never manifest itself at normal loads or usage. When you run a certain application with very unusually high loads, you will experience it. If yuo run those sort of apps, then you need the patch -> or else, no.

I'm running WCG 24/7 since 2-3 days at 2.42GHz and 1.98GHz on NB & HT. No problems whatsoever, no errata. The lockups are because these retail chips simply seem to have a speed limit.

That mainly, or, less likely, the BIOS is very bad yet.

Quote:

---

Originally Posted by KTE

Why would AMD cover 2.4GHz errata when that's outside their 2.2GHz/2.3GHz warranty and never guranteed by them? ;)

The errata is present in the silicon. It will never manifest itself at normal loads or usage. When you run a certain application with very unusually high loads, you will experience it. If yuo run those sort of apps, then you need the patch -> or else, no.

I'm running WCG 24/7 since 2-3 days at 2.42GHz and 1.98GHz on NB & HT. No problems whatsoever, no errata. The lockups are because these retail chips simply seem to have a speed limit.

That mainly, or, less likely, the BIOS is very bad yet.

---

so my unstable 235x11 is due to factors other than the errata when simply running FF ?

Yep. :)

Either or a combination of board/BIOS/CPU/NB/RAM.

Pretty sure Sami mentioned this on this board before too.

I can replicate that behavior right now. All I need to do is oc it to 2.55GHz and that's exactly what'll happen. ;)

Well i need your help. Took the MSI for myself ,build it over push power button but ......nothing. I got the first screen about three times but now nothing. It shows a screen that it is busy scanning the drives and then after a while tells me i have a hardware problem.? Running a phenom and Mushkin 1066. Silverstone 1kw PSU. Did BIOS resets as told by MSI site but nothing.

Remove any extra pci cards and test again also try upping the dim voltage to 2 volts.

Quote:

---

Originally Posted by Swanie

Well i need your help. Took the MSI for myself ,build it over push power button but ......nothing. I got the first screen about three times but now nothing. It shows a screen that it is busy scanning the drives and then after a while tells me i have a hardware problem.? Running a phenom and Mushkin 1066. Silverstone 1kw PSU. Did BIOS resets as told by MSI site but nothing.

---

if u selected 1:2.33 ram dividers, it wont work.

at least for my motherboard...

well got it to boot, got into windows but now here something weird. Cannot run dual channel, thus cannot run in green and orange together ... also updated bios to 113 but now every time i start it says Cmos bad checksum and i must press either f or f2, i cannot save my settings.

Quote:

---

Originally Posted by Swanie

well got it to boot, got into windows but now here something weird. Cannot run dual channel, thus cannot run in green and orange together ... also updated bios to 113 but now every time i start it says Cmos bad checksum and i must press either f or f2, i cannot save my settings.

---

how did u do that ?

Used a USB stick with the HP booter , new bios, etc. Also after flashing bios how to i safely reboot, etc...clear cmos..how..

Quote:

---

Originally Posted by Swanie

well got it to boot, got into windows but now here something weird. Cannot run dual channel, thus cannot run in green and orange together ... also updated bios to 113 but now every time i start it says Cmos bad checksum and i must press either f or f2, i cannot save my settings.

---

Are you running in unganged mode?
What RAM divider did you choose? (put it to 1:2 and see if that works fine)
Bad checksum is if it has an error starting and you haven't set the correct date/time (do this). Check your HT ref. speed is stock and make sure your VDIMM is as high as it should be.

Quote:

---

Originally Posted by KTE

Nah don't bother with him. :lol:

Low-k allows more power savings than high-k.

---

Really?

Quote:
" Using high-k/metal gate IBM and its Alliance Partners have been able to successfully shrink the size of a chip by up to 50 percent as compared to the previous technology generation while improving a number of other performance specifications. For example, high-k metal gate chips save about 45 percent total power, an increasingly critical metric in all electronics applications. "

http://www.physorg.com/news116860211.html

I guess there are some differeing opinions on that, but before you 2 guys blast me, I read stuff like that and the new Intel Process, which as you know- is HK/MG.

If I was wrong, that's fine..
JAP- no need to talk to me like that either.

Well KTE thanks for the reply. I am getting frustated here. Over at MSI HQ they assited but said i must not use 113 but here they say it more stable but no matter. Flashed bios correctly, got rid of boot error ,can get into Vista ...all fine but tried to run Dual channel and now pc don't boot. I have not set anything. all default. I am sure my memory was not on 1:2 divider and was running in ungaged mode. What are the Vdimm supposed to be, running a Phenom 9500.

Quote:

---

Originally Posted by muzz

Really?

http://www.physorg.com/news116860211.html

I guess there are some differeing opinions on that, but before you 2 guys blast me, I read stuff like that and the new Intel Process, which as you know- is HK/MG.

If I was wrong, that's fine..

---

That's all true. I don't see where you or anyone is wrong. But this is what I was talking about:
http://www.physorg.com/news85247225.html
http://nepp.nasa.gov/index_nasa.cfm/934/
http://www.fabtech.org/content/view/2079/74/

This "ultra low-k" was too difficult for AMD/IBM to implement yet but was better than the hafnium based high-k we know for power savings. ;)

Quote:

---

JAP- no need to talk to me like that either.

---

JAP? :confused:

Quote:

---

Originally Posted by Swanie

What are the Vdimm supposed to be, running a Phenom 9500.

---

Which RAM do you have (part number would be good)? :)

Mushkin HP2-8500 : 996578 ( the 2 x1 gig kit)
Rated : 5-5-5-18 and 2.2-2.35V

Quote:

---

Originally Posted by KTE

Lower fab node means a smaller manufacturing process i.e., 65nm to 45nm, or 45nm to 32nm, etc.
High-k allows lower power saving, larger leakage but higher frequnecies.
Low-k allows large power savings, much less leakage but lower frequnecies.

---

Huh? Low-k and High-k are used in different parts of the chip. They are not exclusive, far from it, you want both of them. You want a process that uses both high and low k.

Low-k has been around for a while. You want to use a low-k material between the interconnects and between transistors to lower crosstalk and parasitic capacitance. Everybody uses some form of low-k these days, there are a number of off-the-shelf solutions available.

High-k and the associated metal gate are something entirely new. High-k is used for the gate dielectric in order to reduce gate leakage and increase gate capacitance. Everyone is working on it, but only Intel has it in production with its 45nm process.

Some useful reading:

http://en.wikipedia.org/wiki/Low-K
http://en.wikipedia.org/wiki/High-k_Dielectric
http://www.spectrum.ieee.org/oct07/5553/1

Quote:

---

Originally Posted by KTE

That's all true. I don't see where you or anyone is wrong. But this is what I was talking about:
http://www.physorg.com/news85247225.html
http://nepp.nasa.gov/index_nasa.cfm/934/
http://www.fabtech.org/content/view/2079/74/

This "ultra low-k" was too difficult for AMD/IBM to implement yet but was better than the hafnium based high-k we know for power savings. ;)


JAP? :confused:

---

I read that, and others like it regarding that collaboration... in other words it's useless for <65nm.

Everything I've read to date states HK/MG is the way to go for reduced leakage/power consumption, which is what I stated in my earlier post that was rebuffed(which is fine- we are on the net!! heheheh).

Isn't leakage what caused the P4 to run so hot?

JAP= justapost
If someone doesn't agree with ya, and feels ya made a mistake that's fine.

OK, sorry about the OT.

Well i set my divider to 1:2 -still no boot and also upped volts to 2 ,no luck either

Quote:

---

Originally Posted by Swanie

Mushkin HP2-8500 : 996578 ( the 2 x1 gig kit)
Rated : 5-5-5-18 and 2.2-2.35V

---

See if this helps: http://forums.mushkin.com/phpbb2/vie...f191f5754df3d4