Phenom 9500 w/ MSI K9A2 Platinum

[KTE]

I'll give it a go when I get home this morning. I think I tried 1&2 before, no difference really.

That's mainly for dual channel and not single channel. If you still get single channel, then return the board. However, CPU-Z is the only software I've seen which will detect it as dual channel properly.

When I run 1 3870's it detects erm 3278 I think? , then 3072 or so is when I run two 3870s. :s I really hope this isn't affecting my 3d scores. I have yet to run 3dmark yet with my 2 3870s, I think I may do that this morning too.

3D is affected my single vs dual channel but not for 3GB vs 4GB total.

Do you know if I should have this Memory hole thing enabled? it is by default.

Disable it, it shouldn't make much diffrence, it doesn't with mine.

I've considered modding the Timings manually, but I have no idea where to begin. The ram I've got is 5-5-5-15 pc2 8500 I believe.

GX24GB8500ESPDC

I'd need to see the SPD tab info before I can say anything as I mentioned. But you can go into DRAM Config and change the timings to what your RAM is rated for if it boots up differently. Follow the same output as what CPU-Z reads (look at the titles there). I'm really too short of time to go into details.

*edit* umm are the PSU readouts at all reliable with this board? My 12 is around 12.5V which is kind of concerning. I think the 5v was a bit off as well. I've got a Corsair 1000W psu.

Nope. The voltages should be checked with a DMM. There's no other way to know what is correct. Mine read +0.5V for the 12V rail while the DMM read +12.01V idle/load the last I tested.

[mongoled]

I'll give it a go when I get home this morning. I think I tried 1&2 before, no difference really. When I run 1 3870's it detects erm 3278 I think? , then 3072 or so is when I run two 3870s. :s I really hope this isn't affecting my 3d scores. I have yet to run 3dmark yet with my 2 3870s, I think I may do that this morning too.

Do you know if I should have this Memory hole thing enabled? it is by default. I've considered modding the Timings manually, but I have no idea where to begin. The ram I've got is 5-5-5-15 pc2 8500 I believe.

GX24GB8500ESPDC

*edit* umm are the PSU readouts at all reliable with this board? My 12 is around 12.5V which is kind of concerning. I think the 5v was a bit off as well. I've got a Corsair 1000W psu.

Im assuming you missed this otherwise you have hurt my frail ego

So erm, ya
reading thru all this. I had a dumb question. I'm running 1066 @ 800mhz (geil 2x2gb). The OS here (vista64) is only report 3GB, appears to go down when I add my second 3870. :s

Anyone know whats up with that?

Dont know if this will be of help but I think i read somewhere that there is a quirk with the memory allocation when a second graphics card is added. Are both the cards 512mb? Whats the amount of memory Vista is showing in mb? Is it 3072mb? I try to see if i can drag some information up

Will do my best to find time in helping u find this info.

some info.........here

[KTE]

Actually, thanks for the reminder. I totally forgot about that post mongoled.

So it's k, your still well safe from your sig.

[JumpingJack]

Again, another nice experiment and interesting data... thanks KTE.

[mongoled]

Actually, thanks for the reminder. I totally forgot about that post mongoled.

So it's k, your still well safe from your sig.

Just read up to page 3 of the thread i linked to, great info regarding the issue of mem allocation above 2gb. Will be seeing more and more people posting these types of issues here and ofcourse the blame game will start and we all know what will happen then


So with your permission allow me to post this once here, the synopsis made by the poster morphine @ TR on behalf of the people who brought the techy info together

Summing it up:

- You can't get much more than 3GB out of 32-bit Windows XP, period. Addressable amount will vary (3 / 3.3 / 3.5GB) depending on the devices you have, mainly your graphics card.

Why? above 3GB is necessary for device mapping, for example, drivers accessing your graphics card RAM as if it were main memory.

- PAE does *not* help in the former case.

Why? Enabling PAE is only good for enabling the NX/XD (DEP) functionality when using modern CPUs, because even if it's enabled, 32-bit Windows will not map stuff above the 4GB boundary, thus falling in the previous situation.

- The "memory remap" option in the BIOS does *not* help you in the former situation. The 32-bit addressable space problem remains with 32-bit WinXP. It helps in other situation (see below).

How can I use more than 3GB of RAM then?

- Your motherboard must support more than 3GB (well duh)
- Your motherboadd must support memory remapping (most modern boards do, look for "Memory remap above 4G" or something similar).

Why? The 3-4GB area is for devices. This "splits" the RAM in two pieces below 3GB and above 4 or so GB, respectively. Don't worry, the PC knows what it's doing.

- Your operating system must be 64-bit OR 32-bit Windows Server, or Linux with proper PAE support.

What? In more realistic/practical terms, think WinXP 64, Vista 64, some 64-bit variation of any Linux distro, etc.

BIG NOTE: If you have a 64-bit operating system but your program is 32-bit only, it will still be only able to address 4GB. That means that, i.e., even if you're running XP64 with 8GB of RAM, Supreme Commander will only be able to use, at maximum, 4 of those 8 gigabytes.

-- This "guide" created with information from, among others: bitvector (mainly) / UberGerbil / derFunkenstein / just brew it! / Convert

Personally this quote should be stickied somewhere so hoping a mod could do a good thing and rip it and pin it up somewhere where it can easily be seen and found.

Sorry to derail the thread KTE, will remove on your request

[KTE]

mongoled: It's fine, as long as the intention is to help.
Just a note though: blargman has Vista 64-bit and 2x 2GB installed yet still OS only sees 3GB.

MSI do not yet list a RAM QVL list and by the looks of how things are panning out with his testing, it seems they don't yet supoport his RAM.

Jack: You're welcome. Thanks.

[mongoled]

mongoled: It's fine, as long as the intention is to help.
Just a note though: blargman has Vista 64-bit and 2x 2GB installed yet still OS only sees 3GB.

re: Your motherboadd must support memory remapping (most modern boards do, look for "Memory remap above 4G" or something similar

Do you know if I should have this Memory hole thing enabled?

I believe this may have something to do with 'memory remapping' as mentioned ^.

Im just going on third party info here, dont have the hardware/software to test unfortunately.

The important thing to note is the following

Having a 64-bit capable OS is ONLY half of the solution

Lets try not to get stuck on this point

[blargman]

So obviously its some sort of memory controller/bios issue that's causing this I guess. The physical addresses are not being "hoisted" properly and remapped. Can I bomb MSI? :s

On an interesting note. I don't know if these means anything, but i remember i looked at the memory tab in CPUZ. and everything was blank but the amount of memory.

[KTE]

re: Your motherboadd must support memory remapping (most modern boards do, look for "Memory remap above 4G" or something similar

I believe this may have something to do with 'memory remapping' as mentioned ^.

Usually having a 64-bit OS is enough for 4GB to show up. I have ran this many times to be aware and I know how Windows works well enough.

But if you have software memory hole feature in the BIOS OR memory hole remapping, it can offer different options.
Hardware/software/disabled for memory hole remapping.
In this case the option software will show you the most amount of RAM.
Or in our case, disabling the software memory hole should show up the most amount of RAM.

In either case it is not working for blargman and yet it should be if the MB/BIOS supported it. The procedure for having 4GB showing up is very simple but it relies on the RAM being compatible with the board in the first place. So far MSI have not certify this RAM to work on this board.

[blargman]

Anything newer than 113? :s I'll be a beta tester

[KTE]

Not that I know of. 113 seems to be the latest and each version had some comaptibility with memory fixed. They were rumored to release a BIOS last week but they didn't and I desperately need one because I'm now getting hard lockups at stock out of the blue and very strange system behavior at random times. AOD is a total no start now.

[blargman]

screen of my setup.
http://208.76.110.197:8080/screens/untitled.jpg

unganged + slots 1 & 2 got me dual channel. thanks alot! memory size issue persists tho with mem map enabled/disabled :s

this board is just confused when it comes to my ram evidently perhaps its really ddr3

[KTE]

Yeah, the BIOS/MB doesn't support your RAM rated speeds/timings. It only picks up your RAM for 800MHz 5-5-5-15-22 2T 1.8V (SPD) and no EPP info like on Gigabyte boards. I suppose you will have to enquire with MSI here: http://support.msi.com.tw/?language=eng

x64 is faster for Phenom, so try testing your CPU and memory perf.

[blargman]

what benchmark should i run? You've been a big help guys, I'll see what MSI has to say. Any thoughts on perhaps trying to set the Memory settings manually? I have no idea what to set them to
GX24GB8500ESPDC :s

I've noticed a marked improvement in speed in COD4 loading, perhaps its mental, but i think it's the dual channel mode ram.

[KTE]

what benchmark should i run? You've been a big help guys, I'll see what MSI has to say. Any thoughts on perhaps trying to set the Memory settings manually? I have no idea what to set them to
GX24GB8500ESPDC :s

I've noticed a marked improvement in speed in COD4 loading, perhaps its mental, but i think it's the dual channel mode ram.

Run something like what I ran to start off with: http://www.xtremesystems.org/forums/...0&postcount=71

RAM timings, I don't know what your RAM is rated for since your system was not reading any EPP information. You should be able to run 800MHz (stock HT ref.) 4-4-4-12-20 at 2.1V with ease though.
This post may help you with your BIOS options: http://www.xtremesystems.org/forums/...90&postcount=6
These timings can be modified in the Advanced DRAM Configuration menu under Cell Menu in the BIOS. Choose DRAM Timing Mode "Both" and then change the 1st, 2nd, 3rd, 5th values to 4-4-4-12 and the tRC to 20. Should bootup fine.

You're welcome.

[KTE]

My Old DMM had became faulty with the P4 somehow reading very weird values so I picked up another new one spare around. 2-3 models above what Shamino likes to use so it measures -40C-2000C temps too at accuracy of +/-1% from 0-40C. Obviously you have capacitance, AC/DC voltage, AC/DC current, resistance measurements and so on.

Now, pay attention to this; On todays testing the CPU VID was higher than before, it was at 1.25V. Naturally than means higher amps are being fed to each core than when I last measured it when it was 1.20V CPU VID.
So as expected, the temps, wattage and amps are also higher. Here's my DMM idle/load voltage readings of the CPU +12V line, idle/load VAC readings of system power draw and temperature readings using its probe compared to HWMonitor readings.

I'm sitting next to a very hot heater and right next to it is a window behind which this is the current outside temp:


Central heater right next to the system, surface temp:


Ambient temp of system:


DIMM temp:


Passive cooled chipset temp:


Full 3hr+ load P95 in-place large FFT stock system 1.25VID 1.240VCore lowest fan speed Zalman 9700 only one fan HWMonitor CPU temp:


Same as above with adjacent reading of temp probe placed on IHS:


Very cool indeed, just like I had stated many times.

Now voltages are always wrong or inaccurate with software. BIOS reads my +12V correctly but EVEREST reads my +12V as 8.18V under load. I knew its badly wrong anyway, but look at the above screenshot and these to compare.

Idle CPU +12V:


Idle system wattage (VAC):


Full load CPU +12V:


Max load system wattage (VAC):



Thanks for reading.

[DaMulta]

I have been watching this thread for a while now. This is as high as I have been able to get my 9500, and the funny thing is that it is a 100% stable.

This is with the 1.1 bios from MSI, now that had a beta OC bios but I would get a bad CMOS Cheaksom with it. It gives the HT, and the 1066Mhz memory divider with it.


This is the HD2900PRO flashed to an XT with a 1.35v bios.

Now I am going to have to RMA my board because the slot 1 PCI-E has never worked. My video card will just set there with a red light.

[KTE]

DaMulta: Your NB speed is too high you'll need to drop it down before you oc from the BIOS to really get to your limit. And use AMD OverDrive to OC aswell. Otherwise you're going great.

My max HT ref with Phenom is 265MHz:
http://valid.x86-secret.com/show_oc.php?id=279946
http://valid.x86-secret.com/show_oc.php?id=276606

Max valid HT and NB is 2160MHz: http://valid.x86-secret.com/show_oc.php?id=280790

Anything above that NB speed does not validate repeatedly.

Max valid/benchable (before 3 days) was 2849MHz: http://valid.x86-secret.com/show_oc.php?id=280406

The max CPU speed now is about 2.53GHz and max fully stable is 2.2GHz. I can run anything up to 2.53GHz quite easily but even after all load testing is passed.. suddenly whilst browsing the system will lockup, so I can't call it stable. No more is possible yet and I've tried at least 18 times. Very very strange behavior, a) max on everything is dropping b) CPU VID and so the wattage/heat has gone up c) day by day it requires more and more voltage for the same frequency d) vdroop has increased by the day e) instability has increased day by day to only 2.2GHz being stable f) lockups have increased dramatically g) oc limit has decreased by a long margin.
I don't know what to make of it. First time experiencing this, ever.

[DaMulta]

I have ran all day on this, and did AMDOD for 3 hours to see if it was stable.

Ran 3dmark06 all day at this speed.

Now I have tried my HT at all settings, and the FSB to ends ends to see if there was a hole that it would not boot on for me. 236 is the max, when I get into windows with amd over driver I move it by one number and LOCk.

What do you mean about NB speed?

http://valid.x86-secret.com/show_oc.php?id=283128

[KTE]

What do you mean about NB speed?

The Northbridge speed. Look at the AOD main page to see it. Put that below 1.8GHz before you start oc'ing through the BIOS. HT can only be same as or below it. It's one of the most crucial things to control when oc'ing and keep below 2GHz if possible on this board.

To test max HT you have to drop CPU multiplier and NB speed. Then you'll see what the max HT ref is.

Also what volts are you feeding it for that speed?
Tried max volts yet? (if yours is cooled well)
Whats your max load temps?

This is my max load power and temps at all voltages being thrown the highest possible at stock speeds:


So it can't get much worse at all. Stock volts is obviously very cool and quite low power drawing.

[DaMulta]

So the HT control in BIOS is the samething? I had it turned down all the way when I started turned up the R Clock. Then when I maxed that out I slowly started to turn up the HT control.

I havent totaly figured out how to turn down the mulity yet. My temp is around 35 load.

I think my volts are at 1.37(around that)

NB is at 1.225

[KTE]

To oc Phenom, you'll have to learn how to use AM2+ P-States through the BIOS because that's what you'll have to use in order to drop CPU/NB/HT multi's. It's been covered here very thoroughly, quite a lot on the topic has been discussed:
http://www.xtremesystems.org/forums/...d.php?t=167977

The NB speed is separate from the HT speed and the NB volts are not the volts the BIOS labels "NB Voltage" ironically. Don't even bother touching that voltage since it won't be needed. There's quite a lot to read before attempting here.

NB clock is always equal or higher than HT clock.

[justapost]

KTE u might find this page usefull, for power consumtion comparison, till you get a Q6600.

At stock a Q6600 seems to use ~22,5W (DC or VAC?) under load ~50W more. So in idle it seems to require ~5W less than a 9500.
Taking into account that intel chipset's require ~25W and amd's new chipset's require ~10W a 9500/9600 system should require ~10W less than an intel system.

I'm curious what you'll find out.

[KTE]

That's very good actually, thanks.

They use a very similar if not the same board as I had for my Intel builds, basically the same HDD and the same CPU cooler. If I was to measure at VRMs the wattage I gave will be lower for the CPU as some of that is lost from my quoted figures, and especially if I factor in my voltages, currents and power factor.

22.5W is DC wattage with power saving running. That's close to what Penryn 3GHz does idle too when power saving puts it to 2GHz, around 21.5W. It drops the multi/volts/speeds. With P4, Core 2 and Penryn, voltage/frequencies matter massively to temps and power as you will see by only increasing or decreasing by 0.1V from a nominal voltage. Even by increasing just the FSB you will see a measurable enough increase in temps/power. You can take a look >HERE< for a quick frequency/FSB/volt/VID effect to Penryn power demand.
You can observe a similar trend with Phenom, but not from 1.0V-1.24V. You will hardly get a power consumption difference even if you dropped the frequency by a full 1GHz. From 1GHz 1.55V to 2.7GHz 1.55V the difference in total system full load power draw is 296VAC vs 330VAC, which is minute.
So from 1GHz to 2.2GHz you will hardly see an idle power consumption difference if you keep the voltage/VID the same (load is obviously different) and even if you dropped the voltage by 0.2V from stock the difference is very small in comparison to what it is with AMD X2, Intel P4, Core 2 or Penryn architecture.

They also did not specify a few very important parameters to me such as idle/full load temps and CPU VID for each core. A higher VID will get higher power consumption with the same core and a lower one will get lower power consumption, such as what you see when you compare my (or your) Phenom at 2.2GHz/2.3GHz/2.4GHz/2.6GHz with reviews of the 9500/9600/9700/9900 or when they overclocked.

I will test the Q6600 hopefully myself soon to find out wattages accurately using the exact same hardware and conditions (apart from board/processor).

The overclocking peak is limited in Intel CPU's by a) chip quality (some are better and some worse for high FSB/high speeds) b) stock VID c) stock voltage d) stock multi e) motherboard f) cooling and thus, considering a) and e) are good, the cooler you go, the higher you can effectively reach (until the chip or cooling limit is reached).

The overclocking peak is limited in AMD CPU's by a) chip quality b) stock VID c) stock voltage d) stock multi. Cooling doesn't seem to be the limit at all because we can go far cooler, the chip stays very cool and yet you won't see a difference in MHz gained.

On a good chip and board;
The limit with Intel CPU's is heat/power/voltage (same as P4).
The limit with AMD CPU's is processor technology (same as A64).

What it effectively means is;
For Intel to speed down the fab size scale is nothing but pure enlarged benefits in every way in terms of higher transistor switching speeds, reliability, less leakage, lower volts, less heat, less TDP, higher frequencies, better performance. They need it and they know it and that's why they push down very fast when they get stuck at the max TDP with their CPU's. I've not seen the chips pushed yet to their peak, I mean absolute peak which someone like my tutor whose a government physicist can test which requires more effective cooling and conditions. Sub -200-250C and you will see their peak with the best chips chips. I believe their architectural peak is like that of P4 and Tejas, around 7-10GHz. And I DO believe Intel is trying to get that dream "Tejas" speed out still, but cannot do it, lower fab node is not improving their TDP enough to meet those demands, so they now need other ways around this (more cores/multiple threads per core/RAM inside CPU aka cache/tweaks) whilst still trying for higher speeds. One node down should get them 4GHz retail dual-core I reckon.

But for AMD, to attain higher frequencies since A64 days, you have to simply change the processing manufacturing technology and design first and foremost because that's where the limit lies, not in heat/voltage/temps. So running from 65nm to 32nm isn't going to have even a 50% benefit for AMD compared to that Intel will have, unless there are radical changes. It won't allow them much higher frequencies, certainly not the same as Penryn at 32nm can get or even what Core 2 at 65nm can potentially get with cold. I believe their architecture peak is like that of A64, around 4-5GHz and that's what they're trying to reach. I don't know what they're doing to circumvent this or even if they want to. With native quad, it throws things backwards from X2 so another two tweaked cores, one node down and you'll probably be able to get X2 speeds again.

[beerking]

Now this is what we want to see.

click