1366 x58 Xeon Enthusiast overclocks club

So based on my own quote above. Will one of these xeons really benefit me? Computer is in my sig as well.
I say yes, it could benefit you, but there could be one big catch. Not all the EVGA boards will support these Westmere cpus without a hard mod. The hard mod includes some soldering, and the info on what to solder is almost impossible to find. EVGA did the mod for some people for a small cost in the past, but I'm not sure they would still do it now.
As for what EVGA boards work, well, there used to be a list somewhere but I can't find it now. I'm still looking but it's hard to find. Let us know what model you have. The model number should have E758, E760...etc..included. This will be printed on the board itself.
You would also need the latest bios update for these Westmere cpus, and of course you would need to to the bios update before removing your current cpu.

edit: I found the info:


From EVGA technician _NickM, these board revisions have support:

E757 - rev 1.1
E758 - rev 1.2
E760 - rev 1.1
E762 - rev 1.1
E767 - all
E768 - all

Ife you have 1.0 rev of E760, you can't use -EP Xeons; and it isn't about updating a BIOS, your board needs to be wired for their support, thus board revisions. Check by looking at your board at the lower left corner near the PCIe slots, it should be in white letters.

http://forums.evga.com/tm.aspx?m=1764611
 
23 x 175 is 4ghz, I have run 21 x 200 for a 4.2 and things were ok.

I haven't tried for real high o/c I figure 2.9ghz going to 4ghz is not a bad o/c as it is.
Plus my temps with AIO Hydro coolers H70, H90, H80i my temps stay between 41c to 54c 100% load.
 
E758-A1
Rev 1.1

So mine is not supported?

Nope and they will charge $50 to do it. I was able to convince them to do it for free but it took a TON of whining and complaining to a point that I just sold mine and got an Asus mobo instead.
 
Is it normal for an x5650 to have trouble holding down 4 GHz? It ran Prime95 in-place FFTs all night without a hiccup. Same with 3DMark FireStrike on a loop (not at the same time). But I got that wonderful WHEA_UNCORRECTABLE BSOD while playing Dragon Age: Awakeings. Normal?
 
Is it normal for an x5650 to have trouble holding down 4 GHz? It ran Prime95 in-place FFTs all night without a hiccup. Same with 3DMark FireStrike on a loop (not at the same time). But I got that wonderful WHEA_UNCORRECTABLE BSOD while playing Dragon Age: Awakeings. Normal?

Well we would need more info on what bios setting your currently using, one thing i have learned is not all asus boards use the exact same settings but im sure we could give you better assistance if we knew all the relevant settings you are trying. before i could suggest alternative settings that might help you i need to know exactly all the settings you using. (as well with people with your exact board) please list all settings or screenshots of all relevant bios settings before proceeding, but with that being said 4ghz should be doable. If feel im pretty good with these things but there are several people checking this thread who are pretty much masters in these things (Deimos, Zoson, DASHlT,) but there are many others:D Some can help you with pm's as well
 
Last edited:
Well we would need more info on what bios setting your currently using, one thing i have learned is not all asus boards use the exact same settings but im sure we could give you better assistance if we knew all the relevant settings you are trying. before i could suggest alternative settings that might help you i need to know exactly all the settings you using. (as well with people with your exact board) please list all settings or screenshots of all relevant bios settings before proceeding, but with that being said 4ghz should be doable. If feel im pretty good with these things but there are several people checking this thread who are pretty much masters in these things (Deimos, Zoson, DASHlT,) but there are many others:D Some can help you with pm's as well

Thanks for the response; I didn't want to presume. I once replied to a thread (different message board, different issue) where I posted all of my specs/settings/etc and then the thread just died, no responses at all :D. Didn't want to kill this thread as well. I'll post photos of my BIOS settings in 30 minutes or so.
 
Oops... I forgot, most of my BIOS settings are on auto right now. Here's my story. A few months ago I got my x5650 to replace my old i7 930. I got it rock solid stable by upping BCLK and CPU voltage up to 4.17 GHz. Days under Prime95, no crashes while gaming, etc. The only instability it had was that every time I would try to wake from sleep it would BSOD. I started investigating to see if it was a clock speed related issue and found that at 4.0 GHz and higher, the problem was reproducible. Unfortunately, when I tried one last validation step, taking it up to 4.05 GHz, the resulting BSOD somehow corrupted the bootloader. I ended up having to reset the BIOS and reinstall Windows. Since then I've been too busy to mess with it but I got sick over break and decided to take the time and tinker. I just brought up clock speed in large hops (3.2 GHz -> 3.6 GHz -> 4.0 GHz) and checked for stability with Prime95 and looping 3DMark.

In all cases my tests where stable; however, the BSOD upon waking from sleep is back (WHEA_UNCORRECTABLE). Can someone look at the BIOS settings below? I don't remember what my "working" ones were from before but they were largely held over from my i7 930. If someone could particularly suggest CPU and PLL voltages to optimize around, that would be great. Also, what stability tests should I use, since Prime95 is clearly not working?

jU5gNQGl.jpg

FwAQom2l.jpg

3dlem2sl.jpg

mwxK17Yl.jpg

Z2z5X5ol.jpg

4VctCVjl.jpg


PS. The RAM voltage is the result of me typing in "1.65". If I set it to auto, the voltage goes slightly below 1.65, and my understanding of these OCZ gold sticks is that they don't like being undervolted.
 
nice work on posting the settings...rampage owners will likely help the most but you could try 1.35 qpi/dram and moving vcore up to 1.35 as well.......also IOH voltage 1.25 and dropping qpi link data rate speed to lowest option.....for whats its worth anyway...im sure rampage owners wil have better suggestions....but you can at least try these ideas:)
 
First of all, change your tuning mode to XMP if your memory supports it, otherwise change it to manual.

Set your memory speed manually, keep in mind that you MUST overclock if you want to use memory speeds above 1333 otherwise no POST.

Set your UCLK to 2 x your memory speed (so 1600 memory should have a UCLK of 3200) you can get a small performance increase by going slightly higher than memory speed.

Set your QPI link to one above the lowest setting (Slow mode really is just that, your system will feel noticeably slower)

Voltage wise, set everything manually to stock voltages apart from:
CPU - start at 1.2V for 4Ghz and go from there.
QPI - Do not exceed 1.35, only increase this above stock if increasing CPU voltage doesn't help stability, also, your DRAM voltage should not be more than +.5v higher than the QPI voltage, so if your ram needs more than 1.65v you need to bump this up as well.
PLL - Lower is better in some cases, in my testing on 4 different CPUs with two different boards, increasing PLL didn't help stability at all so I would suggest leaving this alone unless you want to try lower voltages (I'm running 1.35v but this is very much different per system).
 
Did i get any of it right Deimos?:D......i knew you guys were way better than myself but i tried to help best i could lol...looking at your suggestions i would say hes in good hands (nicely done)
 
Hi all,

I just got the X5670 chip for my 132-BL-E758-TR board and got it up and running after the resister mod. However I noticed something very strange when overclocking, the uncore frequency, as reported by CPU-Z as NB Frequency, is stuck at ~4000Mhz when I'm running at 22x200bclk turbo off.
I've tried all the CPU Uncore mulitpliers from 12x to 18x but the Uncore number won't budge as reported in CPU-z.
I'm having trouble stabilizing the system with CPU VTT voltage staying under 1.4V which is the limit for these chips and am wondering if there's something I'm doing wrong, seems that the Uncore speed is causing instability but the board doesn't seem to want to adjust it for whatever reason.

Thanks for any advice and thoughts!
 
If you're on the latest bios you're SOL. This is part of what I was talking about when I said that older x58 motherboards don't correctly support the Gulftown/Westmere architecture.

You either need a BIOS update with microcode to support 1.5x Uncore, or your board simply physically can't do it. And the Uncore limit for these chips is 1.35v, not 1.4v.
 
Hmm I am on the latest bios which does support gulf towns, in fact even evga will perform the modification under warranty (moving a resistor) to support westmere which I've done. I've seen others running same board and chip and had goods results overclocking so figured it was some config issue I'm missing
 
Actually I just checked the supported chips list on eVGA site and Xeon X5670 is definitely supported by latest bios
 
Key words: "don't correctly support"
Yeah, the MINIMUM uncore ratio is 1.5x memory speed for Gulftown/Westmere, but you can run it faster.

Your board was designed for 45nm cpus which has a minimum uncore ratio requirement of 2x memory speed. See if lowering your memory from 1000MHz to 800 brings your uncore down to 3600MHz.
 
OK got it, just so I understand it correctly, my memory is at 1600 but uncore is at 4000 due to the multiplier on it seems stuck. Does this mean I need to scale the bclk back as I don't think changing the memory ratio will fix it? Currently uncore multiplier 12-16x which is 2400-3600 has been tried with no effect, both are above the 1.5x memory (1600) threshold
 
so funny enough, what I have discovered is that eVGA BIOS has this funny feature where some values would become stuck such as uncore frequency in my case. I forgot about encountering other things stuck with my Bloomfield before and after resetting all values to default and reworking it from there, uncore multiplier is now unlocked and working correctly
 
Okay, so after lurking here for a bit, I've finally decided to pull the trigger on upgrading my i7-950 to an x5600 series - but after reading many pages of this thread, I'd like a little advice on what I should go with from the experts here - 5650, 5660, or 5670? Right now, my 950 is running @ 190x21 to hit 4GHz, on a Asus P6T Deluxe, as I had issues with 20x. In the summer, I clock it down to 200x19 as 3.8 runs at least 6-8*C cooler. For reference, I'm currently at 70-73*C (68*F ambient room temp) under 100% load running 8 threads of BOINC with other misc DC projects also running.

So my question - Knowing I have the known issue with the 20x multiplier not always working well, should I settle for the x5650 and lower it to 19x, get the slightly more expensive and much less available 5660 for the 21x/4.2Ghz capacity, or pay another $25 for a 5670, knowing the 22x may not work (guessing only due to the even multiplier issue)? Right now the best ebay prices I see from reputable sellers are $74/$96/$120...

Thanks!
 
Jathanis,

I don't think you should shun the even multis as every chip is different and getting one is nothing but the luck of the draw. I've owned two X5670s on the P6T Deluxe board and the first chip required pretty much the same exact amount of voltage to stabilize at 4.2GHz with either even or odd multis. The second Xeon preferred even multis and required less voltage compared to odd multis for the same 4.2GHz overclock.

You should take two things into consideration when choosing a CPU in this case. First, what overclock are you aiming at? And second, does your motherboard handle high BLCK frequency well? The reason I went for the X5670 is that my integrated sound card goes nuts once I hit ~185+MHz BLCK, so 24x multi helped me big time on hitting 4.2GHz (175x24). Not to mention, running lower BLCK will require less QPI/VTT voltage, too.
 
Jathanis,

I don't think you should shun the even multis as every chip is different and getting one is nothing but the luck of the draw. I've owned two X5670s on the P6T Deluxe board and the first chip required pretty much the same exact amount of voltage to stabilize at 4.2GHz with either even or odd multis. The second Xeon preferred even multis and required less voltage compared to odd multis for the same 4.2GHz overclock.

You should take two things into consideration when choosing a CPU in this case. First, what overclock are you aiming at? And second, does your motherboard handle high BLCK frequency well? The reason I went for the X5670 is that my integrated sound card goes nuts once I hit ~185+MHz BLCK, so 24x multi helped me big time on hitting 4.2GHz (175x24). Not to mention, running lower BLCK will require less QPI/VTT voltage, too.

pure win. this is what i will try and do as well
 
Hamm I wonder if thats my issue, current x5670 will run at 200x21 for 4.2ghz at 1.285vcore almost stock volts for everything but 200x22 for 4.4 and I can't stabilize for the life of me even at 1.385vcore and 1.35vtt. I didn't try higher volts as performance gains over 4.2 won't be worth the heat and noise
 
Hamm I wonder if thats my issue, current x5670 will run at 200x21 for 4.2ghz at 1.285vcore almost stock volts for everything but 200x22 for 4.4 and I can't stabilize for the life of me even at 1.385vcore and 1.35vtt. I didn't try higher volts as performance gains over 4.2 won't be worth the heat and noise

As regards your 4.4GHz OC, have you tried 210x21 or 183x24 configurations? I've noticed for some reason some chips tend to have one or two "bad" multipliers with which they are very unstable. It may also be your chip prefers odd multis only, but you'll never know unless you try other BLCK and multi combinations.
 
Yep, going to try it when I get home today, I can get 23x multiplier with turbo so theres a few things to be tried. If my chip likes odd, perhaps 23x200 (I wish haha)?
 
Thanks for the reply - from all the i7 guides I've read, I was under the impression that it was a chipset + even multiplier known issue, not chip dependant. Honestly, I'm looking at this upgrade for 3 reasons - first, I want the extra 2c/4t for our team's distributed computing projects, as this upgrade gives me an instant 50% boost in work done. Second, the money I save by keeping my mobo/RAM allows me to spend more to upgrade my poor 9800GTX+ for gaming (my obvious bottleneck here). And third, since I've been having summer heat issues, I like the fact the the Xeons run cooler than the i7s. So as long as I get a solid 4.0 Ghz I can run year round and not have to back off for the summer, I'll be happy! 4.2 to 4.4 would just be a happy bonus on my system - I currently don't see much difference from 3.8 to 4.0, it's only when a power outage causes the bios to reset to 3.06 that I notice my machine running slower...​

As for BLCK, I've ran it at 180, 190, and 200, all without issue. Due to my RAM, I've never attempted to run it over 200, but I'd suspect it has some room. But it sounds like I should go with at least a 5660 for the extra multiplier, because if I get a 5650 and the 20x doesn't work like with my i7, I'll either be stuck at 3.8 or have to push BLCK over 200 to reach 4 Ghz - I about have this? :D

Jathanis,

I don't think you should shun the even multis as every chip is different and getting one is nothing but the luck of the draw. I've owned two X5670s on the P6T Deluxe board and the first chip required pretty much the same exact amount of voltage to stabilize at 4.2GHz with either even or odd multis. The second Xeon preferred even multis and required less voltage compared to odd multis for the same 4.2GHz overclock.

You should take two things into consideration when choosing a CPU in this case. First, what overclock are you aiming at? And second, does your motherboard handle high BLCK frequency well? The reason I went for the X5670 is that my integrated sound card goes nuts once I hit ~185+MHz BLCK, so 24x multi helped me big time on hitting 4.2GHz (175x24). Not to mention, running lower BLCK will require less QPI/VTT voltage, too.
 
So I have had a x5670 for about a month now. I am very happy with my daily driver OC but i do have some questions for the thread (which took me many days to read in my spare time)

First is my board
Its Gigabyte EX58-UD3R (rev1.6). I havent seen many here with it. What is the thoughts on it?
I maybe i am lucky but it seems to have no problem all the way up to blck220 which is as far as i have gone with it (using a low multiplier just to find my limits)

I find the dynamic voltage control to be one of my favorite features. I didnt even use it on my 920 but with this westmere, my chip runs like its stock.
speedsteps down to 2.2ghz at (.991v)
up to 4.2ghz full load on all cores (1.29v),
and 4.4ghz up to 2 cores loaded

So with dynamic voltage this chip operates pretty much like it would if it were a 4.2ghz chip with turbo. I set the volts to normal and dynamic control to (+.1).
BLCK is 175.

Do the legendary asus boards have a comparable feature? I havent seen anyone bring it up. Its a really neat ability that has given me an ultra quiet and rick solid system for a month now. Just think others might find it just as useful as me.

Also want to bring up the CPU PLL. People such as primetime suggest putting this at 1.9v for even mild overclocks. Is this something people are finding necessary? I havent touched it and have a full range of overclocks stored in my bios. I have actaully lowered my CPU PLL after reading this thread just because.....if intel says over 1.9 will degrade your CPU then why run it as high as 1.8 if you dont need to? I am just wondering though, is there anyone here that has seen bumping the CPU PLL up to 1.9v do any good on a westmere OC. Like you know for sure this alone helped you stabilize. If you havem what blck and overclock where you trying to stabilize.

I havent found raising it useful on my chip. Well, not yet.

Lastly, there seems to be a lot of x58 boards giving out in this thread. People remember, these boards are many years old. Sometimes its not worth it to push for such a small bump up in speed. This chips start to require a lot more once your get to the middle of 4ghz. Pushing for that last 100mhz might be pushing these old boards over the edge. See, all motherboards die eventually. Blow up your x58 and your will be out a lot of money. The only options are another old used board, an expensive side-grade to a 4790k, or drop a grand on a haswell E.

These westmeres can do great things for an old platform, but keep in mind how old your board is when your pushing it to the max limits.

Thats my first post here in what looks to be a great forum.
Oh, and hi everyone!!!!!
 
My board will be 5 years old come March.. but it is a EVGA A-1 product with lifetime warranty and they still stand behind it.
 
Do the legendary asus boards have a comparable feature? I havent seen anyone bring it up. Its a really neat ability that has given me an ultra quiet and rick solid system for a month now. Just think others might find it just as useful as me.

Yeah my P6X-58D Premium has such a feature, it's called offset voltage control and the adjustment is + (or -) above the stock voltage however I found a major caveat that could result in a dead CPU so I stopped using it, the offset voltage is highly dependent on the level of overclock and multiplier settings. So if for example, you set a high BCLK and low multi, you will need a much higher increase to the offset (say +.2v) to achieve the same VCORE as a lower BCLK with a higher multi. so if you forget to adjust it after increasing your multiplier you could deliver a much high VCORE than what you intended and kill your CPU (I once ran at 1.5V by mistake but I'm running a custom WC loop so it didn't cause any damage).

I also saw very little difference in temps going from offset to static, perhaps 2C if that. C-States makes the largest difference.

Interestingly this feature is absent from the Rampage III Extreme but it's pretty easy to see why.

Also want to bring up the CPU PLL. People such as primetime suggest putting this at 1.9v for even mild overclocks. Is this something people are finding necessary? I havent touched it and have a full range of overclocks stored in my bios. I have actaully lowered my CPU PLL after reading this thread just because.....if intel says over 1.9 will degrade your CPU then why run it as high as 1.8 if you dont need to? I am just wondering though, is there anyone here that has seen bumping the CPU PLL up to 1.9v do any good on a westmere OC. Like you know for sure this alone helped you stabilize. If you havem what blck and overclock where you trying to stabilize.

I have stated many times that increasing PLL above stock doesn't increase stability. Others have mentioned that lowering the PLL can help in bringing down your other voltages (but this takes a lot of testing). I'm currently running mine way under at 1.35V. Go lower if your board can do it.
 
Originally Posted by Ocre View Post
Also want to bring up the CPU PLL. People such as primetime suggest putting this at 1.9v for even mild overclocks. Is this something people are finding necessary? I havent touched it and have a full range of overclocks stored in my bios. I have actaully lowered my CPU PLL after reading this thread just because.....if intel says over 1.9 will degrade your CPU then why run it as high as 1.8 if you dont need to? I am just wondering though, is there anyone here that has seen bumping the CPU PLL up to 1.9v do any good on a westmere OC. Like you know for sure this alone helped you stabilize. If you havem what blck and overclock where you trying to stabilize.

Its one of those learn as you go things, and i agree with Deimos as well as many others lower or Auto is better it seems for most...why my board/cpu likes it higher god only knows but i seem to be in the extreme minority in this case. A lot of us have asus boards but mine and Deimos (for example) are completely different in many ways and as such will act differently with different settings.I may have started this thread but i have learned so much from the more experienced/ educated who have contributed to this thread:)

As far as blck and multi settings....for me 191 blck and highest multi (22 for me) is best and is what i consider a mild /safe overclock for our aging boards but i see a lot of sigs with 4400 and i imagine they're pretty safe as well (as long as its stable and not using excessive voltage why not?:))
 
Last edited:
So the C states reduce power consumption significantly just by reducing multipliers even with the voltage locked?

My board is just about reverse when it comes to auto voltages. If u bump up the blck while reducing the multiplier, the auto voltage seems to follow the blck and not the overall clock speed determined by multiplier x blck. I once observed it pumping over 1.4v for a sub 3ghz CPU speed just because of the high blck,

T
I quickly learned not to trust or use the auto settings especially if your running 40% + overclocks. I manually set my dynamic voltage to +.1 and its locked it. Had to select all my power states to enable because leaving them in auto just caused my board to enable them in auto.

As for CPU PLL.
If there are people that do know for sure it has helped.... I am just interested in this. My limited experience has seen no real gains from raising it. Unless there is some trick to making use of it.

I have actually lowered it to 1.7v after seeing this thread. I have verified it to be stable but haven't done enough testing to see if its beneficial to my setup or not.

To tell you the truth, I think I am gonna take it pretty easy on my old x58 platform for now.
 
So the C states reduce power consumption significantly just by reducing multipliers even with the voltage locked?

Nope. C-states basically disables parts of the CPU. It has multiple levels and each level disables parts of the CPU and reduces voltage.

Speed-step, which is different and separate from C-States, drops the multiplier.

Also, and I've said this before, heat is a function of wattage, not voltage. As with any electrical circuit the demand for wattage is determined by the amount of work being done. More work means more current, more wattage and more heat. CPU's, being highly complex, have the ability to reduce voltage on the fly to lower wattage but the difference with this feature disabled is not particularly significant.

More info about C-states if you are interested:
Hardware Secrets Everything you need to know about C-states.

C-states have been around since the 486DX4 BTW
 
Back
Top