860K OC Temp/Stability

[slyu9213]

I'm starting to see the final endpoint in finding a stable overclock of my current computer so I am going to use the OP to make an overview of what happened.

My Build.

AMD Athlon X4 860K
Gigabyte FM2A88XM-D3H
Gigabyte Radeon HD 7850 1GB
8GB of DDR3 RAM (1333MHz-2400MHz)
Corsair H100i (4 fan, push/pull)
Cooler Master N200 Case
7X 120MM fans, 1X ~60/70mm fan

- I tried OCing initially by raising the BLCK and then again with just CPU Multipliers
- Highest OC to boot to OS = 4.7GHz, but never was stable even with 1.55v. All sorts of stress programs crashed quickly
- Highest OC to be stable is 4.5/4.6GHz
- Idle temps were 50C+ in the BIOS, Room Ambient Temperature is around 20C
- Corsair Link would not detect the H100i nor it's fans so the fans spun at a low RPM, I got this issue fixed later
- Stock paste of the H100i was used so I reinstalled the cooler (Remove, Clean, Repaste, Reinstall) and the temps dropped significantly almost 10C+. That shows I had bad contact between the cooler and CPU.
- The initial OC was done with RAM running at the stock 1333MHz speeds. I later overclocked the RAM to 1866MHz and 2133MHz. 2133MHz wasn't stable so I decided to use 1866MHz.
- Adding a small 70mm fan to spot cool the VRMs helped drop the temp times, with the addition of curing time for the Ceramique 2 Thermal compound that was used. The FM2A88XM-D3H has no heatsinks on the VRM so I installed my own.

- GSkill 2 x 4GB Ripjaws Z 2400MHz was bought for $64
- Computer would not work at the previously stable 4.5GHz OC at 2400MHz RAM speeds, so CPU was underclocked to 4.2GHz
- Prime95 tests still failed in Blend tests which stress a decent amount of RAM. So I did my overclocks again but this time used Small FFT so I could first find the stable overclock speeds without stressing RAM. Conclusion 4.5GHz with 1.464v.
- I ran Memtest86+ and got nearly 200+ errors very quickly. The 2400MHz RAM has default timings of 10-12-12-31. Seeing that AMD's 2400MHz RAM have CAS 11-12-12-31 I lowered the CAS 10 to 11. Running Memtest86+ for 5 passes and there are 0 errors. Very surprising one number makes that much of a difference in stability.

Stable OC Settings for my PC build.

4.1GHz (1.344v [+0.054])
4.2GHz (1.368v [+0.078])
4.3GHz (1.368v [+0.078])
4.4GHz (1.452v [+0.162])
4.5GHz (1.464v [+0.180])

I will be sticking with 4.5GHz with this build as it was the highest-stable OC through several tries. I will stick with 2400MHz for the RAM and stick with the 11-12-12-31 timings as they are stable in Memtest86 unlike 10-12-12-31 which causes a lot of errors. With those out of the way I am now going to look into overclocking my IMC/NB as that seems to boost memory performances a bit for the 860K. 1800MHz is the default speed 2000MHz should be pretty easy with little to no voltage increases to the NB.

.:edit:.

These voltages are no longer accurate. With F7 BIOS things have changed. +0.180 use to be 1.464v but now is 1.512v. +0,138 will be 1.464v now. Temps are now running higher than before. So far 4.6GHz is lasting longer i Prime95 but still failing at some point. Good news is I might be able to get 4.5GHz at a lower voltage. Lots of more stability testing to do now to find the newest highest stable clock.

.:edit:.

I have made a conclusion for myself that the best Temp reading for my 860K on my FM2A88XM-D3H motherboard is the CPU temps from the Super I/O Controller (ITE IT8620E) and the CPU 0 Temps are the closest for Core Temps. Both of these temps can be found on HWiNFO64. My reasonings is that the numbers from the Super I/O Controller seems to match well with the Thermal Margin given by AMD Overdrive. With my 860K overclocked to 4.7GHz with 1.548 (1.55V) I get average CPU Temps of low 50Cs to low 60Cs. Core Temps are 40-50C. Additionally when AIDA64 shows the same numbers from the Super I/O Controller. If the Maximum Operating Temperature for a 7850K is ~72.4C then a 860K (7850K minus IGP) should be about the same. If I am getting a Thermal Margin from 15-30C (15-30C of Thermal headroom) then that should match with the fact that the 50-60C as there is 10-20C of headroom until it hits ~72.4C. I could be wrong but I won't know until my CPU dies a premature death.

 

[Darkbreeze]

Unless you live in the desert, or someplace where the ambient temperature in your room is like 40-45c, that idle temp is pretty high. I'd address that concern before I worry about going any bigger, or even staying where I am, on the overclock.

I don't like OCCT except as a secondary verification, as it's not a true measure of the CPU stability. It stresses memory and is not traditional in the methods it uses for stressing the CPU. Prime95 on small FFT is the way to do it. 20-30 minutes for basic thermal testing, 1-2 hours for confirmation of thermal compliance. 3 hours for basic error checking and of course, 24hrs for full stability.

Do you know why the Corsair link isn't working? That's a problem and should be resolved before you do much in the way of overclocking. You're probably much better off with a good air cooler than a liquid solution that doesn't work right. I'd get that working correctly if possible.

 

[iron8orn]

24hrs.. you don't even need that long with aida64..

 

[slyu9213]

Unless you live in the desert, or someplace where the ambient temperature in your room is like 40-45c, that idle temp is pretty high. I'd address that concern before I worry about going any bigger, or even staying where I am, on the overclock.

I don't like OCCT except as a secondary verification, as it's not a true measure of the CPU stability. It stresses memory and is not traditional in the methods it uses for stressing the CPU. Prime95 on small FFT is the way to do it. 20-30 minutes for basic thermal testing, 1-2 hours for confirmation of thermal compliance. 3 hours for basic error checking and of course, 24hrs for full stability.

Do you know why the Corsair link isn't working? That's a problem and should be resolved before you do much in the way of overclocking. You're probably much better off with a good air cooler than a liquid solution that doesn't work right. I'd get that working correctly if possible.

The computer is located in the basement of the house. We have our heating system set to 70F I believe and the basement is much cooler than that I believe. Honestly I don't know if the temperature readings are accurate. With AMD OverDrive my Thermal Margin was 70C but with the overclock it can hover around 60C sometimes but still in the mid-60C to 70C.

As far as Corsair Link, I've only read bad things about Corsair Link working on Windows 8.1. I tried the so called 'fix' and it didn't really work with me. The only thing I could do is try plugging the H100i into a different internal USB header or use a different usb cord to plug into an external usb port in the back of the motherboard. I have stopped overclocking and have it at 4.5GHz for now while I try to figure things out. In the mean time I am messing with my RAM to see if it will hit 1866MHz

.:edit:.

I have two good news. First, I tried another fix/workaround instruction and this time it worked. Corsair Link detected the H100i (Pump and Fans). I can now control the fans and pump. Second I was able to overclock my DDR3 1333MHz to 2133MHz. I'm going to now try tighten the timings.

 

[Darkbreeze]

That's an awfully big OC on your RAM. Don't be surprised if one or more modules burns out in a much shorter period of time than should be expected for a memory module. I probably wouldn't take a 1333 module past 1866, if I bothered with it at all. The performance gains from overclocking RAM, or even from just using faster RAM, is minimal in most applications and operating systems. Lower CAS latency settings offer more improvement than the clock speed, and even then you probably wouldn't notice the different of a CAS 7 module versus a CAS 10 module except maybe in benchmarks. Very unlikely you would notice it in real world applications.

 

[slyu9213]

That's an awfully big OC on your RAM. Don't be surprised if one or more modules burns out in a much shorter period of time than should be expected for a memory module. I probably wouldn't take a 1333 module past 1866, if I bothered with it at all. The performance gains from overclocking RAM, or even from just using faster RAM, is minimal in most applications and operating systems. Lower CAS latency settings offer more improvement than the clock speed, and even then you probably wouldn't notice the different of a CAS 7 module versus a CAS 10 module except maybe in benchmarks. Very unlikely you would notice it in real world applications.

Thanks for your reply. Yeah it sounded like too big of a OC. The RAM was stable with MemTest86+ and a few other Memory stressing programs but I got some graphical artifacts running on 2133MHz so I am running them at 1866MHz stable with no artifacts now.

I know that RAM speed has minimal effects on most applications but I learned that the 860K differs a bit. I don't know if it's the lack of L3 cache but the 860K had significant (in my opinion) gains in games from a 860K w/ 1866MHz compared to 860K with 2400MHz. For example in Overclock3D's review of the 860K, a stock 860K with 2400MHz got 17FPS boost in Hitman Absolution, ~5FPS boost in Sleeping Dogs. This is on lower graphical settings and with a fairly weak GPU (R7 250x) though.

Currently what I am doing most with my computer is playing Dragon Age Inquisition. With the 860K OCed to 4.5GHz and with 8GB 1333MHz RAM and 1GB HD 7850 I got 30-40FPS on everything on Low except a few options on High/Ultra (Mantle). With the RAM OCed to 1866/2133MHz the overall FPS was a bit higher in the 50-60s but there were still horrible dips. Thats mostly because Mantle api performs worse than DX11 on Inquisition. In anycase Inquisition on DX11 got a significant amount higher AVG FPS and minimum FPS seems to have increased up 10FPS+, and this is with everything on High with one or two ultra. I'm not saying RAM speed matters in general but in the case of my processor it seems to help a little more (like APUs) than other processors

 

[Darkbreeze]

Are you using a GPU or the motherboard's iGPU?

 

[slyu9213]

The Athlon 860K is a 7850K (APU with a strong IGP) with it's IGP disabled from the factory. It's for the FM2+ socket which the IGP is in the APU itself not the motherboard. So no I am using a discrete Radeon HD 7850 video card not any sort of integrated graphics card but still have a slight performance increase just like Overclock3D and a few other 860K users experience.

 

[Darkbreeze]

I know what the 860K is my friend. This is what I do, day in and day out. However, there are a lot of motherboards that have their own integrated graphics that are not reliant on the CPU to have an integrated graphics processor although I don't think any of the FM2 motherboards have non-APU integrated graphics, or it's extremely rare if they do.

If you are seeing a real world performance increase using the faster RAM, overclocked or actual, I'd say it might be worth it to invest in a faster module, preferably with an already low latency. According to this article, that chip really does sport some decent gains with the faster RAM but it may be that your modules are simply not stable at that speed and if you want to take advantage of the performance bump you may have to do it physically rather than with firmware.

http://www.overclock3d.net/reviews/cpu_mainboard/amd_athlon_860k_black_edition_cpu_review/2

 

[slyu9213]

I know what the 860K is my friend. This is what I do, day in and day out. However, there are a lot of motherboards that have their own integrated graphics that are not reliant on the CPU to have an integrated graphics processor although I don't think any of the FM2 motherboards have non-APU integrated graphics, or it's extremely rare if they do.

If you are seeing a real world performance increase using the faster RAM, overclocked or actual, I'd say it might be worth it to invest in a faster module, preferably with an already low latency. According to this article, that chip really does sport some decent gains with the faster RAM but it may be that your modules are simply not stable at that speed and if you want to take advantage of the performance bump you may have to do it physically rather than with firmware.

http://www.overclock3d.net/reviews/cpu_mainboard/amd_athlon_860k_black_edition_cpu_review/2

I'm sure you knew it too. I just found it weird that you asked if I were using integrated graphics on the motherboard as little to no FM1, FM2, FM2+ motherboards have IGP in the motherboards due to all of them being on the CPU/APU chip itself. Although there are motherboards with IGP still I believe all the newer sockets come with the IGP in the processors now.

That was the website I was referring to about performance. I have been looking at 2400MHz modules for a while but with the current modules now running at 1866MHz I may be able to wait some more. Not sure I want to invest up to ~$300 on CAS 9 2400MHz to pull a few more FPS out. Better off saving a little more and getting a better GPU than what I have, right? Although getting some cheap CAS 10 2400MHz modules and seeing if I can tighten the timings might be an okay idea.

 

[Darkbreeze]

How much RAM are you running? I know some of these configurations have trouble with the higher settings when running quad modules. And what country are you in?

 

[slyu9213]

I am running two 4GB DIMMS out of four slots. The motherboard can take upto 64GB of RAM but 16GB DIMMs don't exist for consumers yet and not worth the money when it comes up. Four 8GB DIMMs (32GB) is cool with the ability to use 16GB as a RAM Disk but I have a SSD for OS + Programs, small 5400RPM HDD for Documents, Videos, Music, etc, and a 7200RPM 1TB HDD for games. If you can use RAM Disk as a pagefile that is what I would use it for I guess.

8GB of RAM is what I am running now so 2 x 4GB 2400MHz is probably ideal budget wise. The motherboard is a Gigabyte F2A88XM-D3H. Not sure if there will be trouble when running quad modules but it is possible. In that case I guess going with two 8GB modules will give me the largest amount of RAM while not populating all four slots.

Lastly I live in the US.

 

[Darkbreeze]

RAMdisk for a pagefile is a no no. What you CAN use it for is cache files, windows temp files, browser temp files and cache, temp map folders for games and a few other things. Honestly, I've played around pretty extensively with RAMdisks and once I moved to SSD's, especially modern one's, I didn't notice any improvement over just using the RAM as RAM. Browsing was very slightly faster and game levels loaded a little faster, but it was mostly not worth it. I've got several paid and unpaid versions and none of them worked better than the others.

http://www.overclock.net/t/1193401/why-it-is-bad-to-store-the-page-file-on-a-ram-disk


I probably wouldn't go beyond 2133mhz as the memory controller doesn't natively support it.

http://www.eteknix.com/athlon-x4-860k-cpu-available-starting-next-week/


This is what I'd go with, and you can try to tweak it up to 2400mhz if you feel the need from there, probably much more easily than trying to tweak 1333mhz modules up to 2133, heh:

PCPartPicker part list / Price breakdown by merchant

Memory: Mushkin Redline 16GB (2 x 8GB) DDR3-2133 Memory ($144.99 @ Newegg)
Total: $144.99
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2014-11-29 04:15 EST-0500

or this:


PCPartPicker part list / Price breakdown by merchant

Memory: Mushkin Redline 8GB (2 x 4GB) DDR3-2133 Memory ($84.99 @ Newegg)
Total: $84.99
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2014-11-29 04:16 EST-0500

 

[slyu9213]

Awesome. I guess you're right. The 860K's memory controller officially only supports 2133MHz but will run 2400MHz when overclocked. Safest way is to get reliable 2133MHz and then overclock them to 2400MHz later. Thanks.

.:edit:.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231585&cm_re=ripjaws_z-_-20-231-585-_-Product

These are probably what I am getting 2 x 4GB 2400MHz, its only $63 right now and the 4 x 4GB and 4 x 8GB sets are on the QVL list for my motherboard. Best way for compatibility.

 

[Darkbreeze]

I have never in my life paid any attention to the QVL. For the most part, neither does anybody else including the product reps from the MB mf'rs who will tell you point blank that they only test enough modules to flesh out a list and that's it. It has absolutely nothing to do with what is EXLUDED from compatibility, and everything to do with "that's enough to show that most modules work with our product". The only thing you really need to worry about with modern memory and motherboards is that the listed specs are compatible, that the memory is or is not ECC depending on your board and that you don't accidentally get a memory part number with the wrong number of chips as some boards don't support some of the more unusual or exotic arrangements.

The rest of what you've stated sounds fine. Personally, I like the Mushkin Redline modules much better than any of the Patriot, Team, Corsair, AMD or Geil modules. I try to use Kingston, Crucial or Mushkin, preferably Mushkin unless the cost is prohibitive since they are entirely manufactured in the USA and I've personally never had a Mushkin module present problems. I'm sure they do, like anything electronic, but I've used them since the early 90's and still have some old DDR1 modules laying around that I'm sure would work just fine if I popped them into an old board. Kingston and Crucial have also been around much longer than the others and Crucial, which is Micron, has been doing it longer than anybody else other than IBM I'm pretty sure.

 

[slyu9213]

I have never in my life paid any attention to the QVL. For the most part, neither does anybody else including the product reps from the MB mf'rs who will tell you point blank that they only test enough modules to flesh out a list and that's it. It has absolutely nothing to do with what is EXLUDED from compatibility, and everything to do with "that's enough to show that most modules work with our product". The only thing you really need to worry about with modern memory and motherboards is that the listed specs are compatible, that the memory is or is not ECC depending on your board and that you don't accidentally get a memory part number with the wrong number of chips as some boards don't support some of the more unusual or exotic arrangements.

The rest of what you've stated sounds fine. Personally, I like the Mushkin Redline modules much better than any of the Patriot, Team, Corsair, AMD or Geil modules. I try to use Kingston, Crucial or Mushkin, preferably Mushkin unless the cost is prohibitive since they are entirely manufactured in the USA and I've personally never had a Mushkin module present problems. I'm sure they do, like anything electronic, but I've used them since the early 90's and still have some old DDR1 modules laying around that I'm sure would work just fine if I popped them into an old board. Kingston and Crucial have also been around much longer than the others and Crucial, which is Micron, has been doing it longer than anybody else other than IBM I'm pretty sure.

Haha. I never looked at the QVL either except when some of my RAM didn't work on a new motherboard. The only reason I looked at the QVL is that I am sick and tired of any return processing. I was trying to keep the purchase in the $60 range honestly. In order to do that with the Mushkin, Corsair, or AMD modules I had to buy one 4GB DIMM which wouldn't be enough to game on. Seeing that the RAM I am going to buy is the same exact model (not the same amount of DIMMs) from the same brand, same speed, same CAS, same voltage, and both are non-ECC I felt I had less chance of having to return them except for DoA. But the real choice was that these 2 x 4GB sticks are only $63.99. Obviously I can't find 1 X 4GB 2133MHz or 2400MHz at $32 so I decided on the G.SKILL. Others with my motherboard or FM2/+ users left good reviews.

Both the memory you provided and the one I am purchasing (already ordered) have great reviews. I don't think I could have gone wrong anyway except the fact that G.SKILL is probably not manufactured in USA (by your comment). The current Corsair XMS3 are the first DDR3 modules I bought and have not bought any new memory excluding RAM for my laptop and macbook since then. Before DDR3 I just used Samsung/Hynix modules that had no heatsinks.

p.s My PC has an All-Black with a hint of aluminum/silver as the color theme so the Red-heatsinks was a deal breaker.

 

[slyu9213]

Well good news is that I think the high temps were due to nonproper installation or the fact that I used the stock Thermal paste on the H100i. Removed, cleaned, and reapplied paste (pea sized-ceramique 2) and the temps are now 35 C at idle compared to 40-50 on idle.

 

[Darkbreeze]

That is much better, but still maybe a bit higher than I'd usually like to see. 10-30c is average and of course will depend a lot on the room's ambient temperature. Still, it's an improvement. The other thing you want to look at is fan configuration. Do you have the fans for the radiator configured to suck air through the radiator from inside the case or blow air through the radiator from outside the case?


Case cooling is important too since you have likely used two fan locations up with the radiator, which limits how much hot air you can get OUT of the case if it's a top mounted radiator since that usually only leaves a single exhaust fan at the back. What is your case fan configuration like? Any pics of all fan locations and direction of airflow would be helpful in assisting you with increasing the cooling efficiency.

 

[slyu9213]

That is much better, but still maybe a bit higher than I'd usually like to see. 10-30c is average and of course will depend a lot on the room's ambient temperature. Still, it's an improvement. The other thing you want to look at is fan configuration. Do you have the fans for the radiator configured to suck air through the radiator from inside the case or blow air through the radiator from outside the case?


Case cooling is important too since you have likely used two fan locations up with the radiator, which limits how much hot air you can get OUT of the case if it's a top mounted radiator since that usually only leaves a single exhaust fan at the back. What is your case fan configuration like? Any pics of all fan locations and direction of airflow would be helpful in assisting you with increasing the cooling efficiency.

The case I am using is a nice and small, micro atx case, Cooler Master N200. The H100i is located at the front of the case. It is being used in a push/pull setup (2x SP120L in the front and 2x Cooler Master fans on the other side of the radiator). Additionally I have a 120mm fan on the side of the case blowing air onto the GPU/VRM area. Finally I have two 120mm fans as exhausts, one in the back and one on the top. I don't really have pictures but if you look up the case I have all the fan locations occupied.

I don't really know what the room's ambient temperature is. Anywhere from 18C to 21C as our heater is set to about 70F, and the room being in the basement it won't be 70F. Currently I have the CPU overclocked to 4.6GHz with VCore at 1.512V (drops as low as 1.46/1.476V while stressing). With the two SP120L set to 2700RPM and the other fans not manually controlled I am getting these temps while running Prime95 Blend Test.

Corsair Link
CPU: 80-90C
Motherboard: 52C and 32C (Two Sensors)
H100i Temp: 30.3C
SSD/HDD: 27C
GPU: 30C

AMD Overdrive
CPU: 31C Thermal Margin

HWinFO64
CPU 0: 38.5C
CPU 0 Package: 87.4C
Motherboard: 32C
CPU: 51C
Temperature 4-6: 45C

The CPU seems to run a little warm to hot no matter what. Another user with the same CPU and Motherboard is having issues with 50C at idle.

.:edit:.

Here's a simple picture showing how the fans work in my computer.

 

[Darkbreeze]

How much thermal paste are you using during application? Fan configuration seems fine. Maybe double check that all fans are in fact blowing the correct direction. You would be surprised how many times I've seen configurations where people were absolutely certain the orientation was correct until they checked and then found, hey, one of my exhaust was sucking in instead of blowing out or visa versa. Just a thought to double check. Also, double check that both of the fans in front of the radiator are positively blowing into the radiator and that both fans behind the radiator are sucking in.


I've actually seen more than one instance where the fans on the pull side of a radiator were not as capable as the pressure fans blowing in and were actually restricting airflow rather than increasing it. With your inner fans not manually controlled and your two pressure fans set to 2700RPM, I'm betting there is a restriction caused by the inner fans.

If you do use four fans in push pull, you always want the more powerful fans on the exhaust side of the radiator pulling so that they are not restricted by the speed of slower fans or fans not capable of equal RPM. A more powerful fan sucking will actually still suck air through a slower fan, but cannot push air through a slower fan due to the pressure and turbulence created by the physical presence of the second fan.

 

[slyu9213]

Well I use to spread thermal paste out onto the CPU but I began using pea-sized to a little less in the middle of the CPU and then letting it spread naturally when attaching/securing the heatsink/waterblock. I double checked my fans and they are all positioned/orientated in the correct direction.

I will switch the pressure fans to the back to suck the air through the normal fans and radiator later. I can see airflow restrictions happening also when the pressure fans are running at 2700RPM while the other fans are running rather slow. Excluding the two SP120L all the other fans are all plugged into a small circuitry board that is powered by one 4pin molex. So there isn't a way to change speeds.

 

[Darkbreeze]

Ah, I see. Ok. Next time you paste, try using only a rice grain sized amount. That's what I've been using for years and I find I get full coverage whether it's a convex or concave lid or heatsink. It's plenty for a true flat surface as well as obviously less or none would be needed if both surfaces were actually truly flat. A pea sized amount might be more than necessary although it's probably ok either way, especially if you went kinda light on the pea estimate.

 

[slyu9213]

I'm about to rearrange the fans for the H100i. What do you think about this? As the H100i is a 240mm or 2 x 120mm radiator I wonder if I should pair the same fans together. For example I have 2 SP120L and 2 Cooler Master Fans. What if I have the SP120L Push/Pull on one half of the radiator and the other two fans push/pull on the other half. By doing that at least there would be no direct problems on the fans. Although having two non-radiator fans doing push/pull together may have some weakness.

.:edit:.

Also is there any point in overclocking with the BCLK settings when my CPU multipliers are unlocked? The only other reason I see for using BCLK would be for memory and north bridge if they were locked but both have unlocked multipliers. Currently I am OCing with BCLK to figure out the stable settings for each tier of speed and what temperatures they run at while being 100% stressed. Once that is done I'll probably test it in games to see what kind of performance differences I am getting.

 

[Darkbreeze]

I never use BLCK as it affects USB, PCI, GPU and the whole system. It's not necessary. Any reachable overclock that can be stable should be obtainable using multiplier and voltage on systems that have unlocked CPUs and the ability to make changes in the BIOS. Using BLCK is just asking for unnecessary problems in other areas.

I think the idea with the fans is a good one, but make sure you've got each pair of fans running at the same speed, which may require Y splitters to ensure each pair of fans receives the exact same signal from the controller.

 

[slyu9213]

It seems 4.5GHz is the best clock for my CPU/Motherboard combo. What I did was OC by raising the BLCK in increments of 5-10 until I found my max clock. With a +0.162v (Stock voltage is 1.26-1.375) I am able to run at 4.5GHz on Prime 95/OCCT with Max temps a little over 70C. Before I reseated the water block I was getting higher temperatures and needed higher Vcore to achieve 4.5GHz. At idle the CPU gets 1.48ish volts and when stressed it's 1.44V. I believe the max recommended Vcore is 1.5-1.55V but I don't plan on getting higher as the additional voltage will raise my temps to more uncomfortable levels.

Now that I feel comfortable about the CPU clock I'm going to look at North Bridge speeds as they seem to give a boost in memory related performance (as Overclock3D states). Once the 2400MHz RAM arrives later this week I will do another stress test run to see if everything is stable and then try tightening the timing a little.

Thanks for the help, suggestions, and good discussion so far.