1210MB of RAM installed?

[Korth]

The latest ROG overclocking update/news mentions a new world-record, der8auer's i9-7980XE (Engineering Sample) overclocked on LN2 to 6.104GHz on all 18 cores with HT enabled. The CPU-Z validation reports total system memory was 1210MB.

1210MB is a strange value, lol.

1x1GB?
2x512MB?

The numbers don't work out. Was some higher memory capacity (say 2x1GB?) installed but not being fully addressed by the motherboard after POST? Has the motherboard or memory been damaged from repeated LN2 exposure? Do the processor iMCs become less stable or partially malfunctional under these extreme conditions?

(And, if so, is this sort of thing "normal" for extreme LN2 overclocking?)

[MasterC]

Hi Korth,

In this case, MaxMem was used to limit the memory capacity used 🙂
http://www.analogx.com/contents/download/system/maxmem/freeware.htm

[haihane]

der8auer, oh debauer. first you destroyed a threadripper just to see what the insides are like,
and now you bravely overclocked the most expensive intel chip (for me it is, spare me the semantics), with LN2! which i haven't even dared to even try in my wildest dreams.

and i see that korth is properly acclimating with debauer now. /wink


(i maybe making an ass out of myself, but i don't care. bless that son of a gun. he got the best job ever!)

[Korth]

MasterC@ASUS wrote:
In this case, MaxMem was used to limit the memory capacity used 🙂
http://www.analogx.com/contents/download/system/maxmem/freeware.htm

Thanx!
I hadn't heard of this tool before. To be honest, I don't really see the value of running it (what benefit it provides) during competitive OC sessions. But I am not a competitive overclocker and I don't know all (or even very many) of the leet secrets, lol.

haihane wrote:
der8auer, oh debauer. first you destroyed a threadripper just to see what the insides are like,
and now you bravely overclocked the most expensive intel chip (for me it is, spare me the semantics), with LN2! which i haven't even dared to even try in my wildest dreams.

Well, these are Engineering Sample or Media Sample parts. Obviously obtained outside normal ($2000) consumer channels, lol probably given to der8auer specifically so he can analyze, advertise, overclock, overvolt, de-lid, freeze, cook, burn, and explode ... I don't think (ASUS or Intel or AMD or whomever) really expects these items to be returned, let alone returned "intact". But I'm only speculating, could be wrong.

and i see that korth is properly acclimating with debauer now. /wink

Yeah. Kinda surprising since I've long been aware of K|NGP|N and Fatal1ty and even (my fave) Dinos22 ... yet never heard of der8auer, lol.
He certainly knows his stuff. He's got plenty of top-tier tech to play with. And he explains it well, has charisma. Always useful for impressing the ladies, "hey baby, let me pop the panel off this thing and show you my nitro-powered 18 cores at 6Ghz, zero to pi in 6.3 seconds ..."

The overclocking scene is strange these days. It simultaneously seems bigger and smaller than it was some years ago. Elite-tier "professional" overclockers are brands in themselves, champion vassals with strong loyalties to their specific sponsors. I agree that it's an awesome job, lol.

When in Rome ...

[Korth]

I find it interesting that Win7 appears to be overwhelmingly preferred for hard overclocking benchmarking.

I can understand linux being too far off the mainstream and perhaps offering too few of the preferred software tools. But not Win8, not Win10? No DX12?

Win7 happens to be my preferred WinOS as well, but it is technically EOL and is increasingly becoming incompatible with (unsupported by) the latest hardware tech. Is it "fair" to proudly advertise a motherboard's/processor's record-setting overclocks on an old WinOS when many of the important new features (selling points) built into the motherboard/processor aren't actually supported by or compatible with the old WinOS?

[Menthol]

Different benchmarks are better run on different OS's, some are best on XP, some Win 7, Win 8.1, 10, some, 32 bit, some 64 bit, some on Pro, some on enterprise edition, there is as much or more tweaking of the OS as there is overclocking to be competitive when benchmarking,
I have been dual booting Win 7 and 8.0 or 8.1 and 10 for a long time preferring 7, now that support ended on Sky Lake I am forcing myself to use Win 10 exclusively on my daily rig and liking it

[Korth]

I admit I still feel it's a bit of a "cheat". As in, technically, if the machine requires Win10 to fully support all hardware functions then using an OS which doesn't fully support all hardware functions (like Win7 or even WinXP) is not a properly representative or meaningful benchmark. Almost misleading.

I do understand overclockers controlling all the parameters and aligning every possible factor to achieve their results, streamlining everything possible, focussing only on aspects that improve performance and discarding aspects which don't, extreme results require extreme approaches. And we all know that software bloat is the overwhelming contributor of underwhelming performance.

And consumers aren't seriously intended to run their desktops on LN2 cooling anyhow, so it's already a special context.
Some overclock records have been made on modded firmware, sometimes provided by the OEM, sometimes custom, which is also not seriously intended (or even available) for consumers to use on their systems.

But it still somehow kinda seems like a "cheat". Perhaps they should also voltmod the mobos, rewire all the VRM hardware, replace it with different VRM hardware entirely, solder on modified processor sockets, use off-board processing to augment the chipset, etc? Overclocking is already out of spec (overspec) by definition. But how much can you really change the platform configuration before it's not even the "same" system, where to draw the line?

I guess I'd feel less "cheated" if the line was drawn closer to where the end-user would actually use (and overclock) his own components. It's a given that people who buy SKL/KBL/etc need to run Win10 to make full use of the their new tech, so it should be given that achieving 6.1GHz under Win7 (or, hypothetically, 10GHz under MSDOS 6.22) is not a "real" measure of maximal motherboard/processor capabilities.

All that being said, I vote for some linux-based overclocking competition. Somebody's gotta be willing to kickstart it.