X79 and/or rampage IV OC'ing thread for those of us still left....

[dalekdukesboy]

Hi all just checking in....hot as hell here last few days and will cool off so not putting A/C in window so I have been on pc/gaming etc sparingly to keep myself somewhat comfortable.

However I will shortly be changing my build (slightly-moderately) to first off

1. improve airflow in case
2. get the damn side panel on case and look less redneck
3. add too/rearrange/clean out the rads...it's dusty here and dust in pc never helps.
4. update/upgrade ssd.
5. OC and test...what else?

Anyway we may have talked earlier about it but I know there is hack to get m2 pci-e drives to work on X79...ALSO which I'm more concerned with is there are a couple drives (Samsung 970-80 series?) I've heard actually work as boot drives on X79 with no bios hacking and I'd like to go that route obviously and want to confirm that.

 

[drizzler]

The latest NVME based PCIE-SSDs that boot out of the box from legacy systems without NVME Support (the stock X79 uefis lacking support) are the the Samsung 950 Pro and the Plextor M6e.

Newer ones are missing the inbuild option rom to boot so you will have to mod the nvme option rom to your mbs uefi (win-raid.com). For most of the x79 boards you are able to find already modded ones.

WIth this mod, every nvme ssd should work just fine and will be able to be used as boot device ( you will ofc need an pcie adapter to use them).

Another method is to use something like Clover or DUET but there is really no need for this with the option of nvme modding the uefi. (this is more common with x58 boards for example cause you can not mod the nvme support).

 

[dalekdukesboy]

The latest NVME based PCIE-SSDs that boot out of the box from legacy systems without NVME Support (the stock X79 uefis lacking support) are the the Samsung 950 Pro and the Plextor M6e.

Newer ones are missing the inbuild option rom too boot so you will have to mod the nvme option rom to your mbs uefi (win-raid.com). For most of the x79 boards you are able to find already modded ones.

WIth this mod, every nvme ssd should work just fine and will be able to be used as boot device ( you will ofc need an pcie adapter to use them).

Another method is to use something like Clover or DUET but there is really no need for this with the option of nvme modding the uefi. (this is more common with x58 boards for example cause you can not mod the nvme support).

I was pretty sure 950 pro was last one and good to know plextor m6e works too. I just have latest bios and even though I have an older WD black m2 already here I just am weary of bricking my bios (never done it but was close). I suppose I have multiple bios chips so I could always just make sure to keep one that is current away from being flashed as a backup.

 

[1000Gbps]

Guys, any idea how to properly test a couple of Xeons if a CPU core or cores are faulty, but on LOWEST possible frequency? Or should I test them at 100MHz steps to highest one allowed by Intel?

 

[drizzler]

Ohh sure I know that PCIe 2 is enough even for 3090 but in that case you could potentially start loosing few frames here&there in some games/benchmarks while with "weaker" GPU like the Fury there is literally no difference at all....no there are no errors at all(in between 104,3-104,9 the win will just froze on higher bclk it will start with downgrading PCIe and everything will work totally normal) I believe that actual problem is probably HIGH PCIe link speed as after certain point of BCLK/OC this motherboard simply can not handle that speed anymore so it will downgraded on gen 2 but as I said and show on pic above that "downgraded" PCIe Gen 2 speed is now when is OC actually closer to the PCIe Gen3 speed then to the normal/regular PCIe Gen2 speed....Not sure but maybe you are correct and some other GPU could handle more BCLK/OC tho' I still believe it's more due to the limitation of the X79 motherboard....
Yeah this board is great for OC have also option for strap BUT it's like 100,125,150 unfortunately those Xeons can not handle strap on 125 or more like some unlocked I7.......

Hey sorry bringing this topic up again but while digging around some other stuff i found a post of someone having the same problem like you and so i took a look in the hidden PCI System Sub-Menu of the Rampage IV BE that maybe has to be enabled wthin your bios (i have modded myself and i think most of the common mods floating arround have not enabled this but maybe i am wrong).

Within this menu you have settings for automated device bandwith handling and you can even force the speed. Despite it says PCIE Gen2 Settings it works for gen3, i think this is just a relict from the first days of x79 where it was lacking support of gen3. I was able to force my Titan X Maxwell to 2,5 GT/S or PCIe 1.1.

So the interesting settings for you should be hardware autonomous speed and width, if your gpu allows the disabled setting to work. But this should be the culprit since the bclk oc comes with pcie oc that maybe causes some minor link instability.

 

[lexluthermiester]

I was able to force my Titan X Maxwell to 2,5 GT/S or PCIe 1.1.

Do the same in the board UEFI. Change the "Target Link Speed" from "Auto" to 5GT/s, save & exit. Should force the card to PCIe2.0 speeds.

 

[drizzler]

Some other stuff i want to share (maybe this is well known but from reading this topic i can not remember it was clearly stated) and it left me with some questionmarks :

I switched my 64gb running at 2233 with 16gb 2666 / 2800 sticks and i was running them at 2666mz (125 strap needed ofc) and i got some performance improve but switching to 2800 results in a performance decrease.

I have done much X58 OC and such problems are also happening with X58 and you mostly overcome this by adjusting the skews and/or pll voltages (it is mostly an uncore or qpi problem, bclk and pcie are not tied together at x58) but i was not able to fix this problem by adjusting skews under cpu paradise and i was not able to find some dmi related pll voltages).

So i was searching for some other bandwith results and found one from 2014 that has nearly the same settings i use.

My results with 2666 strap 125/125 are :

The one i found :

So i just lowered the bclk to round about 122 like this dude has done and voila :

So what we see here is a huge gain in ram performance (minor loss in latency caused by lower freq) so sticking straight to the strap presets may give you worse performance. And this is not just an AIDA related thing, i see a 4gb/s gap with SysSandra too. And yes, the ram is running with the same settings and it is 100% stable.

With 2800 the things even get worse because i was not able to find a bclk giving better performace, mostly the known bug occurs in a massive drop of write performance.

So my explanation is that the strap concept is some kind of broken due the lack of having a clear seperation of clockgen for pcie and bclk which cause overlapping / interference and was just some marketing to reach higher ddr3 speeds. (sure you still can adjust but people launching xmp and thas is it will maybe never recognize).

And some other thing that is maybe just CPU-Z related :

First of all it stated about QPI, something that is non existend with x79 anymore, here we have DMI 2.0

But despite of this his linkspeed is 4875 while mine is "only" 3999 or 3911.

Readings from HWinfo :

CPU Current: 4523.1 MHz = 37 x 122.2 MHz @ 1.1259 V
CPU Bus Type: Intel QuickPath Interconnect (QPI) v1.1
Number of QPI Links per CPU: 2
Maximum Supported QPI Link Clock: 4800 MHz (9.60 GT/s)
Current QPI Link Clock: 4000 MHz (8.00 GT/s)
CPU Bus Type: Intel Direct Media Interface (DMI) v2.0
Maximum DMI Link Speed: 2.5 GT/s
Current DMI Link Speed: 5.0 GT/s
Number of Overclocking Bins: Unlimited

So we have some qpi reading on a single socket board with a single socket cpu, but with mine it is not the maximum possible like it is with the dudes screenshot. Something i do not understand.

Anybody a clue?

 

[basco]

comparing different aida versions can be a pain- just saying

 

[dalekdukesboy]

Guys, any idea how to properly test a couple of Xeons if a CPU core or cores are faulty, but on LOWEST possible frequency? Or should I test them at 100MHz steps to highest one allowed by Intel?

Sorry I saw your post...but I have NO idea what answer to this is. Well, OK not true I have an opinion but not sure it's based in solid fact lol. I would test them all at stock values NO overclock or underclock and see if they test out. Then if everything seems ok play around with overclocks you think are reasonable and see if you get errors and bluescreens, bad boots etc.

Hey sorry bringing this topic up again but while digging around some other stuff i found a post of someone having the same problem like you and so i took a look in the hidden PCI System Sub-Menu of the Rampage IV BE that maybe has to be enabled wthin your bios (i have modded myself and i think most of the common mods floating arround have not enabled this but maybe i am wrong).

Within this menu you have settings for automated device bandwith handling and you can even force the speed. Despite it says PCIE Gen2 Settings it works for gen3, i think this is just a relict from the first days of x79 where it was lacking support of gen3. I was able to force my Titan X Maxwell to 2,5 GT/S or PCIe 1.1.

So the interesting settings for you should be hardware autonomous speed and width, if your gpu allows the disabled setting to work. But this should be the culprit since the bclk oc comes with pcie oc that maybe causes some minor link instability.

View attachment 206460
View attachment 206461

Some other stuff i want to share (maybe this is well known but from reading this topic i can not remember it was clearly stated) and it left me with some questionmarks :

I switched my 64gb running at 2233 with 16gb 2666 / 2800 sticks and i was running them at 2666mz (125 strap needed ofc) and i got some performance improve but switching to 2800 results in a performance decrease.

I have done much X58 OC and such problems are also happening with X58 and you mostly overcome this by adjusting the skews and/or pll voltages (it is mostly an uncore or qpi problem, bclk and pcie are not tied together at x58) but i was not able to fix this problem by adjusting skews under cpu paradise and i was not able to find some dmi related pll voltages).

So i was searching for some other bandwith results and found one from 2014 that has nearly the same settings i use.

My results with 2666 strap 125/125 are :

View attachment 206462


The one i found :

View attachment 206466


So i just lowered the bclk to round about 122 like this dude has done and voila :


View attachment 206464


So what we see here is a huge gain in ram performance (minor loss in latency caused by lower freq) so sticking straight to the strap presets may give you worse performance. And this is not just an AIDA related thing, i see a 4gb/s gap with SysSandra too. And yes, the ram is running with the same settings and it is 100% stable.

With 2800 the things even get worse because i was not able to find a bclk giving better performace, mostly the known bug occurs in a massive drop of write performance.

So my explanation is that the strap concept is some kind of broken due the lack of having a clear seperation of clockgen for pcie and bclk which cause overlapping / interference and was just some marketing to reach higher ddr3 speeds. (sure you still can adjust but people launching xmp and thas is it will maybe never recognize).

And some other thing that is maybe just CPU-Z related :

First of all it stated about QPI, something that is non existend with x79 anymore, here we have DMI 2.0

But despite of this his linkspeed is 4875 while mine is "only" 3999 or 3911.

Readings from HWinfo :

CPU Current: 4523.1 MHz = 37 x 122.2 MHz @ 1.1259 V
CPU Bus Type: Intel QuickPath Interconnect (QPI) v1.1
Number of QPI Links per CPU: 2
Maximum Supported QPI Link Clock: 4800 MHz (9.60 GT/s)
Current QPI Link Clock: 4000 MHz (8.00 GT/s)
CPU Bus Type: Intel Direct Media Interface (DMI) v2.0
Maximum DMI Link Speed: 2.5 GT/s
Current DMI Link Speed: 5.0 GT/s
Number of Overclocking Bins: Unlimited

So we have some qpi reading on a single socket board with a single socket cpu, but with mine it is not the maximum possible like it is with the dudes screenshot. Something i do not understand.

Anybody a clue?

Don't be sorry for that first post!!! That was great! You are getting into all sorts of stuff I've forgotten and/or didn't even know about! I largely found and got my 1680 due to that guy (Mrs. Shorty) and a couple threads with him and others doing all sorts of tests with it. However I either missed or forgot he actually underclocked the base clock to get his memory bandwidth so high.

Also I particularly am grateful you showed the hwinfo menu ( I had it once on my pc but deleted it cause I wasn't using it etc) and that's very good information to have on forcing speeds of certain things and the bandwidth settings I didn't even know were accessible and there! Great stuff! This is exactly sort of stuff I want in this thread and per usual your posts are particularly pertinent with good solid info and things we can tweak/improve on with our setups.

Unfortunately I tried what you did and I get a mixed bag....the Read is great, Write blows and falls into that weird bug that happens with some settings, and the Copy....is ok but still not incredible either. What am I missing?

 

[Zyll Goliat]

Hey sorry bringing this topic up again but while digging around some other stuff i found a post of someone having the same problem like you and so i took a look in the hidden PCI System Sub-Menu of the Rampage IV BE that maybe has to be enabled wthin your bios (i have modded myself and i think most of the common mods floating arround have not enabled this but maybe i am wrong).

Within this menu you have settings for automated device bandwith handling and you can even force the speed. Despite it says PCIE Gen2 Settings it works for gen3, i think this is just a relict from the first days of x79 where it was lacking support of gen3. I was able to force my Titan X Maxwell to 2,5 GT/S or PCIe 1.1.

So the interesting settings for you should be hardware autonomous speed and width, if your gpu allows the disabled setting to work. But this should be the culprit since the bclk oc comes with pcie oc that maybe causes some minor link instability.

View attachment 206460
View attachment 206461

Cheers&THX for suggestion but already been there and I try that...just in case I did tried today again and it's the same thing....no change oh high BCLK OC it will work only on PCIe GEN 2 or GEN 1....but who cares everything else working just fine so I keep it on 115 BCLK......
BTW Why do you Forcing PCIe Gen 1.1 on your Titan Maxwell?You can loose a bit of performance on Gen 1.1 even on X16 link...

 

[drizzler]

Cheers&THX for suggestion but already been there and I try that...just in case I did tried today again and it's the same thing....no change oh high BCLK OC it will work only on PCIe GEN 2 or GEN 1....but who cares everything else working just fine so I keep it on 115 BCLK......
BTW Why do you Forcing PCIe Gen 1.1 on your Titan Maxwell?You can loose a bit of performance on Gen 1.1 even on X16 link...

I just wanted to showcase that the gen2 submenu settings are affecting gen3 cards / pcie lanes by forcing gen 1.1 in the menu. Could have done a better writing to prevent this misunderstanding, it was not intended to tell someone to force 1.1 for fixing any issues.

Sorry I saw your post...but I have NO idea what answer to this is. Well, OK not true I have an opinion but not sure it's based in solid fact lol. I would test them all at stock values NO overclock or underclock and see if they test out. Then if everything seems ok play around with overclocks you think are reasonable and see if you get errors and bluescreens, bad boots etc.




Don't be sorry for that first post!!! That was great! You are getting into all sorts of stuff I've forgotten and/or didn't even know about! I largely found and got my 1680 due to that guy (Mrs. Shorty) and a couple threads with him and others doing all sorts of tests with it. However I either missed or forgot he actually underclocked the base clock to get his memory bandwidth so high.

Also I particularly am grateful you showed the hwinfo menu ( I had it once on my pc but deleted it cause I wasn't using it etc) and that's very good information to have on forcing speeds of certain things and the bandwidth settings I didn't even know were accessible and there! Great stuff! This is exactly sort of stuff I want in this thread and per usual your posts are particularly pertinent with good solid info and things we can tweak/improve on with our setups.View attachment 206490

Unfortunately I tried what you did and I get a mixed bag....the Read is great, Write blows and falls into that weird bug that happens with some settings, and the Copy....is ok but still not incredible either. What am I missing?

hmm yeah you maybe have to lower the bclk further or maybe use 126 or anything else to get rid of the write bug with the 32:3 divider. 122.25 is the max i am able to set without getting the performance hit you showcased here.

I also use blck skew -1 and -10 / +10 within the cpu paradise menu but from my experience this has no impact.

And this is what me really bothers me because it looks like there is some issue with signal quality or signal reading with the 32:3 divider and higher bclk which is not fixable by any skew or pll setting, it is just luck to hit / find the right bclk.

To clarify : by 32:3 divider i mean the divider that sets 2133 ram at 100 strap and 2666 with 125 strap.

The next higher one giving 2400 with 100 strap is well known broken at showing the same performance hit in write speed showcased here.

I also had some runs with seeing this performance hit with 100 strap and 2133 ram divider for reasons i was not able to find or reconstruct.

 

[dalekdukesboy]

I just wanted to showcase that the gen2 submenu settings are affecting gen3 cards / pcie lanes by forcing gen 1.1 in the menu. Could have done a better writing to prevent this misunderstanding, it was not intended to tell someone to force 1.1 for fixing any issues.



hmm yeah you maybe have to lower the bclk further or maybe use 126 or anything else to get rid of the write bug with the 32:3 divider. 122.25 is the max i am able to set without getting the performance hit you showcased here.

I also use blck skew -1 and -10 / +10 within the cpu paradise menu but from my experience this has no impact.

And this is what me really bothers me because it looks like there is some issue with signal quality or signal reading with the 32:3 divider and higher bclk which is not fixable by any skew or pll setting, it is just luck to hit / find the right bclk.

To clarify : by 32:3 divider i mean the divider that sets 2133 ram at 100 strap and 2666 with 125 strap.

The next higher one giving 2400 with 100 strap is well known broken at showing the same performance hit in write speed showcased here.

I also had some runs with seeing this performance hit with 100 strap and 2133 ram divider for reasons i was not able to find or reconstruct.

Bingo! This is from yesterday before you even posted this I myself wondered if I just lowered the bclock to 121.7 vs exactly at 122 which is what I had in the first run with the meh results...and it worked!

Still not spectacular because the timings aren't dialed in perfectly and this is also high speed single sided ram...I have some nice microns I binned from earlier posts I may swap out if after tinkering with this stuff I can't get 70k across the board...I believe this is hynix memory but not sure.

 

[dalekdukesboy]

Bingo! This is from yesterday before you even posted this I myself wondered if I just lowered the bclock to 121.7 vs exactly at 122 which is what I had in the first run with the meh results...and it worked!


View attachment 206595

Still not spectacular because the timings aren't dialed in perfectly and this is also high speed single sided ram...I have some nice microns I binned from earlier posts I may swap out if after tinkering with this stuff I can't get 70k across the board...I believe this is hynix memory but not sure.

Not a huge difference but slight uptick just by changing the timings to CR1 from CR2. Definitely going to have to play around a bit and see what I can get out of this set of ram (at best) and put microns in just to see what they can do, haven't had them even in to test/play with for quite some time.

 

[dalekdukesboy]

Very disappointing news...not a big deal just a bummer! My Ballistix which are microns are causing the board to throw the 53 code which I looked up means "incompatible memory" . I'll have to go back and see if I ever used it in this board but it's possible I used them exclusively in the regular Rampage IV or they worked minimally on this board if that I'll have to check. Regardless, tried couple times reset bios etc same code, put the Klevv one sided 2800 mhz high speed back in booted fine first try. So, I'm revisiting the 2800 speed and seeing if I can get much out of that vs. lower speed and tighter timings. This is what it does at roughly 4.6 ghz.

Not bad...latency obviously good but bandwidth is not amazing will play with dividers and timings see what it'll do.

Went back to 2600 mhz and tightened from Cas 11 to Cas 10....it is about 1k better than my best results with CR1. So if it's at all stable I may with some more prodding get near 70k across the board.

 

[drizzler]

Just for terms cause you are writing about single sided : i think you mean single ranked, right?

Dimm can be single sided or dual sided and single ranked, dual ranked, quad ranked and octo ranked.

I had some klevvs too which where dual sided and single ranked @ 2800 which are most of the higher clocked ones. My dominators platinum are dual ranked @ 2800, it is sad that you can not tell by labels most of the time how they are organized. Only Samsung is doing this very well, at least with their server dedicated ram.

And you suffer from the same performance hit @ 2800 i experienced when i was running them @ 2800, i saw something like 66 66 66 with cl 12 14 14 35 which is way less performance compared to the 76 68 71 with 2600 11 13 13 28. The higher bclk needed with 125 strap is messing this up.


i am able to boot with 162.5 / 166 but only with cpu speed of 4 ghz, anything above results in cpu initialize failure or usb initializing failure.

The most pain in the ass about finding the right bclk that you can not adjust the blck on the fly to see if it affects this issue.

Starting with a bad 125 / 125 and adjust it on the fly due oc panel within windows to 122 / 125 will change nothing in terms of ram performance, i am only getting the better numbers with booting up 122.

 

[dalekdukesboy]

Just for terms cause you are writing about single sided : i think you mean single ranked, right?

Dimm can be single sided or dual sided and single ranked, dual ranked, quad ranked and octo ranked.

I had some klevvs too which where dual sided and single ranked @ 2800 which are most of the higher clocked ones. My dominators platinum are dual ranked @ 2800, it is sad that you can not tell by labels most of the time how they are organized. Only Samsung is doing this very well, at least with their server dedicated ram.

And you suffer from the same performance hit @ 2800 i experienced when i was running them @ 2800, i saw something like 66 66 66 with cl 12 14 14 35 which is way less performance compared to the 76 68 71 with 2600 11 13 13 28. The higher bclk needed with 125 strap is messing this up.


i am able to boot with 162.5 / 166 but only with cpu speed of 4 ghz, anything above results in cpu initialize failure or usb initializing failure.

The most pain in the ass about finding the right bclk that you can not adjust the blck on the fly to see if it affects this issue.

Starting with a bad 125 / 125 and adjust it on the fly due oc panel within windows to 122 / 125 will change nothing in terms of ram performance, i am only getting the better numbers with booting up 122.

Yes. I actually mean single ranked vs. single ranked, just been a while and honestly forgot the proper terminology. I believe microns were dual ranked and have tighter timings and overall better results at similar speeds but just can't achieve the high top speed. I never tried booting with 166 UNDER rated bclock but it always failed at 166 and I never bothered to keep trying just assumed it was something no cpu I ever had could/would do. Interesting about the on the fly results, not surprised though the board needs to restart and whatever magic happens when it resets all the internal timings associated with bclock is when you see the difference. Your numbers with 2800 mhz is almost identical to mine....I get roughly 65, 62, 60 something like that. Just meh, not even worth the extra mhz and very slight increase in latency.

Well try try and try again always good to check everything you can think of! I had 3rd or 4th newest bios not newest one for my board...no idea how that happened but I double checked and updated it to newest available. After that it started booting with the Microns just wasn't "seeing" all 4 sticks and it was not wanting to boot at all once I changed settings to higher frequencies...I went in and manually changed main timings...then it booted fine! So this memory isn't listed on compatible list for this board as I had said but I knew I had gotten it to work you just can't let auto settings work because it sets them totally wrong once you go above the 1600 mhz the Ram is sold as.

So very consistent across board not as high as I'd like at 2275 mhz but decent...I tried 2600 mhz and oddly it had the write bug where it had 2/3's of what it should be at! Had 74k for read and almost 70k for copy so that was odd. May just have to tinker with exact bclock and maybe 2600 will work as well as 2275.

 

[VeqIR]

Do Xeon E5 1680 v2 not work well with XMP profile enabled with quad channel RAM? I don’t know if I got a bum chip, but my RAM used to be stable in default XMP profile 1 (2133MHz) with an i7 4930k, and since swapping in the Xeon, I can’t run Memtest86 without WHEA errors when XMP is enabled. Turning it off, it runs completely error-free. Trying to overclock the RAM is not working manually either so far. Just running it at the base 1333MHz at this point.

P.S.: the motherboard is an ASUS RIVBE with the latest custom BIOS.

 

[dalekdukesboy]

Do Xeon E5 1680 v2 not work well with XMP profile enabled with quad channel RAM? I don’t know if I got a bum chip, but my RAM used to be stable in default XMP profile 1 (2133MHz) with an i7 4930k, and since swapping in the Xeon, I can’t run Memtest86 without WHEA errors when XMP is enabled. Turning it off, it runs completely error-free. Trying to overclock the RAM is not working manually either so far. Just running it at the base 1333MHz at this point.

P.S.: the motherboard is an ASUS RIVBE with the latest custom BIOS.

It may not be memory the board likes, it can be finicky about Ram and especially certain settings as I'm finding out. How many gigs of Ram is it? So you had same mobo but a non Xeon chip and it worked? I have had both in this board and if anything Xeon has been best at overclocking itself and has the best memory controller, which considering it was a couple grand new as a chip it should be essentially the best overall silicon they had at the time, was 2,500 I think new something like that. What brand memory and speed etc? I just had to play around with my memory to get it to work as you may have read.

Speaking of, I tightened timings and rebooted at 2600 and got among the best results I ever got. Excellent bandwidth across board and minimal latency! Stable? God only knows lol. Haven't tried any testing with new/old memory. I may be disappointed it can't handle it but I will try and at worst it's benchmark stable if nothing else.

 

[VeqIR]

32gb in 4 sticks. It’s G.Skill RAM that has been in use for a few years with XMP profile enabled using an i7 4930k. The only thing that changed was swapping the processor to the Xeon recently—one from an ebay listing, pulled out from a Mac (seller with great feedback). I gained 2 cores and lost all overclock so far =/. What voltages are you using? Seems like the XMP profile wants to raise the voltages higher than what my Xeon likes, and I haven’t been able to find a stable configuration with manual voltages. Changing voltages, running Memtest86, encountering errors, changing voltages again, etc. I’m running the processor at stock too, with lowest load line voltage compensation setting, so at least it’s running super cool. Sad, but I’ll probably just upgrade the the whole system in a year or two; lack of Windows 11 compatibility and poor/no overclocking might be the last drop in the bucket for that system.

 

[bobbybluz]

Once I get back up to speed and being somewhat fully functional again (spent some time in the hospital, got out Tuesday) I'm going to do my Cooler Master Cosmos II build with one of my Sabertooth X79/E5 1680 V2 setups in it. I have all of the parts now, just need the energy to finish cleaning up the case and assembling everything. It's nice to be back home on a real PC again instead of squinting into my Android phone.

 

[dalekdukesboy]

Once I get back up to speed and being somewhat fully functional again (spent some time in the hospital, got out Tuesday) I'm going to do my Cooler Master Cosmos II build with one of my Sabertooth X79/E5 1680 V2 setups in it. I have all of the parts now, just need the energy to finish cleaning up the case and assembling everything. It's nice to be back home on a real PC again instead of squinting into my Android phone.

Hope you're feeling better and on the mend chap! Look forward to hearing more of your exploits with new/old rig you've been a consistent poster here for which I'm grateful. Kinda cool I have one of the oldest and longest threads on here all about hardware that was "obsolete" even the 5-6 years ago I started it

 

[Duc996]

HI ... my name ist Thomas and I´m from Munich/Germany.

A few months ago I decided to rebuild my old X79 System. Everthing worked fine and my 3960x was running slightly overclocked with 4200mhz.
A few days ago I was lucky and grabbed a 1680v2 for only € 100,- on local Ebay. I immediately switched CPU´s and started playing around with it.
And now comes my problem ... is it normal that this CPU got a somehow strange Boost behavior? In CB or any other heavy load programm all is fine and it run´s with the set Clockspeed. But in any Game it just boosts to 3000mhz ... sometimes max clock while loading. In Firestrike for example just the Physics Test runs with max clock while all others remain at 3000mhz. Sure, I can disable C-States but that´s not the point. Or did I miss a BIOS setting? My System Specs are ...
E5 1680v2, Rampage IV Gene, 32Gb Dominator Platinum 2133, TitanX Pascal, 750W Seasonic ... running a custom water loop.

That´s it ... maybe someone with more experience could lend me a hand

Regards
Thomas

 

[agent_x007]

@Duc996 Change Windows power plan to "high performance".

 

[Duc996]

Wow ... thank you
Running now smooth like Babyoil. But now I wonder why the 3960x doesn´t need that setting ...
Anyway ... thanks alot

 

[Duc996]

After tweaking and testing I found the sweet spot of my 1680v2. It seems that I was lucky and got a good CPU ... it only needs 1.256 mv for 4,4ghz
The max temperature was 51 C while running Aida64 Stresstest for about one hour (thanks to my water loop)

All in all I´m pretty happy with that "old" CPU. In the next days I´ll try to find it´s max stable clock. But 4.4 with 1.256 mv will be my 24/7 setup anyway ...

Greetings
Thomas