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

[Aquinus]

And yes Broadwell e is for x99 but that's the tock, of the tick-tock cycle so even though I haven't seen evidence of new motherboard platform it seems it should be the rational next step by how they operate.

https://www.techpowerup.com/forums/...-processors-to-get-yet-another-socket.224271/

Older article though. By that time, I might be in a perfect position to do another HEDT build and maybe even put it under water. We'll see though.

 

[dalekdukesboy]

Yeah, no real details or dates etc there so that is pretty vague but thanks I obviously didn't see that. Like most that posted there unless the architecture or instruction set is a big improvement it almost looks like it will be same performance or even less than a highly overclocked x99...it even has less cache which never can be a good thing. So, who knows seems intel is really keeping us in the dark hoping they saturate us with x99 for as long as possible before giving away when the next socket is coming out. Also, I know they are having issues with die shrinks and even 14nm was delayed and I did see an article that 10nm is scheduled for 2nd half of 2017 but appears that is just a hope at this point and beyond that may be a long time before they can get good enough yields to get 7nm. So my instincts are that unless they can really invest in new tech they are pretty stuck as far as die size goes and there may be much smaller jumps from platform to platform from now on especially while AMD is sucking ass and giving them no competition to even worry about. So why should they push their architecture and spend on R & D when they can give us just small improvements and make hay until AMD gives them something to worry about but for now sun is shining.

 

[Random Murderer]

I'm in the process of catching up on this thread. I messaged @dalekdukesboy to see if he ever got an answer about the PCIE switches on the RIVE and figured others could benefit from this info, as well as giving me an easy "in" to the thread, lol. So here is what I had typed to him and decided to post here instead:

I'm glad there are still people loving the X79 platform, I built this system in late 2011/early 2012 and was hoping for a long lifespan, and it has provided just that. I started with a 3820 that ended up being a pretty good clocker, ran 4.9GHz its whole life and had an IMC that would run 2520 stably. Later, I swapped the 3820 for this 4930K and swapped my 4x 4GB sticks of Samsung Green "Wonder RAM" for the 4x 8GB of G.Skill I have now. I still have the 3820 along with the RIVE BIOS chip that still has the 4.9GHz OC on it, and I'm still using the Sammy Greens in two other systems.
Since building this PC, those switches have seen a lot of use for troubleshooting and benching.
So here's the low-down on those switches and LEDs beside the switches:
The RIVE has 4 PCI-E paths wired directly to the CPU, the top red slot is one and has 16 lanes, the second red slot and the black slot below it share a path and have 8 lanes, the third red slot is the third path and switches (automatically) between 16 lanes and 8 depending on whether the bottom slot is populated, and the bottom red slot is the fourth path and has 8 lanes. Each switch corresponds to the path in the same order, top switch for top path, bottom switch for bottom path, etc. The switches allow you to physically disable the pathway to the processor to, in effect, remove a PCIE device from the system without physically removing it. The lights next to the switches correspond to the same paths, if a light is on that means the path is enabled via switch and that path is populated.
Let me tell you, when you're trying to troubleshoot a system with three or more GPUs, or switching between benching multiple cards, those switches and lights are a godsend. At one point I was running three 7970s in Crossfire and a GTS250 for PhysX, and drivers can be super picky about installation order with a hybrid PhysX setup. The ability to switch off the GTS250 by itself or leave the 250 as the only connected card simply by shutting down and flipping a couple of switches was a great time saver. The switches also came in super handy when I was tinkering with GPU BIOSes, particularly while running Crossfire on the 7970s, as I could be 100% sure of the card I was flashing by shutting off all the other lanes. The feature was implemented by Asus in an effort to help facilitate sub-zero, multi-GPU benching, but those switches have far more uses than just extreme bench runs.
Keep in mind, the PCIE 1x slot is exempt from those switches, as the 1x slot does not have its PCIE lane provided by the CPU, rather it is routed to the PCH.

In all of my years overclocking, I have easily spent the most tinkering time on this board. If you have any questions regarding clocks, voltages, timings, etc., don't hesitate to tag me in a post and I'll answer to the best of my knowledge. Let us keep an active community around this platform like the X58 platform has seen, and share our knowledge and tricks to hopefully extend the lifespan of this platform even further.

 

[dalekdukesboy]

Hey, I just want my thread to live and not die and provide me and ultimately others as well to have a place to comment on a platform that is older yet good enough to do most of what anyone wants...but it's still old 2011-2012 so many people have dumped it or never bothered with it. Anyway, I am envious you have a 3820 that runs those crazy high clocks and can run memory over 2500...but at the same time? I admit that made me wonder, also what kind of cooling do you have to get 4.9 ghz and what kind of voltage? So yes I have a lot of questions for you...and currently my 1st pci-e slot is toast as well as a GPU so I'm simply happy I only need one card and I can still run 16 lanes on 3rd slot and everything else seems fine. Anyway I figured out the lights I think you're talking about when I had issues with the 1st lane and noticed they light up depending on which slot is populated and you have manual switches you can shut them off with which I played with and verified. So actually that may be helpful to others so thanks, but that I knew...however all the details about your 3820 and high memory ghz has me interested. My ES 3960 will boot above 4.6 ghz but stable that is highest I've gone and especially with high memory clocks is 4.6 or so and even if it can do better the juice required seemed useless considering even at that speed I hardly have any real use for it. However, I'm still interested in your results and the details of them because half of reason I have this setup is simply to benchmark and push the limits and just enjoy doing so...not a practical application.

 

[Random Murderer]

The 3820 would bench past 5GHz, but 4.9 was the 24/7 overclock. I recall the 24/7 clock being 38x 129 BCLK, RAM at 2407 (18.66 divider). Due to the 3820 not having an unlocked multiplier, I had to run the 39 multiplier and ~125 BCLK to get 4.9 and 2500+ on the RAM. At 38x 129, I recall having had to raise the PLL to 1.86ish, the VCCSA to just under 1.3 volts, and VCORE was at 1.45. Cooling started as a Xigmatek Aegir heatsink with two fans, that held the CPU at 4.6 with ease, later I upgraded to an H100i and that got the 3820 and the 4930K to the limit of where I was willing to push VCORE before hitting a temperature ceiling. I do recall needing a dedicated fan for the VRMs on the RIVE for anything above 4.8 on the 3820, that heatsink got toasty, lol.

EDIT: Found a picture of the system a few months after I had built it, right after it got its second 7970:

 

[dalekdukesboy]

Wow, that's pretty amazing. I got the pll up to 2.0 though maybe I can lower it but also that bclock is amazing my setup will do 2 different straps and with the 100 strap I can get it up to 104 or so which got me 2500 mhz on memory with 12 cas timings but multiplier was fairly low. However on the 133 strap you can't go up or down oddly it just won't do it...yet dead on it would but with less stability than the 100 strap. Also I don't have a new bios for the ES doesn't support anything beyond a certain bios so no idea how much an older bios helps or hurts me. Vrm has a fan that is what I had to do mostly for the high pll and vccsa voltage etc that's almost a given. 1.45 volts is amazingly low for 4.9 ghz that almost shouldn't be possible to boot nevermind be anything resembling stable. I know it's 4 cores versus 6 so that may help somewhat but other than for the extra cores why did you opt for the 4930k? Especially since I see you have it at 4.5 ghz as the everyday clock...how good is that with the memory? I'm also curious what settings you had and have memory at with 4930k and 3820k everyday as well as max stable.

 

[Random Murderer]

Wow, that's pretty amazing. I got the pll up to 2.0 though maybe I can lower it but also that bclock is amazing my setup will do 2 different straps and with the 100 strap I can get it up to 104 or so which got me 2500 mhz on memory with 12 cas timings but multiplier was fairly low. However on the 133 strap you can't go up or down oddly it just won't do it...yet dead on it would but with less stability than the 100 strap. Also I don't have a new bios for the ES doesn't support anything beyond a certain bios so no idea how much an older bios helps or hurts me. Vrm has a fan that is what I had to do mostly for the high pll and vccsa voltage etc that's almost a given. 1.45 volts is amazingly low for 4.9 ghz that almost shouldn't be possible to boot nevermind be anything resembling stable. I know it's 4 cores versus 6 so that may help somewhat but other than for the extra cores why did you opt for the 4930k? Especially since I see you have it at 4.5 ghz as the everyday clock...how good is that with the memory? I'm also curious what settings you had and have memory at with 4930k and 3820k everyday as well as max stable.

Max stable on the 3820 was a hair under 5GHz, I believe it was 4.96, but I wanted to keep the 24/7 VCORE at or under 1.45V so backed down to 4.9GHz. The 4930K is the same way, I've had it stable at 4.7GHz, but wanted to keep the 24/7 VCORE at or under 1.4V and 36x 125.125 BCLK is where that ended up. I would have used 45x 100 BCLK, but the 24.00 divider for RAM gives oddly bad bandwidth and latency on the RAM. I chose to upgrade (side-grade?) to a 4930K for a few reasons, first and foremost being more cores which has helped with school (CAD and the like, heavy multi-tasking, audio processing, etc.), the second being a stronger IMC to play with. I haven't fully flexed the IMC's muscles just yet, but it does POST above 2600, so I know it's a stronger IMC than the 3820 had. Once I have some RAM I know will run 2600/2666 stable, I'll toss them in and see what this IMC can really do. Plus, overclocking IVB-E is a cakewalk compared to SB-E.

Also, you have a 1.33x strap? I was sure the straps were 1.00x, 1.25x, 1.66x, and 2.50x. I use the 1.25x strap most of the time, it performs better with RAM ~2400 for the reason I mentioned earlier. On the 1.00x strap, don't expect much above 104 BCLK. The same can be said for the other straps, at the 1.25x strap, don't expect a BCLK much higher than 104 x 1.25 = 130 BCLK. SATA and PCIE start getting a little wonky at BCLKs higher than that, so even if the CPU is stable, you risk data loss/corruption.

Somewhere in my OC notebook for the 3820, I have a couple of conversations I had with Shamino, a pro-overclocker turned ROG engineer. He passed along some really helpful tips regarding secondary voltages, as well as small tweaks to open up your OC headroom a bit more on the RIVE specifically. As soon as I find that notebook (I think I know where it is, but haven't seen it since I moved in late 2014), I'll be sure to post up what's relevant.

 

[dalekdukesboy]

it's probably and it IS 1.25....I was guessing off hand without numbers in front of me so don't get excited Yes that 104 was basically my limit and even anywhere in that area I did get data corruption/bluescreens and just other issues I knew was from the bclk overclock...So...you confirm +4 is MAX you can push bclock...but then that means you were well over +4 to get 2525, unless you had like 104.25 or something. But in reality that's pretty close to where I was stable at with memory I was just a hair over 2500 so I guess same difference there. Also biggest question beyond what I already asked is what program or tests did you use to ascertain it was stable?

 

[Random Murderer]

it's probably and it IS 1.25....I was guessing off hand without numbers in front of me so don't get excited Yes that 104 was basically my limit and even anywhere in that area I did get data corruption/bluescreens and just other issues I knew was from the bclk overclock...So...you confirm +4 is MAX you can push bclock...but then that means you were well over +4 to get 2525, unless you had like 104.25 or something. But in reality that's pretty close to where I was stable at with memory I was just a hair over 2500 so I guess same difference there. Also biggest question beyond what I already asked is what program or tests did you use to ascertain it was stable?

Actually, I had to go the other way to get 2500+ speeds, I used the 1.25x strap and lowered the BCLK below 125 while using the 21.33 RAM divider.
I forgot to answer your question regarding the 4.5GHz 4930K with RAM, it churns out great bandwidth with low latencies, RAM is at 2335 10-12-12-31 1T.
As far as stability testing goes, I used to be a hardcore P95 advocate, but since AVX was implemented in P95 it has just been too harsh on overclocked CPUs. My usual stability test on X79 consists of 50 runs of IBT at standard stress level, and if that passes then I'll run HCI Memtest set to 1500MB(the max allowed in the free version) per iteration, one iteration per thread (so 12 windows of HCI Memtest in my case) until all iterations hit 200% test completion. It has worked very well for me so far.

 

[dalekdukesboy]

I'm also surprised that a simple AIO was sufficient to keep your chip stable at 4.9 ghz even with a relatively low 1.45 volts that's still a lot of heat if you use the cpu and memory at all...I have a very slightly better AIO ( if I remember the model numbers correctly) h110i extreme which is either top model or 2nd, I only don't know thanks to their changing their numbering scheme over last few years and old vs new sku's mixed together which has confused noobs like me who never followed AIO coolers much till I decided I was going to buy one.

 

[Random Murderer]

I'm also surprised that a simple AIO was sufficient to keep your chip stable at 4.9 ghz even with a relatively low 1.45 volts that's still a lot of heat if you use the cpu and memory at all...I have a very slightly better AIO ( if I remember the model numbers correctly) h110i extreme which is either top model or 2nd, I only don't know thanks to their changing their numbering scheme over last few years and old vs new sku's mixed together which has confused noobs like me who never followed AIO coolers much till I decided I was going to buy one.

It was very close to hitting the thermal point where I would normally back off, but it did manage. The H110i is a 2x 140mm radiator, the H100i is 2x 120mm.

 

[dalekdukesboy]

Actually, I had to go the other way to get 2500+ speeds, I used the 1.25x strap and lowered the BCLK below 125 while using the 21.33 RAM divider.
I forgot to answer your question regarding the 4.5GHz 4930K with RAM, it churns out great bandwidth with low latencies, RAM is at 2335 10-12-12-31 1T.



As far as stability testing goes, I used to be a hardcore P95 advocate, but since AVX was implemented in P95 it has just been too harsh on overclocked CPUs. My usual stability test on X79 consists of 50 runs of IBT at standard stress level, and if that passes then I'll run HCI Memtest set to 1500MB(the max allowed in the free version) per iteration, one iteration per thread (so 12 windows of HCI Memtest in my case) until all iterations hit 200% test completion. It has worked very well for me so far.

Interesting...standard on IBT I can pass just about anything, however if I do MAX memory it takes the hands of God to get more than a run or two to pass. I was operating as in my earlier posts here with 9-10 gigs of 16 gigs being used in burntest which I realize is a big ass strain on the platform. I actually found hci memtest very handy to fine tune my memory and it failed quickly so I knew if I was on wrong track with my settings. If I did my quick math correctly you test 1/2 your memory but just run test 200% to make up for that.

 

[Random Murderer]

My test is 12x 1500MB, so 18,000MB of RAM. If I had less ram, I would evenly divide it between each thread, but 18,000MB is the maximum I can test on the free version without having more threads. When I ran 16GB RAM, I used to run each thread at 1200MB for a total of 14,400MB. At least I know it's a very rigorous strain on the RAM and IMC, so I test CPU stability with IBT and RAM/IMC stability with HCI, and if the system passes both then I consider it stable. This method has not let me down yet, at least on X79.

 

[dalekdukesboy]

Makes sense, I used the MCI memtest similarly but I only have 16 gigs so I used 80-90% of memory with it and it definitely helped me figure out what was what on stability and what was simply not possible to get stable. However IBT with standard is 1 gig of memory I think...so I admit even 50 runs of that especially with 16 or 32 gigs of ram is a bit minimal on stressing the system, that's just my thought though, since if you stress under IBT you stress the cpu plus whatever amount of memory you choose.

 

[FreedomEclipse]

I had a play around to see if i could bring those USB 3.0 ports on the back of my board back to life... Nope. I think they are pretty much dead. Only thing i can really do now is probably use one of my PCI-ex slots for a USB 3.0 expansion card. to fill in the gaps. This motherboard is one foot in the grave Ive had a look at X79 boards on ebay and even for second hand ones, they arent really worth the asking price.

 

[Random Murderer]

I had a play around to see if i could bring those USB 3.0 ports on the back of my board back to life... Nope. I think they are pretty much dead. Only thing i can really do now is probably use one of my PCI-ex slots for a USB 3.0 expansion card. to fill in the gaps. This motherboard is one foot in the grave Ive had a look at X79 boards on ebay and even for second hand ones, they arent really worth the asking price.

What have you tried as far as fixing them? I'm willing to bet the fuses are blown, it's a simple fix if you can solder SMD.

 

[FreedomEclipse]

What have you tried as far as fixing them? I'm willing to bet the fuses are blown, it's a simple fix if you can solder SMD.

Repositioning the motherboard just to make sure it hadnt been earthed as last week I installed an Antec Spot cool 100 to cool the my VRMS/Mosfets and thats about it.

Im not sure if anythings blown as the issue only effects the rear USB 3 ports but the ports at the front of my case still work perfectly fine. So the chipset is still very much alive. Do you have a video or guide on where and how to solder on new fuses?

 

[dalekdukesboy]

I had a play around to see if i could bring those USB 3.0 ports on the back of my board back to life... Nope. I think they are pretty much dead. Only thing i can really do now is probably use one of my PCI-ex slots for a USB 3.0 expansion card. to fill in the gaps. This motherboard is one foot in the grave Ive had a look at X79 boards on ebay and even for second hand ones, they arent really worth the asking price.

No, the sign that the platform is indeed very potent in performance despite age is exactly this, the mobo's and even the higher end cpu's still command hundreds of dollars apiece despite being 2 gens old and 4 years old which is relatively ancient in tech world time! It is amazing, plus shows that intel hasn't really advanced all that much since x79 and thanks to AMD spanking their monkey with cheap but relatively crappy processors (relative to intel) and intel being complacent no one has been pushed for much innovation, just die shrinks and updates of usb speed/ports etc.

What have you tried as far as fixing them? I'm willing to bet the fuses are blown, it's a simple fix if you can solder SMD.

Yeah, maybe but soldering even if you somewhat know what you are doing is one shaky hand or even just a mistake of soldering something unintentionally bricks your board. But yeah if you know what you're doing with the solder and know your board well enough to know what it is your soldering it's worth a shot, but still no guarantees there isn't more extensive damage causing issues.

Also Random Murderer I accidentally had one of my replies in with a quote of yours so not sure you saw it...you told me your RAM speed etc and I said I had basically the same exact results except I got it atm at 2400 mhz memory. But mainly I wanted to go back to how you test for stability, I'm just wondering your thoughts of using 1 gig of memory on IBT when you've got 16-32 gigs of memory. Do you (or anyone here) know/think of the validity of using more and most of your memory or even all of it? I know with high OC's much easier to pass lower gig of memory but just wondering what you guys all think of that.

And I'd add to your issues of USB ports but I am very much a NOOB and have no ability to give you good advice on soldering etc on the physical board itself. However if some of your ports work...that's not so bad particularly if enough work so you aren't hampered for using your mobo. the ONLY reason I'd worry is if it was a sign that the board may be on its' way out due to multiple things possibly failing due to age and electrical wear and tear particularly being overclocked. However if everything else seems fine, I personally wouldn't worry too much and assume it's just fuses like RM suggested or issues on board near those ports but doesn't affect everything else. Just my common sense and practical experience opinion working with many boards but admitting soldering/repairing and understanding the layout of boards is something I never touched/learned.

 

[Random Murderer]

Im not sure if anythings blown as the issue only effects the rear USB 3 ports but the ports at the front of my case still work perfectly fine. So the chipset is still very much alive. Do you have a video or guide on where and how to solder on new fuses?

I don't have a guide, but you're looking for polyfuses near the USB ports that don't work. You'll need to test each fuse for continuity, if it does not have continuity then the fuse has blown. I have attached some pictures of polyfuses, they come in different colors and markings but are all pretty generally the same shape. The fuses are pretty small, this is not something I would recommend if you have shaky hands, aren't good at de-soldering, don't have a good bit of experience soldering, and/or have bad eyesight. You could always bring it to an electronics shop near you and see if they'd be willing to replace the fuses for you.

This image shows them installed on a Raspberry Pi, shorting them is a pretty common mod on earlier Pi boards. The black piece in between the fuses is a capacitor, and the silver structure on the left of the picture is a dual USB plug, so you have some sort of size comparison.

Also Random Murderer I accidentally had one of my replies in with a quote of yours so not sure you saw it...you told me your RAM speed etc and I said I had basically the same exact results except I got it atm at 2400 mhz memory. But mainly I wanted to go back to how you test for stability, I'm just wondering your thoughts of using 1 gig of memory on IBT when you've got 16-32 gigs of memory. Do you (or anyone here) know/think of the validity of using more and most of your memory or even all of it? I know with high OC's much easier to pass lower gig of memory but just wondering what you guys all think of that.

Use as much as you can. I never said I don't give it a test at maximum available memory, just that my go-to test is usually 50 runs at Standard then HCI.
This G.Skill will run the timings I have it set to to almost 2500MHz, it's just that my 24/7 OC, at least as of right now, has the RAM running at a hair above 2333. I haven't had much time to tinker with my own PC in a few months...

 

[dalekdukesboy]

Use as much as you can. I never said I don't give it a test at maximum available memory, just that my go-to test is usually 50 runs at Standard then HCI.
This G.Skill will run the timings I have it set to to almost 2500MHz, it's just that my 24/7 OC, at least as of right now, has the RAM running at a hair above 2333. I haven't had much time to tinker with my own PC in a few months...

Ok thanks! I wasn't trying to imply right/wrong...I was actually really asking you because including your x to the nth power knowledge beyond mine on the fuses on boards and blackberries etc (very cool btw thanks for info!) but I also was bringing it up because I have struggled with what the F@#K do I call "stable"? I honestly do not know if it's simply ludicrous to be testing 10-13 gigs of memory plus stressing cpu at 100% which takes forever simply to do a run or two but also I can pass many runs at 1 gig for example with settings I know are barely bootable. And I can have settings that pass a quick 4 gig pass but fail at 9-13 gigs no matter what...Honestly it's me wondering why this platform seems so inconsistent and sometimes ok it's on me I miss a setting or two!, but sometimes I change ZERO and I go from stable at 10 runs with 90% of my memory used to barely passing anything 1mb or more. I'm sure if you go back near beginning of this thread you will see a few instances of my frustration right after posting a wonderful screenie of a 9 gig 10 pass run of IBT passed with flying colors only to see within a post or two me saying it suddenly can't pass jack squat and I don't remember changing a damn thing.
Anyway, I never got my RAM stable at CAS 10 at 2500 mhz had to ease timings to CAS 12 so that's pretty impressive! Then again I think that was with your Ivy bridge vs my sandy bridge which you clearly are showing the nice advantage of having. However that said I can't for the life of me figure out why this platform seems so flakey. When I have time and am ready to not get too annoyed maybe I'll revisit the IBT tests and maybe there's some variable or something on the bleeding edge of stability in my settings that I am missing...about all I can come up with short of a component that's faulty but if that were the case usually something stops working or gets worse fairly quickly with a highly clocked system.

 

[Random Murderer]

Ok thanks! I wasn't trying to imply right/wrong...I was actually really asking you because including your x to the nth power knowledge beyond mine on the fuses on boards and blackberries etc (very cool btw thanks for info!) but I also was bringing it up because I have struggled with what the F@#K do I call "stable"? I honestly do not know if it's simply ludicrous to be testing 10-13 gigs of memory plus stressing cpu at 100% which takes forever simply to do a run or two but also I can pass many runs at 1 gig for example with settings I know are barely bootable. And I can have settings that pass a quick 4 gig pass but fail at 9-13 gigs no matter what...Honestly it's me wondering why this platform seems so inconsistent and sometimes ok it's on me I miss a setting or two!, but sometimes I change ZERO and I go from stable at 10 runs with 90% of my memory used to barely passing anything 1mb or more. I'm sure if you go back near beginning of this thread you will see a few instances of my frustration right after posting a wonderful screenie of a 9 gig 10 pass run of IBT passed with flying colors only to see within a post or two me saying it suddenly can't pass jack squat and I don't remember changing a damn thing.
Anyway, I never got my RAM stable at CAS 10 at 2500 mhz had to ease timings to CAS 12 so that's pretty impressive! Then again I think that was with your Ivy bridge vs my sandy bridge which you clearly are showing the nice advantage of having. However that said I can't for the life of me figure out why this platform seems so flakey. When I have time and am ready to not get too annoyed maybe I'll revisit the IBT tests and maybe there's some variable or something on the bleeding edge of stability in my settings that I am missing...about all I can come up with short of a component that's faulty but if that were the case usually something stops working or gets worse fairly quickly with a highly clocked system.

I found SB-E to be extremely picky about certain voltages, VCCSA and PLL specifically. You could play with these voltages and raise them to help with stability, but if you didn't keep them within certain deltas in relation to specific other voltages you could have stability issues, or if you raised them too high you could have stability issues. A lot of my time spent clocking the 3820 was time spent finding the perfect balance of these voltages, a kind of Goldilocks effect. My PLL was raised to 1.86 (1.8625V IIRC) because if it was raised beyond 1.87 I was unstable, and if it was below 1.85, I was unstable. The instabilities caused by playing with these voltages and having them set wrong was also a bit of a crapshoot, as the instabilities themselves weren't always immediately apparent. This is why I kept a notebook with all the settings changed between each stress test, how long the test ran for before finding instability, how the system reacted to the instability (failed test? BSOD and what error code for the BSOD? etc.), any anomalies I noticed, etc. It was a very arduous process to dial in 4.9GHz and 2400 on the RAM stably, it took me weeks of coming home from work and immediately starting tinkering with the PC to get it stable because of these finicky voltages and my stubbornness in not wanting to raise VCORE higher than 1.45V.

A couple of things I remember off the top of my head regarding SB-E overclocking that may help you:

Memory related: CPU Clock Gen Filter should be set to 20uF if you're trying for DDR3-2400 or higher speeds. In the DRAM timing page you should have Rampage Tweak set to Mode 2 for overclocking purposes (Mode 1 is for enhanced RAM compatibility, Mode 3 is for IVB-E). Disable "Enhanced Training" for all RAM channels in the DRAM Timing page near the bottom. Secondary timing tRFC(aka DRAM REF Cycle) has a rather large impact on SB-E memory stability anywhere above 2133MHz. A tRFC of 147 worked best for my 4x 4GB setup at 2400ish. A huge bane of trying to get RAM stable above 2200 or so: RC Warm Boot Fast. Once you have found your RAM is stable, IMMEDIATELY reboot into BIOS and enable RC Warm Boot Fast.

Voltages: PLL can be raised a bit to help facilitate high BCLKs or high RAM clocks, but don't exceed 1.9V, and know that small changes can have huge impacts on stability. VTT helps with overclocking, Secondary VTT does not, but they must be within .3V of each other, so if you have raised VTT and it is more than .3V higher than Secondary VTT, you must raise Secondary VTT to stay within .3V. VCCSA is tricky, it must be kept below VCORE, but no more than .3V below VCORE, and no more than .6V below VDIMM, i.e. if your VCORE is set to 1.4V and your VDIMM is set to 1.75V, VCCSA must be between 1.15V and 1.4V. Also in regard to VCCSA adjustment, a little change goes a long way, test for stability after each change and note whether stability improved or worsened.

Other than that, the usual stuff applies: disable all spread spectrums, play with clock skews, and don't use more voltage than your cooling can handle, this goes for the VRM heatsink as well. While stress testing, touch the VRM heatsink and make sure it's not too hot, I found on the 3820 that starting around 4.75GHz it would get rather hot and I had to put a 60mm fan on it for stability reasons.

 

[dalekdukesboy]

Ok, you saying "weeks" of tinkering and the fact you need voltages literally spot on and not too low or too high confirms my thought...x79 is ridiculously picky...you literally need to give it EXACTLY the right voltage where most platforms you can be slightly on high side and you're fine, if you're a little low it may work but eventually under stress it fails or gives errors programs crash etc. But admittedly you said exactly what I am finding that it is extremely bitchy in my words and if you're off, it just does what it wants...cry and fail under even slight stress. Anyhow, most of what you said I knew or did to some extent due to my own weeks and months tinkering but all good info and a couple things I hadn't considered or honed in on enough so thank you! Also I have set my pll over 1.9 and I've heard varying answers to that including up to slightly over 2.0 is actually ok, and I've heard otherwise. Clock skus I have on auto for everytime I manually touched them things got worse so just too much to figure out what is right vs. wrong, vrm heatsink I got about 170 cfm fan right over it so I got that way overcovered lol. Anyway thanks for the insight and some of it validates everything I've done, and a few things are a bit different so that gives me some new avenues to look at I hadn't really thought of or simply didn't know the full importance of.

 

[Random Murderer]

Ok, you saying "weeks" of tinkering and the fact you need voltages literally spot on and not too low or too high confirms my thought...x79 is ridiculously picky...you literally need to give it EXACTLY the right voltage where most platforms you can be slightly on high side and you're fine, if you're a little low it may work but eventually under stress it fails or gives errors programs crash etc. But admittedly you said exactly what I am finding that it is extremely bitchy in my words and if you're off, it just does what it wants...cry and fail under even slight stress. Anyhow, most of what you said I knew or did to some extent due to my own weeks and months tinkering but all good info and a couple things I hadn't considered or honed in on enough so thank you! Also I have set my pll over 1.9 and I've heard varying answers to that including up to slightly over 2.0 is actually ok, and I've heard otherwise. Clock skus I have on auto for everytime I manually touched them things got worse so just too much to figure out what is right vs. wrong, vrm heatsink I got about 170 cfm fan right over it so I got that way overcovered lol. Anyway thanks for the insight and some of it validates everything I've done, and a few things are a bit different so that gives me some new avenues to look at I hadn't really thought of or simply didn't know the full importance of.

I wouldn't say that X79 is inherently picky, but I would say that SB-E is, at least from my experience. I couldn't believe how much easier it was to overclock IVB-E, even going from a 4c/8t SB-E to a 6c/12t IVB-E. The only voltages I had to play with to get this 4930K stable were VCORE, VDIMM, and a slight increase on PLL to 1.8125V. Maybe part of the reason I look back so fondly on clocking the 3820 was simply because it was a challenge and required a good bit of learning and tinkering... this 4930K just felt too easy after having played with the 3820.

 

[dalekdukesboy]

Ok, well I say x79 as = SB-E cause that's all I got! lol. So maybe fine point (and you're probably right) but it's what I've got so to me it's everything unless I pick up a cheap ivy bridge to chuck in and give it new life...for about 600$ I can get best memory and literally best x99 mobo out there arguably...the PROBLEM is if I want 8-10 core processor not just another 6 core which won't be that much of an upgrade from present would set me back 1000-1600 bucks depending on which on you get! The x99 processors are without any competition and for the moment are priced as such...so even if I get everything but the cpu I'm worried by the time the processor prices are better used or new I may have a platform that AMD or INTEL replaced with something much cheaper and comparable or even better. Yeah, well easy is welcome when ALL you have is SB-E and it is as accommadating as a divorcing wife who is broke and you are a billionaire. Admittedly the successes are nicer in a sense and I had them as you see on this thread but yeah SB-E works well when it works, it's just getting it there overclocked is a process. Also ivy is a die shrink so 1.9 volts on pll is probably limit but sb-e you can definitely go up to 2.0 and bit beyond and even higher with extreme cooling. Also with that in mind I have an ES so I have older bios so not sure how much that hurts me because the ES is not supported in later bioses and you just get 00 and no boot when you try...I experienced that early on and looked in threads and found the answer to that mysterious problem for me.

 

[Random Murderer]

Ok, well I say x79 as = SB-E cause that's all I got! lol. So maybe fine point (and you're probably right) but it's what I've got so to me it's everything unless I pick up a cheap ivy bridge to chuck in and give it new life...for about 600$ I can get best memory and literally best x99 mobo out there arguably...the PROBLEM is if I want 8-10 core processor not just another 6 core which won't be that much of an upgrade from present would set me back 1000-1600 bucks depending on which on you get! The x99 processors are without any competition and for the moment are priced as such...so even if I get everything but the cpu I'm worried by the time the processor prices are better used or new I may have a platform that AMD or INTEL replaced with something much cheaper and comparable or even better. Yeah, well easy is welcome when ALL you have is SB-E and it is as accommadating as a divorcing wife who is broke and you are a billionaire. Admittedly the successes are nicer in a sense and I had them as you see on this thread but yeah SB-E works well when it works, it's just getting it there overclocked is a process. Also ivy is a die shrink so 1.9 volts on pll is probably limit but sb-e you can definitely go up to 2.0 and bit beyond and even higher with extreme cooling. Also with that in mind I have an ES so I have older bios so not sure how much that hurts me because the ES is not supported in later bioses and you just get 00 and no boot when you try...I experienced that early on and looked in threads and found the answer to that mysterious problem for me.

Check out the Xeon E5 1650 v2. It's 6c/12t, IVB-E, and overclockable. You may be able to pick one up cheaper than a 4930K.