• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Overclocking my 1100T - Dealing with my Heat Wall

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

JustCallMeDan

New Member
Joined
Feb 3, 2011
Hi All!

I figure it's only fair to post my results so far, with thanks to Dolk's great article on overclocking the Phenom II, and hopefully prod a few old hands into a little advice as to where I should go next to help squeeze the last juice out:

Here's my hardware setup:

AMD Phenom II X6 1100T BE (Thuban)
Asus Crosshair IV Formula
OCZ Reaper PC3-16000 4 x 2 GB CL8 Edition (Note: deliberately overspecced RAM for OC headroom, which I figured would be especially important given I have 4 sticks and so far everyone has said that OCing 4 sticks of RAM is troublesome at best)
Corsair H70 Push/Pull as Exhaust
OCZ ModXStream-Pro 700W 80Plus
Antec Sonata Series Case - So old I forget what model! But I DO love rubber screws/plugs silencing things.

Oh yeah, and an XFX 7800 GT... Can you tell which component I haven't upgraded yet? :D

This is, effectively, going to be used as a workstation. I test trading strategies with a multi-threading backtesting platform - thus a high speed X6 with loads of high-speed memory (each thread will require over 250MB) is my best bet to decrease time spent running simulations. Think of it like the whole purpose of the machine will be to run Prime95 for three days straight - stability is of PARAMOUNT importance. Sure, I'll get out the games when I'm not running simulations, but that's not what the kit is primarily for.

Putting this thing together sure was fun... The 10cm deep H70 radiator and fans could sit inside the case, but would be completely blocking my PSU air intake (this one faces down into the case), so I had to take a hacksaw to the case and remove the rear fan grill to allow me to thread the H70 CPU block through the hole and mount the radiator/fans to the outside.

I'm running Windows 7 Ultimate 64-bit for the first time - and am pleasantly surprised with it so far. Glad I skipped Vista to be perfectly honest.



Right, first things first - Asus BIOS 1102 on this board included an automatic overclocking utility which I used first to see where it took me - which was straight up to 3.7GHz (16.5 multiplier with 225 bus speed) and auto voltages (Core Voltage of 1.38V according to CPU-Z). Stable as can be, so from here I push on.

CPU-Z revealed my DRAM was running well under specifications by default, so timings got 'loosened' to manufacturer specification and the FSB/DRAM ratio was set to 1:4. Correct me if I'm wrong, but 450ish MHz at CL6 won't perform as well as 900MHz at CL8, right? I hear that AMDs like lower latencies over a few MHz clock increase, but this is a 100% increase in data rate here! And I'm still under manufacturer specifications (data rate of 1800MHz < 2000MHz). Like I say, correct me if I'm wrong!

Next, I raise the multiplier in the OS through AMD Overdrive with wprime 32M tests, as described by Dolk, up to 18.5 before instant fail on wprime. Didn't bother even trying to raise voltage there given the fact I'd raised from default 3.3GHz up to 4.1GHz (24% OC) and it failed near instantly. Instead, I concentrated on getting 18x (~4.0 GHz) stable. My overnight Prime95 crash was of the BSOD variety, indicating memory or NB voltage, if Dolk is to be believed... Which right now I do!

Reboot and set CPU multiplier to 18 in the BIOS, climbing CPU Voltage to ~1.39V, CPU-NB Voltage to 1.29V. Problems even booting. Why? BIOS had auto set DRAM voltage WAY too high, and memory errors (I'm guessing!) were surely from this - 1.9V is just unnecessary - these sticks are specified to run at 1.65V! Hard coded DRAM voltage at 1.65V, and I'm now an hour through a Prime95 torture test without issue. Had to find a utility to actually LOG temperatures - after all, if I get a BSOD on an overnight torture test, I'd kinda like to know what did it... So I finally found SpeedFan (run as Administrator to actually log, it just doesn't do it otherwise!)

So, my overnight torture test will consist of Prime95 running with SpeedFan logging hardware details.


After only an hour of torture testing, temperatures are high though - this flat gets warm (low ceilings, at least we save on heating bills...). And it's sunny right now, which is when I usually abandon the flat myself, let alone expect the PC to stay cool!!

Ambient mobo temperature seems steady under load at around 30 degrees. Unfortunately, the CPU temp hits 64 degrees - not suitable for 24/7 IMO. And that's with the side of the case off.


So, to summarise, I am now at, precisely:

CPU Speed: 4045.6 MHz
CPU Mult: 18
CPU FSB (whatever you want to call it... CPU-Z still calls it FSB): 224.8MHz
HTLink: 2022.8MHz
NB: 2247.6MHz
Core Voltage 1.44V-1.452V (Yeah, LLC is still on, voltages fluctuate)
DRAM Freq: 899MHz
FSB/DRAM: 1:4
DRAM Timings: 8-9-8-30 (as per manufacturer spec)
CPU Core Temp: Load Max 48C; Idle: 24C
CPU Temp: Load Max 64C; Idle: 35C
(I understand CPU Core Temps are calculated and the CPU Temp is a sensor reading - I read on a post here that someone contacted AMD to ask which was the 62C limit for the CPU - they say the sensor temp, so I definitely need to reduce this)


As if on cue, BSOD arrives :)

Anyway, that figures given the temperatures.


Right. So where to go from here? The way I see it, I need to lower temperatures, which means lowering the voltage, which means lowering the CPU OC. I have two available routes here:

1) Drop the multiplier (next step down is 17.5 x 225 MHz = 3.9 GHz) and lower voltages until stable and peak CPU Temp at load doesn't exceed 55C

2) Since my RAM is running under spec, maybe I should try 15.5 x 250 MHz = 3.8GHz at (1:4 seems to be the maximum available FSB/DRAM ratio in the BIOS, so this would get my data rate up to 2000MHz)



Q) Will raising the 'FSB' to 250MHZ lead to other issues? Remember, stability is of paramount importance :)

Q) Does anything else in my setup look wrong to you more experienced folk?



Anyway, I hope my experience helps others!
 
OP
JustCallMeDan

JustCallMeDan

New Member
Joined
Feb 3, 2011
Bah. Less than an hour into another Prime95 run and CPU Temp reached 59C on 'Auto' voltages:

CPU Speed: 3820.9 MHz
CPU Mult: 17
CPU FSB: 224.8MHz
HTLink: 2022.8MHz
NB: 2247.6MHz
Core Voltage 1.4V ish
DRAM Freq: 899MHz
FSB/DRAM: 1:4
DRAM Timings: 8-9-8-30 (as per manufacturer spec)
CPU Core Temp: Load Max 45C; Idle: 24C
CPU Temp: Load Max 59C; Idle: 35C

Also tried the FSB at 250MHz, lowering the CPU mult to 15.5 in the OS (AMD OverDrive) and the system just hanged. Regardless of the cause, the voltages need to go down, not up... So reducing again... 13 x 250MHz... (i.e. marginally underclocked CPU to test high FSB) system instantly halted. So, I expect 225 to be my max feasible FSB. At least this says good things about the mobo's bios overclocking (I expect the algorithm assumes the multiplier is locked and only plays with the FSB)

Now trying 16.5 x 225 (3.7 GHz) with 'Auto' voltages...
 
OP
JustCallMeDan

JustCallMeDan

New Member
Joined
Feb 3, 2011
Load temp got to 55C.... Gently replacing the side of the case, it peaks back at 59C nearly instantly. Clearly I need better case airflow. Ah well.

Well... stability first. Backing down to 16 x 225 MHz = 3600 MHz...
 

Marshmallow64

Member
Joined
Apr 27, 2009
Location
California
I think you need a new case. You have a great cpu cooler, but your temps are pretty high. If I remember correctly the sonata case is not too great for air flow.
 
OP
JustCallMeDan

JustCallMeDan

New Member
Joined
Feb 3, 2011
Yeah, it's a pretty ancient case... No front fan mount - just a 'mid case' intake fan drawing air across the HDDs...

Here you can see the rig as is (with and without flash washing the image out...):

IMG_0068.jpg
IMG_0069.jpg

As you can see, I had to cut away at the case to fit the H70 radiator and fans... I don't relish the prospect of fitting it in a new case, although that's definitely my limiting factor here.

Well, this OC seems stable at load at 56C - A little higher than desired, but I reckon it's well out of danger. If it goes overnight successfully (which I suspect it will at this relatively low OC), I'll post benchs, screenshots etc. For posterity, I'm certainly not in danger of breaking any records with this case...
 

Marshmallow64

Member
Joined
Apr 27, 2009
Location
California
You should be able to get higher overclocks with that setup. Like around 4.0 ghz. Well the radiator is on the outside so that is a good thing. I was picturing it on the inside. Good luck with the prime run tonight! =)
 

Khan

Member
Joined
Dec 22, 2006
might want to take the voltage off auto, I think it might be setting it too high.

you should be able to get at least 3.6 at the stock voltage which is around 1.34 I think for that chip.

everything I've seen of those is that they run really cool.

also, are you getting those temps off coretemp?
 
OP
JustCallMeDan

JustCallMeDan

New Member
Joined
Feb 3, 2011
I think it's mainly down to poor case air flow and being in such a warm flat! Especially from 1pm through 4pm when the sun's pretty much directly incident on the floor to ceiling windows... But definitely more aggressive airflow would help - that front fan is NOT at the front of the case - there's an HDD in between it and an elegant but poorly ventilated front facade. Also, it's only 1300rpm. It's just helping to direct the flow. I don't plan to run the PC with the side off - it's too close to furniture on that side anyway.

All in all, that's got to be some significant negative pressure with the case closed - I have effectively no intake (120mm 1300rpm fan about 10cm away from a poor intake vent), with two 120mm exhaust fans - one through the PSU (probably low rpm) and the radiator pull fan at around 2000rpm.

I've got a spare temperature sensor that I'm going to put in there before I torture test the current mild OC - should be able to get a good ambient case temperature reading out of it, and fairly indicative if I dangle it right about 10cm in front of the radiator pull fan. Hopefully SpeedFan can log it too...

As for voltage, it's sitting at 1.38V now. Does that seem high for a 9.5% overclock? I'm finding it hard to locate any information on the stock voltage for this chip, interestingly.

Yes, I was using coretemp to give those readings, the cpu temp came from speedfan, which agrees with my mobo reading (Asus PC Probe II, bundled with the board), and is always higher. Right now, while installing the latest Windows SDK and other stuff (i.e. not much for six cores!), the core temps are around 25C with a CPU temp of 34C. Naturally, I'm watching the higher of the two temperatures!
 

Khan

Member
Joined
Dec 22, 2006
I think it's mainly down to poor case air flow and being in such a warm flat! Especially from 1pm through 4pm when the sun's pretty much directly incident on the floor to ceiling windows... But definitely more aggressive airflow would help - that front fan is NOT at the front of the case - there's an HDD in between it and an elegant but poorly ventilated front facade. Also, it's only 1300rpm. It's just helping to direct the flow. I don't plan to run the PC with the side off - it's too close to furniture on that side anyway.

All in all, that's got to be some significant negative pressure with the case closed - I have effectively no intake (120mm 1300rpm fan about 10cm away from a poor intake vent), with two 120mm exhaust fans - one through the PSU (probably low rpm) and the radiator pull fan at around 2000rpm.

I've got a spare temperature sensor that I'm going to put in there before I torture test the current mild OC - should be able to get a good ambient case temperature reading out of it, and fairly indicative if I dangle it right about 10cm in front of the radiator pull fan. Hopefully SpeedFan can log it too...

As for voltage, it's sitting at 1.38V now. Does that seem high for a 9.5% overclock? I'm finding it hard to locate any information on the stock voltage for this chip, interestingly.

Yes, I was using coretemp to give those readings, the cpu temp came from speedfan, which agrees with my mobo reading (Asus PC Probe II, bundled with the board), and is always higher. Right now, while installing the latest Windows SDK and other stuff (i.e. not much for six cores!), the core temps are around 25C with a CPU temp of 34C. Naturally, I'm watching the higher of the two temperatures!


if its clocked at 3.6, I think stock is 1.30 or so, I'd give it 1.33 and see how it run at that speed. and as long as your core temp is below 55 its fine, the cpu temp can go a decent ammount higher, but its usually within 15 degrees of the core temp give or take a few.

in fact, try it at 1.3 and see if it runs stable at 3.6, might work.
 
OP
JustCallMeDan

JustCallMeDan

New Member
Joined
Feb 3, 2011
Stable

Overnight run was fine - 8 hrs of Prime95 without error. Didn't bother with the extra temperature sensor in the end though, after all it won't measure airflow, which I suspect to be the culprit. I already have a motherboard sensor giving me a near constant reading of 30C.

CPU Temp under load climbed to steady 55-57C according to the SpeedFan log.

Here are some shots:

SpeedFan-Readings.JPG
CPU-Z-CPU.JPG

CPU-Z seems to quote a higher voltage than SpeedFan above.

CPU-Z-Memory.JPG

Still not quite PC3-16000 DDR3-2000, but close. Could possibly try to tighten timings given the RAM underclock - these are spec.

PCProbe.JPG

Here I see my mid-case chassis fan is actually 1600 rpm rather than 1300 rpm as I thought. Guess I underestimated it. It's still not sufficient intake though.



So, later I'll try reducing my CPU voltage to 1.30, 1.33, 1.35 etc. and tightening RAM timings by a cycle each - for now, I'd better start USING this machine!
 

Khan

Member
Joined
Dec 22, 2006
looks pretty solid, definitely tighten up the ram a bit, I'd try 7-7-7-22 trc 35.

but if the temps are good at 1.4v, then yeah, you can definitely get it down more. at that voltage you could probably run the cpu at 4ghz though.

but every step foward is progress.
 

QuietIce

Disabled
Joined
May 7, 2006
Location
Anywhere but there
:welcome: to OCF!


On Phenom II processors, X2-X6, the cpuNB usually needs to be increased for stability once you start going over 3.6 GHz. This may be part of the problem in addition to your temps.

I'm finding it hard to locate any information on the stock voltage for this chip, interestingly.
AMD technical document #43375 lists the vCore and cpuNB voltage ranges for various processors. All X6's (they don't specifically list the 1100T) have a range of 1.250-1.475 vCore and 1.050-1.175 cpuNB VID.