TR Forums

I'm just posting this because I have always enjoyed looking at build log pictures. So I'll do a little giving back.

This is an ASUS X99 Sabertooth i7 5820K build which will have 64GB of DDR4 and an EVGA 980Ti Classified going into it. It's definitely not a budget build. But like all of my builds I tend to take a money saving approach to things. Best example is probably that I got 2x dual-channel 32GB kits just to save $50. Because you pay a premium for 64GB quad-channel kits.

The other part which is really sad is that the GPU isn't coming to me in the mail until later next week. I will be putting it through its paces on memtest86 using one of many (extremely) anemic graphics cards I have on hand once I get all the innards wired up. I do have my GTX 670 I can slap in there for a spin. But why even bother...

So this is only a build log in progress. So far.

http://imgur.com/a/sys28/layout/horizontal

Thanks for the photos, I expect some attempt at overclocking considering the heatsink.

I've always thought the plastic cover on those ASUS boards looked a bit odd - did you leave it on?

localhostrulez wrote:

I've always thought the plastic cover on those ASUS boards looked a bit odd - did you leave it on?

The plastic cover is suppose to help the thermals of the components, and prevent dust infiltration.

Sure, the 5 year warranty justifies the Sabertooth. As for the plastic armor, I will consider that term legitimate. Think about the risk of dropping the screwdriver or some expensive component. It literally helps to prevent things from going bad. Also the backplate on the board is very nice. The board is also insanely heavy. It might weigh 4 or 5 pounds.


I have fully assembled it now and it took all day and all night, the second half of which I did not spend finagling the camera on my atrocious $10 tripod.

I reckon I'll go take a finished photo later on when I finally get the 980Ti installed. Meanwhile the moment of truth awaits me. I also cannot decide what to do first (assuming everything starts up okay), OC or run Memtest.

Well, my first OC attempt I guess I messed some setting up, I tried to change too many things at once. There's too many things to tweak with Haswell-E! I think it was my cache ratio setting that made my first attempt not even POST...

I figured that since I care about power consumption I only care about what clocks I can achieve while leaving EIST and C-states and turbo enabled. (Edit: Hahahahaha, nope, this turned out to be way too difficult!!)

Now I just have it at +0.1V offset and it is Prime95'ing 12 threads stably so far, running 4.0Ghz, hovering around 1.1V, drawing 260W out of the wall. Will ignore cache ratio/voltage until later...

I suppose to save some time I should crank to 1.25V, set 44 multiplier, and see if that's stable, if not 43 multiplier, then start lowering the voltage. But also I'd like to switch the voltage adjustment to Adaptive instead of Offset since I dont want +.1 or +.2V when the machine is idling... This way when the machine is idle, it is running with ~0.72V at 1.2Ghz and under 80W out of the wall. Hopefully once I get my 980Ti it will become even more power-efficient when idle.

And switching my +0.1 from offset to adaptive causes P95 to immediately freeze and BSOD. plot is thickening. In fact, I tried +0.2 with adaptive and P95 results in the same insta-freeze.

Maybe I really will have to maintain separate BIOS profiles, one for full performance (running workloads, playing games: 4.3+Ghz, all the time), and one for cool & quiet computing (browsing, development, mild or no OC). I was hoping that I could get something that could do both! The other option is maybe I'll start setting my machine to sleep when I go to sleep. Laaaaaaame. What'll end up happening is I'll run the OC 24/7. Will just make it a somewhat conservative one.

Another update.. now I can see my Kill-a-watt is telling me idle at 1.15V at 4.1Ghz is 120W out of the wall. At some point I guess I could just throw up my hands and say that the 50% additional power consumption could be worth the buttery smooth stability and awesomeness that is perpetual 4 gigahertz. Besides, it's got to throttle down by a factor of *well* more than 50% (down to 1.2ghz) ordinarily.

Interestingly 1.25V at 4.3Ghz is exactly the same kind of thermal behavior as before when idle... 121W, though this goes up to 305W with P95 going. Still stays quite cool and quiet at idle while still hauling ass at 4300Mhz (and the same voltage)... I'll try 4.4 at this voltage next after backing up this BIOS config to USB. And once the first Prime95 test passed it's on to the 8K FFT's and we're at 360W from the wall now. Damn this thing is sweet.

The FMA3 instructions P95 throws at this CPU really make it put out a lot of heat. It's more than I'm comfortable with and creeps past 90C on most of the cores eventually, after a few minutes. The fans are all set to max by the mobo at this point and it's pretty clear I need to put the CPU under water if I intend to run this type of torture-test computational load where all the data is available in cache and it cranks till it potentially self destructs...

I'm also testing IBT (IntelBurnTest) but as i understand it this was written/updated in 2012 so it likely has not been coded or even re-compiled to use the FMA instructions so it tops out at 325W at the wall.

The truth is that most computational loads, the sort that I'll use, are not and will never reach this level of torture level and I'm content with it being limited in a thermal capacity sense, enough not to be running out and buying water loop equipment tonight anyhow.

Now if only that darned 980Ti would just get to my door I could play some games, which I know with certainty I can do at least at 4.3Ghz.

For my 4790K have it at 4.5 GHz for the Cache, and IA cores with adaptive voltage adjustment but I went the other way -0.063 for the IA cores. Keeps the voltage between 1.173-1.177 depending on the voltage assigned to each core unless AVX2/FMA3 come into play then the voltage gets bumped to 1.22/.25. Still haven't messed with the cache voltage, I tried to but didn't get very far. Have all power managements settings turned on, and the cache set to downclock on idle.

I went with the negative offset to keep heat, and power consumption in check. The hotter the CPU/GPU get results in increased power consumption leading to more heat. Don't need anymore heat considering Haswell, Haswell-E, Broadwell, and Broadwell-E have a integrated VRM on die already adding more heat.

Yeah I don't know, all I can say is I do not have an entire week, nor even the two days I already put in, to spend tweaking my hardware, regardless of how sweet it may happen to be... I tried a few settings that (to me) clearly should have worked, but any time that I used Adaptive voltage, Prime95 made it lock up instantly.

That said, my chip is now happily skipping along at 4500Mhz with 1.25V which I think means that it is a decent to very good chip based on my recent armchair travels...

I do know that at these speeds, feeding it too many FMA/AVX type instructions too fast may cause me to run out of thermal capacity with this heatsink, the most massive air heatsink ever designed... Perhaps I should have considered water cooling a bit more than I did, hmm? I still would like to keep this rig low-maintenance and highly robust, though, and adding a liquid system inside of it is simply incompatible.

It would appear that this build has been an unqualified success so far. I shall begin to ease off the voltage while keeping this clock ratio. I haven't seen many folks claiming to be running a 5820K at 4.5 with 1.2V or even 1.25V...

I ran some AIDA64 benchmarks. It's pretty sweet to see it near the top, only underneath various 16/20 core setups and consistently beating the stock 5820K by 32%. This level of overclockability is completely on track with the original OC monster that I had, the E6300.

Also the fan speed control capabilities of the Sabertooth are phenomenal. I can get it so the machine sets all fans running at ~400rpm when the machine is idle. This is quieter than a typical hard drive or water pump. At any rate, since my fans are 200mm and 140mm the pitch is low, and the noise although present is a nice soothing low windy kind of white noise. And I can even have it turn all the case fans off (and control exactly the temperature that is required to switch them back on) too.

Congratulations on the nice build. The case is much nicer than the older Antec I'm using.

Thanks! From the looks of it, this thing will be able to last me 10 years, probably. Esp with a OC'd 980Ti if I SLI it or replace it 3 or 5 years from now.

Reckon I ought to back it down to whatever frequency allows me to keep under or near 1.2V if I'm running this all day for 10 years...

Only components I see you needing to replace in the future is the GPU at least a couple of times, adding more DDR4, and changing out storage drives. Other wise the CPU should hold you over for at least 5 years possibly longer at the current pace CPU performance is headed/going.

Yeah, especially if you want to keep it a while, I wouldn't bother overclocking. Save that for later when you really need it. If CPU performance ever becomes much of an issue again, that is. Not sure about games and CPU bottlenecks, but a mobile Penryn and a good SSD choked by SATA II still feels pretty snappy for general use, and that's '08/09-era tech. My 5 year old Sandy Bridge i5 (4 year old PC) is still running quite happily at this point.

biffzinker wrote:

Only components I see you needing to replace in the future is the GPU at least a couple of times, adding more DDR4, and changing out storage drives. Other wise the CPU should hold you over for at least 5 years possibly longer at the current pace CPU performance is headed/going.

I think he will be fine with his 64GB's of system memory but I would have went with a top of the line AIO water cooler like possibly Cooler masters Nepton 280XL but since it seems like he got a golden chip I do not think it will matter much. Plus 4-5 years down the line if he does need more cores he can upgrade it with a used Zeon or even a big Broadwell socket 2011 chip.

Anyway it will serve him well for many many years.

localhostrulez wrote:

My 5 year old Sandy Bridge i5 (4 year old PC) is still running quite happily at this point.

Yup, the ITX build in my sig with the stock-clocked 2500K (wait no, I have it at 4Ghz at Turbo... I dunno, I am tired of dealing with Turbo/EIST now ) definitely is adequate for my needs. My three year old Macbook (which cost more than this thing btw) has (a slow) ivy bridge in it and still keeps up with everything I can throw at it, except for games. The Macbook does not tolerate games well, as it can only convect so much heat.

It's only been this "I want to play with nice hardware" itch I caught. After all this is a ritual of mine that I engage in every few years. And I'm doing this fancy photogrammetry stuff for work and it takes ages to render. Probably still hard-pressed to see a 3x improvement in that with the new system owing to sequential data processing dependencies, but then again 3x is a huge improvement by any measure... I do hope that the stages which require GPU processing will be able to benefit from the 980Ti.

Anyhow that was the last straw as it were, in the box of excuses to pull the trigger on this $2k system. When the current line of graphics cards were released I looked at them and was sure I could never justify $600 on a graphics card. But... I did spend $400 on my 670. I got my $400 worth out of it... and it continues to serve me well today. It's very quiet (so long as nvidia drivers are installed) and is decent enough about not drawing huge power while idle. It also continues to perform excellently for games at 1080p resolution.

So when the 980Ti got to the price point that the GTX 680 used to be at, not that long ago, it just sort of makes sense, it will let me do something I wouldn't even be able to think about otherwise, which is run certain apps in 4K resolution (neat thing about 4K gaming is you can just downgrade yourself back to 1080 once the games start to destroy your hardware again... by realistic standards, things still look great in 1080p HD). Well, now I wait. My body is ready. As is my rig.

I wish your new PC a long and happy productfull life and hope the cuda cores on that 980ti help your renders. I still rec spending the extra 50$ on the Evga 980ti classified. The classifieds are Just so nice with the multiple bios and lots a power phases spaced out across a larger than normal pcb. Again worth the extra 50-55$ IMO

At this point I'm super happy with (almost) everything and when the GPU comes, that will only be even better.

The single thing that is concerning me is that stability tools (essentially, just Prime95, as other tests do not come close to its intensity) are able to make the temperatures reach 90C with alarming speed. I'm talking the small in-place FFT's cranking it to that temp instantly, and I really don't want to damage this nice chip so I'd rather not attempt it again till I get the heatsink working better. And this was back when I was on 4.3Ghz, but similar voltage to now. (Now its 4.5ghz 1.223V)

I'm trying to find information about how well I can expect the D15 to cool the OC'd 5820K but any anecdotes I can find (newegg reviews is filled with them) indicate workable temps, I don't know how reasonable this is but when I see someone claiming prime95 topping out at 75 or 80 degrees with my exact same setup gives me hope that I can reapply my thermal paste and improve my own numbers. I guess it could be similar if for these people's numbers they were using Blend and only having it go for 5 minutes.

Anybody here have anecdotes about having a 5820K or the 5930K on a D15? My 5820K is apparently on a path to thermal runaway ONLY under Prime95 small FFTs, but my temps otherwise also seem high. Like, yeah I can switch to water and keep the core under 60 celsius no matter what I do to it, but for now I want to hold out hope that I can somehow get P95 to run to completion or to run overnight without actually killing the chip. D15 is supposed to be capable of this.

Nice build, I miss having that case (all the front panel buttons / fan controller failed)

Yeah the fan controller I tried and wasn't so keen on it, the mobo tweakable fan control is way nicer. I hope the buttons wont die though but i guess i hardly turn my computers off so they dont get used much.

And I read about how new Prime95 versions push Haswell farther than other "realistic" tests, on account of the FMA/AVX instructions. But, I'm kind of skeptical here since if I configure a system which can handle everything except post-version 26.6 Prime95, then i'll only be stable until the day that something rolls around that is "as stressful as" post-26.6 Prime95...

It does bring my temperature observations in line with what I saw on newegg reviews though, with 26.6, Prime95 can't push past 70 degrees, with the CPU fans kicking in. I also did find a benchie that crashes, the handbrake x264 bench part of ASUS Realbench.

APWNH wrote:

The single thing that is concerning me is that stability tools (essentially, just Prime95, as other tests do not come close to its intensity) are able to make the temperatures reach 90C with alarming speed. I'm talking the small in-place FFT's cranking it to that temp instantly, and I really don't want to damage this nice chip so I'd rather not attempt it again till I get the heatsink working better. And this was back when I was on 4.3Ghz, but similar voltage to now. (Now its 4.5ghz 1.223V)

I'm trying to find information about how well I can expect the D15 to cool the OC'd 5820K but any anecdotes I can find (newegg reviews is filled with them) indicate workable temps, I don't know how reasonable this is but when I see someone claiming prime95 topping out at 75 or 80 degrees with my exact same setup gives me hope that I can reapply my thermal paste and improve my own numbers. I guess it could be similar if for these people's numbers they were using Blend and only having it go for 5 minutes.

Anybody here have anecdotes about having a 5820K or the 5930K on a D15? My 5820K is apparently on a path to thermal runaway ONLY under Prime95 small FFTs, but my temps otherwise also seem high. Like, yeah I can switch to water and keep the core under 60 celsius no matter what I do to it, but for now I want to hold out hope that I can somehow get P95 to run to completion or to run overnight without actually killing the chip. D15 is supposed to be capable of this.

I'm going to guess that Prime95 is causing the PCU to increase the voltage over the 1.223V when the vector units are sustaining heavy load. Your keeping and eye on the voltage while running Prime95? Unless you set the voltage in the EFI to manual it will still boost the voltage.
Edit: A link to HWiNFO download page

biffzinker wrote:

I'm going to guess that Prime95 is causing the PCU to increase the voltage over the 1.223V when the vector units are sustaining heavy load. Your keeping and eye on the voltage while running Prime95? Unless you set the voltage in the EFI to manual it will still boost the voltage.
Edit: A link to HWiNFO download page

Ah good point, I had not attempted Prime95 v28 since I went back to Manual voltage.

I just tried it and instant BSOD. Looks like 4.5 isn't stable at 1.223 this is where my luck runs out I guess. Back to trying to get it stable.

If you can get it stable at 4/4.2 GHz I would call that good enough for now. Considering the defaults for the processor are base frequency 3.3 GHz, max turbo frequency 3.6 GHz, and the TDP for your chip is already 140 watts.

Sure, 4.5 is a nice round number though, but I'd like to have it stable under 1.2V I think 1.2V is my line I'm going to stay under. So it looks like 4.4 or 4.3 probably. We'll see....

OK, I knocked it down a peg to 4.4.

You're totally right. I did not keep my eye on Vcore when I tried P95 earlier. Now that I have manual voltage, my cooler is keeping the temps in check, 82C or so is as high as any core gets. This is much more comfortable for me.

However I just had another BSOD about 2 minutes into it there. This time it did pass the Realbench Handbrake encoding test. I'm not sure why people say P95 isnt the best for stability testing because at least for this system (and all my other systems in the past as I recall) P95 will BSOD a system that is otherwise stable everywhere else...

At 4.3 I appear to be stable. It seems that some of the tests in P95 just really do generate that much heat (enough to reach 90C), and many builders do not get around to this point in their testing to see the extreme temps. I think I will run 1.2V @ 4.3.

I'm done, or, I have to be... there is work that I need to do that I've been neglecting.

APWNH wrote:

Sure, 4.5 is a nice round number though, but I'd like to have it stable under 1.2V I think 1.2V is my line I'm going to stay under. So it looks like 4.4 or 4.3 probably. We'll see....

OK, I knocked it down a peg to 4.4.

You're totally right. I did not keep my eye on Vcore when I tried P95 earlier. Now that I have manual voltage, my cooler is keeping the temps in check, 82C or so is as high as any core gets. This is much more comfortable for me.

However I just had another BSOD about 2 minutes into it there. This time it did pass the Realbench Handbrake encoding test. I'm not sure why people say P95 isnt the best for stability testing because at least for this system (and all my other systems in the past as I recall) P95 will BSOD a system that is otherwise stable everywhere else...

Most likely their pointing to the boosted voltage from AVX2/FMA3 you get running a heavy workload that Prime95 sustains. In the real world your unlikely to be hammering the chip's vector units like Prime95 does, and a lot of complied software doesn't use AVX2, most is just starting to use the older AVX. Take for example running Handbrake through an encoding job only boosts the voltage up over the normal voltage a few times. I switched to running Realbench as a stress test myself.

The last thing Im going to do I think will be to convert both exhaust fans in the case into intake by flipping the top and back fans, and swapping my side panel's window for the mesh cover to turn the side into a passive exhaust. Also will be rotating the D15 by 90 degrees, giving it a little more clearance near the GPU. However this will blow the CPU's exhaust directly onto the GPU, which is far from ideal, but at least my GPU will have a back plate and I think it will generally be able to survive.

That plus I should be doing this testing while putting my PSU out of eco-mode so that it will actually switch on its own fan instead of letting all the PSU heat rise back into the case!!!

Finally got that done... well it's definitely running cooler too, I used less thermal compound this time, really small dab. The extra fresh air and PSU cooling itself probably helps as well. I can even run the ultra high temp new prime95 routines without hitting 90, it dances up to 85 (remember this is with the special subset of Prime95 workloads that just does horrible things... other loads never surpass 70 degrees), but I reckon that's about as good as I'll ever get it. Important part is that it is rock STABLE while doing this too.

This is 1.200V, 4.3Ghz (43x100) BTW.

Still more I can do such as

1. Uncore/cache OC to beyond the nominal 3000Mhz
2. Figure out how to keep this stable OC while enabling SpeedStep to save some power and to stay even cooler
3. RAM OC

But it's very much diminishing returns at this point, plus I'm kind of burnt out working on this for 3 straight days. And with more OC tweaking to come when the 980Ti arrives.

APWNH wrote:

At 4.3 I appear to be stable. It seems that some of the tests in P95 just really do generate that much heat (enough to reach 90C), and many builders do not get around to this point in their testing to see the extreme temps. I think I will run 1.2V @ 4.3.

I'm done, or, I have to be... there is work that I need to do that I've been neglecting.

Is this some kind of stealth post to a past post? I'm on to you now.

If you want to you can run the below at a command prompt with admin privileges or a batch file to clear out the event logs from when you were getting BSODs during your overclocking attempts. It doesn't do anything else but clear ALL of Windows event logs btw.

Huh, cool. What's so bad about having those event logs around though?

I guess if there are 64GB core dumps hanging around I could stand to clean them up (nope, my disk would be full, haha) but... I am kind of already thinking of wiping the partition already because I do want to get WIndows 10. Just have to find out where to obtain a copy that works properly, first.

And nah, i'm not referencing anything. I just have real actual work that I've been (still) putting off while I do all this tweaking. You know it is starting to make a lot of sense now that only teenagers and college age folks have the kind of free time to tweak hardware like this. It is friggin' exhausting!! I mean, if I only had typical levels of free time, I could only have accomplished all I did here in maybe a week, week and a half. It quickly becomes clear that for an actual professional whose time is worth something, it basically makes sense to just spend more money on something that is ready to go out of the box.

I think part of it is just me being an idiot and leaving my monitor/kb/mouse and new rig on the floor of my bedroom and having to be hunched over here while working on it. The thought was that it would motivate me to get it done quicker but I think it has just tired me out quicker.

Not as young as I used to be...

Never mind if your going to a clean install. I've have had some past builds go together fairly quickly allowing time to overclock, and figure out stable clock speed so as to not cut into my time needed else where.

You should be able to do a clean install of Windows 10 from a USB drive with your 7/8.1 product key you can skip doing the upgrade, and go straight to booting from USB then clean install you might want to wipe your past/current partitions in setup before the installer kicks over to copying files.

Edit: It's the Recovery partitions from past Windows installs is while I'm suggesting deleting partitions, then letting Windows 10 Setup auto create partitions.
Edit2:Sorry to hear your not your younger self, 37 myself so maybe I'm not as old I feel like sometimes.

Hm I was under the impression that Win10 was only gonna let you do it as an upgrade, and we all know that upgrading Windows is about the worst possible thing you can try...

Well i'm sure I'll figure this problem out easily enough.

Use to be that way before, now you can download the current iso put in the 7/8.1 product key during setup then Windows will later auto activate. Any future reinstalls will automatically activate unless you make significant changes to your system.
Edit: Link to Windows 10 media creation tool

Pretty sweet. Feels even faster than Win7. Thumbs up.