Looking for Suggestions on my E4300 Overclock on BadAxe2

[chuchu]

Hi Guys,

Great forum. I’m new to this and want some expert input. For my rig, see my sig. Objective is to build a stable workstation class machine which is quiet and which the stability of an Intel chip on an Intel MOBO, backed up with ECC ram and Raid 1.

I chose the E4300 because of its high OC potential and it’s low cost. Objective is to get a decent stable air overclock that can just keep on ticking 24/7. The best number is not my objective. I would love to get to the low 3000Mhz in CPU, but as you will see below, maybe it just was not meant to be with this processor.

After much reading on this forum, here are the steps I followed (and used gtj’s memory calculator extensively – thanks!) :

Step 1
I started with stock voltage all around on Ref Freq = 200. Got as high as 250 fsb, for a CPU Freq of 2250. Could not get to 260 fsb on stock voltage. Would not post.

Step 2
I reset the Ref Freq to 266. Got as high as 325 fsb, for a CPU Freq of 2925. At fsb of 330 to 335 I could not get Orthos success. At fsb 340 it would not post. By increasing CPU voltage to 1.3375, I could get Orthos to pass a short 7.5 minute test at 335 fsb or CPU speed of 3015, but obviously with higher temps, as you will see in the chart below.

Step 3
I reset the Ref Freq to 333. Was able to start successfully testing at a fsb of 330, but immediately at fsb of 340 I had to start applying more CPU volts to get the OC to work. At fsb of 350 I needed 1.4 volts. Got as high as 360 fsb with 1.475 volts but with absolutely insane temperatures. I also tinkered with other voltages, but that did not seem to work. Rather, applying more CPU volts exclusively was what made the higher overclocks work.

I have attached a chart below showing the limits of steps 1, 2 and 3, using the memory calculator (temp & volt columns added).


Conclusions & Questions
My conclusion is that for the sake of running cool, stable, and long term 24/7, I should run this rig according to my results in Step 2. Namely at a fsb of 325 on the 1066 strap with a CPU speed of 2925. I could get more CPU speed by applying more CPU volts on the 1066 strap but temps ramp very quickly with this processor. I would also get more CPU speed on the 1333 strap, but at the expense of more stress on the Northbridge, heat from the CPU, and volts applied to the CPU. Also, if you look not only at CPU speed, but at FSB Tput and DDR2 Tput, the gains on the 1333 strap are incremental for FSB Tput, and negative for DDR2 Tput, as you move from 1066 strap to 133 strap. As I write this, I am into the 8th hour of, so far successful, Orthos testing at the CPU speed of 2925.

My questions to those more knowledgeable are as follows:

1. Any comments on my process? Any tweaking suggestions?

2. The only other thing I can think of is lapping, but I am not sure I want to take apart the rig to do that. I read that lapping does not add much to a E4300. For example, I might end up lowering temps, but I am at the point of having to apply more volts, and doing that increases the CPU temps dramatically with this processor. So at the next +10fsb with more volts, the temps may just come back.

3. Do I gain or lose anything by going to the latest BIOS? Once the Watchdog is turned off, does it stay off with rebooting?

[newls1]

You are incorrect with saying "I would also get more CPU speed on the 1333 strap, but at the expense of more stress on the Northbridge." by switching to the 1333 strap, you RELAX the northbridge...

[gtj]

Welcome.

Looks spot on to me. Good reasoning and conclusion.

As for the BIOS, 2333's only flaw is the HDD light stays on if you disable the Marvell controller. If that's not an issue for you, just stay there. If you do upgrade, then yes, if you disable the watchdog it stays disabled.

Oh yeah, you have 2 typo's in your sig... You probably have 16GB and 290GB arrays not 16MB and 290MB. :)

[chuchu]

Thanks Guys,

Based on Newsl1's comment would it be your recommendation that I go from 325fsb on the 1066 strap to 330 on the 1333 strap?

Please take a look at the revised chart I am attaching showing the comparison between the temps, FSB Tput, and the DDR2 Tput. Note that I can run both of these at stock.

Which one is better for performance?

Which one is better for longevity?

Does the 1333 strap stress the Northbridge less?

How much does lower DDR2 Tput on the 1333 strap matte vs higher DDR2 Tput on the 1066 strap?

[Getttosmurf]

first of all why dont you change to a highend cooler and bump your vcore ? would result in higher clock and most likely lower temps , thats stock intel cooler right ? 2970 vs 3015 stay at 2970 the small increase in freq isnt worth those temps.

My 2c


edit: hmm read your sig , ThermalRight Ultra 120 w Noctua NF-12 at full speed , dont know much about that cooler but temps seem really high

[chuchu]

I am running a high end cooler, specifically the Thermalright Ultra 120.

[Getttosmurf]

just noticed that in your sig , dont know much about that cooler, never tried it but temps seem way to high.

Edit: try reversing your Antec 120mm TriCool and see what hapens

[chuchu]

I will change the fans around and see if that helps.

In the meantime, I did some very quick searches on this site for 1333 strap vs. 1066 strap, and on my first quick glance, it looks like the 1066 actually gives better performance. So unless someone with more experience can recommend otherwise, I think I will stick with 325fsb on the 1066 strap.

[gtj]

Stick to the 1066 strap. Unless you're going to be running your FSB significantly above 400, the 1333 strap is pointless.

More info...

http://www.xtremesystems.org/forums/showthread.php?p=2073058#post2073058

[mezcal]

Might also try 200:400. That would keep your ram slightly under spec (650) but use the best internal timings. Assuming you can get back to 325 of course.

[chuchu]

It's a wrap on this project. Changing fans makes no difference. The only temp difference is seen when I ramp up the TriCool case fan from Low to Medium to High. A degree or two for each notch.

Special thanks to gtj. Couldn't have done this would your memory calculator!

rk

[Getttosmurf]

if i where you i would investigate those high temps , look at the charts here http://www.anandtech.com/casecooling/showdoc.aspx?i=2943&p=4


It's a wrap on this project. Changing fans makes no difference. The only temp difference is seen when I ramp up the TriCool case fan from Low to Medium to High. A degree or two for each notch.

Special thanks to gtj. Couldn't have done this would your memory calculator!

rk

[chuchu]

I am open to suggestions. I am also familiar with that review since that is where I first learned of the this HeatSink.

My Arctic Silver 5 was a couple of years old, but I don't think that should matter.

So the possibilities are:

1. Maybe the E4300 is peculiar in its heat curve? Are there any other user experiences or data available as to the E4300? On average, where does it top out in speed and heat?

2. I could rip out the MOBO and re-seat the HeatSink again.

3. I could do number 2 in conjunction with lapping.

4. Maybe there is just a limit for this particular processor?

My only concern is devoting time without achieving results.


Your suggestions are welcome.

[mezcal]

So the possibilities are:

1. Maybe the E4300 is peculiar in its heat curve? Are there any other user experiences or data available as to the E4300? On average, where does it top out in speed and heat?

2. I could rip out the MOBO and re-seat the HeatSink again.

3. I could do number 2 in conjunction with lapping.

4. Maybe there is just a limit for this particular processor?

My only concern is devoting time without achieving results.


Your suggestions are welcome.

Haven't seen much in the way of compiled data for this chip. Bits and bytes scattered everywhere as usual (love the internet.) Can tell you a little about mine though. @333x9-boots ok, fails orthos within a few minutes. 312x9-boots ok, fails orthos in a few hours, 300x9-seems stable forever. Temps are 60ish (orthos blend, priority 9) in TAT. Ram always at 200:400 and all voltages stock.
I didn't see any mention of your temp monitoring program. Given your analytical approach to the overclock, I assume you have read the assorted threads here and elsewhere regarding how the various monitoring apps interact with the 4300. :fact:
Keep going, you're doing my work for me :toast:

[mezcal]

I'm running the oem intel cooler btw.

[chuchu]

I'm using Coretemp 0.95 and Intel Thermal Analysis Tool to check temps. The temps are disappointing, especially with this cooler. I was expecting better temp results.

Some have reported better results with the e4300, but many seem to be in our range. Maybe the only solution is the removal of the IHS, and here is a recent thread:
http://www.xtremesystems.org/forums/showthread.php?t=136776&highlight=E4300

Before we go that far, it would be comforting to know more than one person has succeeded with this. Has anyone else out there managed to pull off removal of the IHS from the e4300? If so, what batch was you processor from?

[gtj]

Before you think about removing the IHS, consider this...

With the IHS removed, the die is actually even with or LOWER than the retention mechanism. If your HSF contact area doesn't fit INSIDE the retention mechanism it won't make contact with the die.

I would say that your best bet is to re-seat the HSF. When you remove it, check that you had even contact between the HSF and the IHS.

[zodden]

That load temp with such a good cooler is what bothers me. Before you start lapping I would first just see if you can start over. Re apply a new coat of AS5 and then re-seat.

[mezcal]

Objective is to build a stable workstation class machine which is quiet and which the stability of an Intel chip on an Intel MOBO, backed up with ECC ram and Raid 1.

I chose the E4300 because of its high OC potential and it’s low cost. Objective is to get a decent stable air overclock that can just keep on ticking 24/7. The best number is not my objective. I would love to get to the low 3000Mhz in CPU, but as you will see below, maybe it just was not meant to be with this processor.


Please reread your original post-and stick with the original goal! I think it's the right one.


I'm using Coretemp 0.95 and Intel Thermal Analysis Tool to check temps. The temps are disappointing, especially with this cooler. I was expecting better temp results.

Some have reported better results with the e4300, but many seem to be in our range. Maybe the only solution is the removal of the IHS, and here is a recent thread:
http://www.xtremesystems.org/forums/showthread.php?t=136776&highlight=E4300

Before we go that far, it would be comforting to know more than one person has succeeded with this. Has anyone else out there managed to pull off removal of the IHS from the e4300? If so, what batch was you processor from?

I agree your temps are not great given your setup but I also have to say that gtj is right. Physically modifying your chip, while it may be neat and is certainly compatible with XS, is not moving you down the road to your stated desired results-24/7 stability and not chasing that last 4 hz. It may help lower your temps of course but...
I wouldn't even consider it myself. Bear in mind that you are already well over a 50% increase from stock. I certainly understand that the (oc) bug bites but I would seriously reconsider hacking up your chip. If you do, however, I'd love to hear how it comes out! :clap:

I've got to build 2 more ps harnesses for mine tomorrow (atx molex pins finally arrived) and then I'll be glad to swap more info with you if you want to keep going. I really wanted 3 Ghz out of this rig at stock voltage... :p:

[Getttosmurf]

your oc will most likely increase with lower temps but leave your ihs on.
lapping might help a bit but you should not have to with that cooler unless your ihs or hs is crazy concave/convex.
Look at the mark your cooler leaves if you lift it straight up from the cpu when/if you decide to reseat.

e4300 has the same or lower power consumption then an e6600, e6700 x6800 or am i wrong here , should be easy to check this at intels website.
And still you have higher temps then people feeding alot more voltage thus increasing the consumed wattage of their cpus way beyoond what yours is consuming at stock vcore.

This make me belive that you have 1. bad contact with cpu 2. realy high case temp. 3. Thermal sensors on your mb that are way of showing much higher temps then your actuall temp.

Just trying to help you get most value out of your hw = ) and those temps dont seem right to me.
Im no expert so perhaps im wrong but hopefully someone with more experience could correct me if that is the case here :toast:

[Xilikon]

I also believe it's not normal to get temps that high.

Mine, running a very similar setup as yours but with a Thermalright SI-128. I was able to run at 3.38 GHz with load temps hovering around 60-65C when the heatsink is seated properly, all this using 1.4875v. If you used the push-pins retention mechanism, ditch it and get a real bolt-down LGA775 mechanism for your Ultra-120 since if well installed, you should beat me. Before getting the bolt-down one, I ran it and got 70-72C and when I swapped the mechanism, I noticed one of the pins is not 100% in, which is normal due to the :banana::banana::banana::banana::banana:y nature of this one.

[Forceman]

For what its worth, I have an E4300 with the Ultra-120 also, and I have mine at 370x9 with load temps around 72C (in a somewhat warm room) with 1.3625 vcore. I was not real happy with my temps initially so I reseated my HSF and found that it was making really poor contact with the chip. The Ultra-120 was not very smooth on the bottom, with noticeable machining marks so I lapped it with little success. Then I tried lapping the CPU and found that it was pretty significantly concave, and I mean pretty significantly, much more than I would have thought it would be. After lapping my temps dropped about 7C under load.

My load temps seem higher then I would have expected, but it is completely stable so I'm not all that concerned about it - and I'm much happier after lapping. I'd recommend checking how good the contact is with the chip and see if you might need to lap.

[chuchu]

Forceman, I am curious, what are your idle temps (say using CoreTemp 0.95 or Intel's Thermal Analysis Tool) ?

[Forceman]

Forceman, I am curious, what are your idle temps (say using CoreTemp 0.95 or Intel's Thermal Analysis Tool) ?

Going from memory, I believe it was around 40C using TAT and Coretemp 0.95 with load temps around 71C using Orthos small FTTs (for max heat production), and around 68C using blend. The idle temp seems quite high to me and I'm not real happy about it, but I don't have very good case airflow now so I'm living with it. With the case side off I was getting temps about 5C lower at idle and load - so improving case flow is my new project. Before lapping I was idling around 45C and loading at 75-78C.

On a side note, has anyone done any analysis on what kind of impact the thickness of thermal compound has on temps? I've always gone with a less is better approach, but after seeing how un-flat the IHS was before I lapped it I'm wondering if I might be going too thin. I might redo my HSF tonight and see if it helps any more - I might have too thin a layer of AS5 on there.

[eagle_]

Before you think about removing the IHS, consider this...

With the IHS removed, the die is actually even with or LOWER than the retention mechanism. If your HSF contact area doesn't fit INSIDE the retention mechanism it won't make contact with the die.Well, this happened to me. I still don't have a solution for it (i posted on the original thread).

[chuchu]

After thinking about this for a few days, I decided that the best and safest way to try to deal with my high temps was to lap the processor.

So last night I lapped it as follows:

320 Grit (dry), and then 600 (dry then wet), 1,000 (dry then wet), and 2,000 grit (dry then wet). There was little that could be done to the base of the Thermalright Ultra 120. Just some polishing with the higher grits, and that’s it. I then re-installed with Arctic Silver 5, but this time I used the thin line method which Arctic Silver 5 recommends, as opposed to the dab in the middle method.

Preliminary results are encouraging. At FSB325 (CPU speed 2925), idle temps drop from 35-37 Celsius to 30-32 Celsius (i.e. about a 5 degree Celsius drop at idle). At Orthos blend load, temps drop from 62-65 Celsius to 50-52 Celsius (i.e. about a 12 degree Celsius drop at load).

I have only done preliminary testing beyond fsb325, and as before, more volts are needed to make it work and temps start to rise fast. When I have time on the weekend, I may do more testing a higher volts, however I must say, I find running at fsb325 (CPU=2925) with cool temps to be very appealing.

My conclusion is that lapping should become standard operating procedure for virtually everyone.

[mezcal]

Wow Chu, that's much better! Hard to believe lapping made such a large difference-that cpu spreader must have been severely concave!
I finally got my bx2 box back up. Pushed the 4300 to 9x312 and now have:
Idle-33-35
Orthos blend (priority 9)-52-54
Orthos small fft (priority 9)-56
3DMark06-9125
This box is still in the garage so ambient is lower and airflow is better than where it will eventually live. It seems solid (1 hour of orthos and it didn't crash during 3dmark) but I'll orthos it overnight before making any stability claims (or trying to push it further.)
Also, I'm pleased with how well the bx2 controls the case fans. I now have the upper-chamber antec tricools (switch on high) connected to the fan headers. Reasonably quiet when poking along but when TAT hits 50 degrees it sounds like a helicopter taking off! Much too loud for normal use but 99% of the time it won't be that hot.
BTW, my cpu is same batch/week as yours. Must be flatter though...

[Getttosmurf]

Thats one big drop in temp from laping = ) glad you didnt settle with the previous temps now try and push it a little more = )


After thinking about this for a few days, I decided that the best and safest way to try to deal with my high temps was to lap the processor.

So last night I lapped it as follows:

320 Grit (dry), and then 600 (dry then wet), 1,000 (dry then wet), and 2,000 grit (dry then wet). There was little that could be done to the base of the Thermalright Ultra 120. Just some polishing with the higher grits, and that’s it. I then re-installed with Arctic Silver 5, but this time I used the thin line method which Arctic Silver 5 recommends, as opposed to the dab in the middle method.

Preliminary results are encouraging. At FSB325 (CPU speed 2925), idle temps drop from 35-37 Celsius to 30-32 Celsius (i.e. about a 5 degree Celsius drop at idle). At Orthos blend load, temps drop from 62-65 Celsius to 50-52 Celsius (i.e. about a 12 degree Celsius drop at load).

I have only done preliminary testing beyond fsb325, and as before, more volts are needed to make it work and temps start to rise fast. When I have time on the weekend, I may do more testing a higher volts, however I must say, I find running at fsb325 (CPU=2925) with cool temps to be very appealing.

My conclusion is that lapping should become standard operating procedure for virtually everyone.

[chuchu]

After pushing it a bit more, the limitations of my system appear to be FSB of 335 on the 1066 strap (i.e. CPU speed a bit over 3000), and FSB of 370 on the 1333 strap (i.e. CPU speed of about 3330). Interestingly, these were about the exact same limitations I ran up against before lapping, albeit at higher temps.

So as mentioned before, what I think I will be doing is letting up on the gas, and running this rig nice and cool on the 1066 strap at FSB 325 for a processor speed of approx 2925.

Thanks to everyone for their comments. Your help drove me to the conclusion that I should lap, and I am happy that I did.

[mezcal]

And a screenie.

57278

Looks good so far. I'll try a little more later.