... would be this puppy:
https://youtu.be/Mc7r3n8CsQU
Bet all you two-account chaps and chapettes are drooling now!
https://youtu.be/Mc7r3n8CsQU
Bet all you two-account chaps and chapettes are drooling now!
Hardware & Technical For Elite, my ultimate Christmas present...
- Thread starter Tyres O'Flaherty
- Start date
you need a backup solution
https://www.caseking.de/en/8pack-or...8086k-5-2-ghz-extreme-oc-system-sipc-186.html
https://www.caseking.de/en/8pack-or...8086k-5-2-ghz-extreme-oc-system-sipc-186.html
"i7 8700k at a blistering 4.8Ghz".
Meaning they haven't even bothered to de-lid and drop some liquid metal in there. Or even done much beyond a very safe overclock on the 8700k.
With this kind of cooling you would have to have faired very very poorly in the silicon lottery to not break 5ghz, a good one even without being delidded.
I'm running mine at 4.9 with an off the shelf AIO, and no I haven't delidded my cpu either. if I do I would expect it to at least hit 5-5.2ghz.
But I'm already at the point where I suspect I might need better VRMS on my mobo to reach that and for now 4.9ghz is good enough and I still don't have load temps higher than 77c. Gaming peak temps are low 60's.
idle temps at about 27-29.
But it's not warm, I barely break 18c room temp this time of year.
It was almost stable at 5 but at those volts I was also seeing core temps at the low 90's Celsius, still not quite near throttling down levels of heat but not worth it imo.
I also suspect my improvement is not so much from peak clock but more from the fact that all cores run at this frequency and never step down to 3.7ghz.
I'm sure it's a sweet rig.
But for this kind of dosh I would prefer to build two seperat rigs of my own.
But I find that almost as much fun as playing.
So that could just be me.
And you could probably still afford a full high end HOSASAT (Hands on stick and stick and throttle)
And a motion platform to boot if you would.
Meaning they haven't even bothered to de-lid and drop some liquid metal in there. Or even done much beyond a very safe overclock on the 8700k.
With this kind of cooling you would have to have faired very very poorly in the silicon lottery to not break 5ghz, a good one even without being delidded.
I'm running mine at 4.9 with an off the shelf AIO, and no I haven't delidded my cpu either. if I do I would expect it to at least hit 5-5.2ghz.
But I'm already at the point where I suspect I might need better VRMS on my mobo to reach that and for now 4.9ghz is good enough and I still don't have load temps higher than 77c. Gaming peak temps are low 60's.
idle temps at about 27-29.
But it's not warm, I barely break 18c room temp this time of year.
It was almost stable at 5 but at those volts I was also seeing core temps at the low 90's Celsius, still not quite near throttling down levels of heat but not worth it imo.
I also suspect my improvement is not so much from peak clock but more from the fact that all cores run at this frequency and never step down to 3.7ghz.
I'm sure it's a sweet rig.
But for this kind of dosh I would prefer to build two seperat rigs of my own.
But I find that almost as much fun as playing.
So that could just be me.
And you could probably still afford a full high end HOSASAT (Hands on stick and stick and throttle)
And a motion platform to boot if you would.
There was a time in my life where I loved that too, not gaming but for recording studio stuff, built my own with 4x 10,000 rpm Raptor drives in RAID-0 config et al. was a sweet rig in deed.
The stuff today is out of this world, really! LOL
Not for this Christmas for me.
There was a time in my life where I loved that too, not gaming but for recording studio stuff, built my own with 4x 10,000 rpm Raptor drives in RAID-0 config et al. was a sweet rig in deed.
The stuff today is out of this world, really! LOL
And now you just grab a few SSD's and the major reason for RAID's is the failure recovery potential.
What astonishes me though is mostly how simple it's gotten it's honestly easier to build a rig now yourself than it is installing a shelf from Ikea..
And AIO's are practically as good as most custom water cooling solutions used to be.
Like I mentioned I built myself what I somewhat consider a beast for my uses last year.
But it's also silent, unless it's going full tilt, still not as loud as my Core 2 duo was, that sounded like an industrial vacum cleaner.
Bloody hell! I was going to disagree with you until I saw the price; !!
I’ve got my 8700k to 5Ghz but discovered it wasn’t stable when I used Prime95. Jumping from 4.8 to 4.9 is a pretty be deal temp wise if you want to pass Prime95. I’ve settled on 4.8 with my NZXT Kraken.
Scan are a really good company generally and if they overclock and delid they are the ones that void the warranty so effectively take responsibility. £14000 though lol, I see your point!
I'd probably omit the LGA-1151 components entirely, spend the money on a good water chiller (set to keep the coolant at 1C over the dew point), and just run the 7980XE at ~5GHz on all cores.
Stability, and how picky one should be about it, is a point of contention among overclockers.
Personally, I have overclocked systems doing my wife's bioinformatics work, where a single job can take weeks at a time and errors that slip through could invalidate months or years of work, costing grants or resulting in the rejection of publications, so I'm extremely picky. Other people, who may simply want to get through a gaming session without a BSOD or lock up, have a bit more wiggle room.
Stability, and how picky one should be about it, is a point of contention among overclockers.
Personally, I have overclocked systems doing my wife's bioinformatics work, where a single job can take weeks at a time and errors that slip through could invalidate months or years of work, costing grants or resulting in the rejection of publications, so I'm extremely picky. Other people, who may simply want to get through a gaming session without a BSOD or lock up, have a bit more wiggle room.
Last edited:
Then they install windows on it.
The lady seem quite excited reading gout the specs.
Imagine forgetting to remove that thin plastic cover on the cip before installing!
I remember when building a computer actually meant just than ...
[video=youtube;VCzbn1XH2FE]https://www.youtube.com/watch?v=VCzbn1XH2FE[/video]
The lady seem quite excited reading gout the specs.
Imagine forgetting to remove that thin plastic cover on the cip before installing!
I remember when building a computer actually meant just than ...
[video=youtube;VCzbn1XH2FE]https://www.youtube.com/watch?v=VCzbn1XH2FE[/video]
[video=youtube;s9F5fhJQo34]https://www.youtube.com/watch?v=s9F5fhJQo34[/video]
Ugh, when you said "puppy", I thought the video was going to show a cute young dog.
WAIL
WAIL
I've never been big into overlocking and only really started doing it on my last PC. Until I used Prime95 I was completely unaware that instability meant that a computer could come up with the wrong answer to a calculation and carry on working. I'd assumed that it would just lock up.
I'd side with you on the picky front. If a computer is going to start making mistakes when running too fast then I'd rather it run a bit slower and have it compute correctly.
For my overlocks I've decided it must pass prime95 for 10 mins and not go over 80c. For me in real world use it should be very stable and never get near thermal throttle.
Over Six-Thousand Euros? And it isn't even stable. And it's over-kill. For that kind of money it should come with a bl-w j-b.
It is not a bug, it is a feature, well known and cherished amongst the well known suspects, the bunch of very serious people, bank-st-ers.
Ops, sooo sorry, but we demand just a wee additional 2.5 billion to safe us from spontaneous ruin! - whispers... and you from not being re-elected, har har har! - Yeah, we know we already got 1.5 billion last week, but as I said...
That's overclocking for serious people.
Last edited:
The overwhelming majority of errors won't crash a system and most won't even result in overt problems. There is a very large gulf between "suicide run", "bench stable", "game stable", and "willing to bet a life or career on this being the right answer". Errors are common, even in stock systems, and mostly go unnoticed. Error rates are even written into the specs of many components, most notably hard drives (with typical consumer hard drives flipping a bit every 10^14 to 10^15 bits written and enterprise drives being an order of magnitude or two better). A large portion of bit rot and silent corruption could be prevented by more rigorous stability testing.
There is a certain chance for an error, that will start off very near (but not at) zero and increases exponentially as components near various limits. The purpose of stress/stability testing is to push closer to those limits than normal use can, and do enough enough work to prompt an error in a reasonable period of time. If no errors can be found in relevant tests, then one can tentatively call a system stable. I say 'tentatively' because stability isn't something that can be proven, only disproven.
While there are a variety of broadly useful stress testing applications, none are comprehensive or infallible. Even if a broad selection of tests pass individually, there still exists potential for overt instabilities in combination testing, when multiple subsystems are loaded simultaneously. I normally use two or three different tests for each component/subsystem, then test with the most demanding of each all running together. How long I test will depend on how important I consider the data/work on the system in question to be. I'll spend a weekend testing a gaming system, with maybe 40 hours of testing total. A bioinformatics workstation might take me a month of work to validate--a week to break in the system with extreme loads and identify/replace defective components; followed by systematic testing and identification of the limits of each component; settling on reasonable margins to account for wear & tear over it's expected lifetime; validating everything together in worst case environmental conditions; then another long period extreme stress testing, after which I compare peak apparently stable settings to the results I initially recorded immediately after the break-in period. Assuming nothing has changed, I do a final combination test and should I be unable to find any problems after a ~72 hour run, I'll drop the clocks to the bottom of the aforementioned margin and consider it good enough to use.
There are those who consider an acceptable, or even ideal, stress test to be using real-world applications that the system will actually run. While such applications should absolutely be tested, relying on them as a stress test is extremely reckless as it's completely impractical to run a long enough test to reliably forecast future stability. For example, if I have a system I'm building to transcode video, and it's unstable enough to fail an eight hour transcode half the time, it could still transcode for 24 hours straight with a 1-in-8 chance of passing. A more demanding set of tests could have the same chance of failure every few minutes, allowing problems to be discovered during the allotted test period, rather than after the system had been doing real work for days or months.
Anyway, ten minutes of Prime95 really doesn't say much. If temps are an issue, you probably need a larger AVX offset. If temperature is under control and you aren't using excessive voltages, there is no reason not to test longer. It's also wise to try a few different tests, as well as test GPU and I/O concurrently with a CPU/memory test. Don't need to go overboard to find most likely issues, but a few hours, maybe overnight, would be much more confidence inspiring than a 10 minute run. Of course, if that system doesn't do anything important, then it doesn't really matter.