Numba WAN in the world.

[Avacado]

Whelp, iv'e done just about all I can. I am #2 rookie overclocker in the world right now. Some dude from France came and blew me out of the water and I should have had 1st. Will possibly get Alder Lake and see if I can compete again. I am looking to get more members active in HWbot. I know @Bastiaan_NLis getting back into it. 

 

 

Have #11 9900k in the world for Cinebench R23. I know I could get way better than this, but I am terrible with mem overclocking. Most of the people ahead of me have lower core OC's, but better mem speeds/timings. I know I could get top 5 if I had some help here. 

 

Edited January 22, 2022 by Avacado

[The Pook]

at least for CB the difference in points probably has more to do with the OS than RAM, CB doesn't care too much about RAM.

 

look into Windows 8.1 Embedded Industry Pro, I run it as a bench OS since getting W7 working is a PITA on Z170+. I bugged mllrkllr88 for help far too many times and gave up  a clean and (mostly) unmodified 8.1 EIP install gets me about a 40 point bump in R15 compared to a optimized W10 install with my 10900K at identical clocks. 

 

Windows 10 Ameliorated might get a similar boost since it has a ton of stuff stripped out but I haven't tried it yet: 

Download - AME

AMELIORATED.INFO

---

A stable, non-intrusive yet fully functional build of Windows 10 to anyone, who requires the Windows operating system natively

 

 

if you have B-Die, look into the OCN DDR4 thread and find someone else running B-Die and just steal their timings at a clock you can run. Even if your kit sucks and tops out at like 4000, find someone running 4266 and steal their timings. 

 

if you're running 16x2 and not 8x4 you can steal mine but it's my daily clocks, I don't really have a HWBOT profile figured out for it yet: 

 

 

you can run a lot more aggressive RAM OCs if you limit your RAM in MSCONFIG to as little as you can without losing performance in whatever benchmark you're running. 4-5GB is plenty for CB

Edited December 4, 2021 by The Pook

[damric]

Well the #1 Rookie is a cheater. It took me all of 5 minutes to see that he has done a bunch of shady things against the rules. I would report his submissions, but the staff guy in charge of that rarely does anything when I report things but he does always seem to nickel and dime me for the slightest things.

 

Of you look at his 3Dmarks, he has manipulated the drivers and is flagged in the 3Dmark database.

 

His CPU-Zs are missing a version number in the screenshots, which makes me wonder if any of his CPU scores are real.

 

I bet there's more but I only spent like 5 minutes looking at the profiles.

 

There's tons of cheaters on HWBOT. I too should have assloads more golden cups. Don't let it discourage you. 

 

You under water for all of your tests? I think you can still get away with slushing the radiators to stay in Enthusiast league. You can also spray duster cans into air coolers. Also remember you only need to be barely stable enough to submit your score.

 

I saw @Bones joined the forum recently and I'm sure he can give you better advice.

[Avacado]

2 hours ago, The Pook said:

at least for CB the difference in points is probably has more to do with the OS than RAM, CB doesn't care too much about RAM.

 

look into Windows 8.1 Embedded Industry Pro, I run it as a bench OS since getting W7 working is a PITA on Z170+. I bugged mllrkllr88 for help far too many times and gave up  a clean and (mostly) unmodified 8.1 EIP install gets me about a 40 point bump in R15 compared to a optimized W10 install with my 10900K at identical clocks. 

 

Windows 10 Ameliorated might get a similar boost since it has a ton of stuff stripped out but I haven't tried it yet: 

Download - AME

AMELIORATED.INFO

---

A stable, non-intrusive yet fully functional build of Windows 10 to anyone, who requires the Windows operating system natively

 

 

if you have B-Die, look into the OCN DDR4 thread and find someone else running B-Die and just steal their timings at a clock you can run. Even if your kit sucks and tops out at like 4000, find someone running 4266 and steal their timings. 

 

if you're running 16x2 and not 8x4 you can steal mine but it's my daily clocks, I don't really have a HWBOT profile figured out for it yet: 

 

 

you can run a lot more aggressive RAM OCs if you limit your RAM in MSCONFIG to as little as you can without losing performance in whatever benchmark you're running. 4-5GB is plenty for CB

Thank you for all that.  I don't think my kit is B-Die, however I can't get anything past my XMP 3600 profile. I have tried quite a lot and the damn thing just resets BIOS. I will look into alternate OS installations. I just bought a DDR4-4000 kit off Amazon. Will try that when it arrives. 

 

MOBO - Gigabyte Z390 AORUS Pro

RAM - F4-3600C16D-16GTZSW

 

 

Edited December 4, 2021 by Avacado

[Avacado]

22 minutes ago, damric said:

Well the #1 Rookie is a cheater. It took me all of 5 minutes to see that he has done a bunch of shady things against the rules. I would report his submissions, but the staff guy in charge of that rarely does anything when I report things but he does always seem to nickel and dime me for the slightest things.

 

Of you look at his 3Dmarks, he has manipulated the drivers and is flagged in the 3Dmark database.

 

His CPU-Zs are missing a version number in the screenshots, which makes me wonder if any of his CPU scores are real.

 

I bet there's more but I only spent like 5 minutes looking at the profiles.

 

There's tons of cheaters on HWBOT. I too should have assloads more golden cups. Don't let it discourage you. 

 

You under water for all of your tests? I think you can still get away with slushing the radiators to stay in Enthusiast league. You can also spray duster cans into air coolers. Also remember you only need to be barely stable enough to submit your score.

 

I saw @Bones joined the forum recently and I'm sure he can give you better advice.

I had not caught that, thank you for pointing it out. I have reported him. Only the 9900k's and 3070 are under water. I have my golden 9900k @5.3GHz on water in the test bench with my 3070Ti. I don't think i'll be water blocking it. 

[The Pook]

25 minutes ago, damric said:

Of you look at his 3Dmarks, he has manipulated the drivers and is flagged in the 3Dmark database.

 

haven't looked at his scores but flagged scores doesn't necessarily mean cheating. LOD and/or Tesselation adjustment is allowed in some benchmarks even though 3DMark flags them

[Avacado]

3 minutes ago, The Pook said:

 

haven't looked at his scores but flagged scores doesn't necessarily mean cheating. LOD and/or Tesselation adjustment is allowed in some benchmarks even though 3DMark flags them

His non-version CPU-z scores are disqualifying for all of his results. Also his DDR4 running at 5026 with a CAS of 16 is concerning to me. 

[J7SC_Orion]

First of, 'grats on your HWBot performance ! 

Further to what @The Pook wrote, there are indeed differences between HWBot and for example 3DM / HoF rules on what mods are allowed. I've competed in both before, but it pays to double-check their respective rules when you want to submit a result to both.

 

Years ago, I did a lot of HWBot / Elite league (below), and for GPU-type subs, there were actually 'non-public' versions of NVidia-Inspector floating around which allowed for much more drastic steps (ie. LOD levels) compared to the NVidia Inspector versions you can download publicly. 

 

On bench OS, in addition to 'Ameliorated' above, check > this and > this Win 10 bench setup out which were posted by a fellow at the 3090 thread, with others reporting decent results (make sure to read both).

 

On RAM, hopefully, the 4000 kit you're getting from Amazon will boost your results further. I have multiple  4x8 Samsung B-die 'GTZR' kits (including a 4000 CL15), and they all either run much tighter timings and/or more MHz than spec. Your limiting factor should not be the RAM, but the IMC of your CPU, for example.

 

The GTZR kits were not even on the Threadripper / Ryzen QVL when I got them (might have changed now) and some of the kits were tagged 'Intel Z270/370' etc, but they worked just the same (and better) on my Ryzen systems than on my Intel Skylakes. The point is that you can use > this Ryzen DRAM calculator for both Ryzen and Intel and get awfully close on 'fast' settings in that app to optimize your RAM for your Intel 9900K, while saving a ton of time and frustration...RAM optimization being the rabbit hole that it is.

 

I had chatted before with @Bastiaan_NL about the 'good 'ol days' of my HWBot adventures. I did it for several years, mostly in the 'Elite' (> subzero LN2 etc) league...

At that (long-ago) time, I also gained sponsorship from select hardware vendors who had supplied some 'Engineering Sample/ES' chips, first-release mobos, RAM, new GPUs and the like.

 

But once you have some decent results in your league, it becomes a bit of a slippery slope re. 'sub-ambient / sub-zero cooling' etc. In my case, it eventually just became too time-consuming, what with work, social life etc. There were also just too many folk in other bench-marking threads who turned hostile just because their nose was out of joint after being displaced. In any case, I still bench - but just against myself -  w/ a new setup. Also, I discovered that I like building new systems more than benching them against 3rd parties.

 

All that aside, once you have tuned your RAM, utilize a bench-optimized OS and have a decent CPU and/or GPU chip to compete with along w/ perhaps some custom v/bios, it is all about 'Watt = heat energy'. It doesn't matter if it is rookie or elite league.

 

Most benchmark apps report some sort of average temps, not starting temps, so as long as you're reported temps are not in the single digit C (or lower) temp range, you can experiment with cooling w/o entering the sub-zero leagues.

 

A simple step would be some QDs for your w-cooling loop which can hook in an additional rad that gets submerged in cold water with 1kg or 2kg  of store-bought water ice cubes (making sure that there is actual flow through the submerged rad, such as an extra pump, but obviously NOT fans).

 

Most box store sell those 6+ L annoyingly squeaky Styrofoam boxes which are useful for that...just make sure it has a Rubbermaid underlay as they eventually do leak.  I did a fair amount of multi-GPU HWBot up to 4000 W total where even 2kg of ice water would only last for a single bench run if that, but with single or dual GPU, you probably have enough for two or three runs. 

 

It was always fun reading the HWBot forum comments on what 'sub-ambient' really meant, ie. if you were benching in certain parts of Canada, Russia, Norway and the like during the winter. Some folks had their systems on the patio outside, connected with tubes and wiring via their kitty's trap door  to the inside, but hey, ambient...

 

If there's a Praxair or similar near you, you might even want to pick up a few kilograms of DICE / Dry Ice. You can add it to the aforementioned Styrofoam tub / water ice setup...it will make a huge racket and have hard cubes dance on your submerged rad's fins for a short while, but it will certainly help with benching temps. It will take some experimentation, but as mentioned, you want to make sure to stay within the reasonable parameters at HWBot for given class - a reported 21 C is probably ok, 5 C, not so much...

 

Good luck, hoping to see your additional results soon !

  

  

 

 

 

 

Edited December 4, 2021 by J7SC_Orion

[ENTERPRISE]

9 hours ago, Avacado said:

Whelp, iv'e done just about all I can. I am #2 rookie overclocker in the world right now. Some dude from France came and blew me out of the water and I should have had 1st. Will possibly get Alder Lake and see if I can compete again. I am looking to get more members active in HWbot. I know @Bastiaan_NLis getting back into it. 

 

 

Have #11 9900k in the world for Cinebench R23. I know I could get way better than this, but I am terrible with mem overclocking. Most of the people ahead of me have lower core OC's, but better mem speeds/timings. I know I could get top 5 if I had some help here. 

 

Looks like you are rocking it bud. Well done

[Bones]

5 hours ago, damric said:

Well the #1 Rookie is a cheater. It took me all of 5 minutes to see that he has done a bunch of shady things against the rules. I would report his submissions, but the staff guy in charge of that rarely does anything when I report things but he does always seem to nickel and dime me for the slightest things.

 

Of you look at his 3Dmarks, he has manipulated the drivers and is flagged in the 3Dmark database.

 

His CPU-Zs are missing a version number in the screenshots, which makes me wonder if any of his CPU scores are real.

 

I bet there's more but I only spent like 5 minutes looking at the profiles.

 

There's tons of cheaters on HWBOT. I too should have assloads more golden cups. Don't let it discourage you. 

 

You under water for all of your tests? I think you can still get away with slushing the radiators to stay in Enthusiast league. You can also spray duster cans into air coolers. Also remember you only need to be barely stable enough to submit your score.

 

I saw @Bones joined the forum recently and I'm sure he can give you better advice.

I'll have to check that guy out - If there are enough of us seeing what they've done as a cheat they'll get it but good, that I can promise you.
However we all know the bot isn't exactly "Good" about taking care of things properly either, just how it is.
Some can slide all day, others get nickeled and dimed to death.

Be aware ANY form of chilled water automatically places you in the "Apprentice" league, be it slush box or any other - Only ambient temp water setups are considered as "Enthusiast" or lower in class.

Be sure to read up on all the class requirements before you find yourself going up against the apprentice guys because some of those can kick some serious butt when they want to. Also know in time you will move up to "Novice" (6 months) and then "Enthusiast" (1 year) and that's across the board for everyone.

Edited December 4, 2021 by Bones

[Avacado]

5 hours ago, J7SC_Orion said:

First of, 'grats on your HWBot performance ! 

Further to what @The Pook wrote, there are indeed differences between HWBot and for example 3DM / HoF rules on what mods are allowed. I've competed in both before, but it pays to double-check their respective rules when you want to submit a result to both.

 

Years ago, I did a lot of HWBot / Elite league (below), and for GPU-type subs, there were actually 'non-public' versions of NVidia-Inspector floating around which allowed for much more drastic steps (ie. LOD levels) compared to the NVidia Inspector versions you can download publicly. 

 

On bench OS, in addition to 'Ameliorated' above, check > this and > this Win 10 bench setup out which were posted by a fellow at the 3090 thread, with others reporting decent results (make sure to read both).

 

On RAM, hopefully, the 4000 kit you're getting from Amazon will boost your results further. I have multiple  4x8 Samsung B-die 'GTZR' kits (including a 4000 CL15), and they all either run much tighter timings and/or more MHz than spec. Your limiting factor should not be the RAM, but the IMC of your CPU, for example.

 

The GTZR kits were not even on the Threadripper / Ryzen QVL when I got them (might have changed now) and some of the kits were tagged 'Intel Z270/370' etc, but they worked just the same (and better) on my Ryzen systems than on my Intel Skylakes. The point is that you can use > this Ryzen DRAM calculator for both Ryzen and Intel and get awfully close on 'fast' settings in that app to optimize your RAM for your Intel 9900K, while saving a ton of time and frustration...RAM optimization being the rabbit hole that it is.

 

I had chatted before with @Bastiaan_NL about the 'good 'ol days' of my HWBot adventures. I did it for several years, mostly in the 'Elite' (> subzero LN2 etc) league...

At that (long-ago) time, I also gained sponsorship from select hardware vendors who had supplied some 'Engineering Sample/ES' chips, first-release mobos, RAM, new GPUs and the like.

 

But once you have some decent results in your league, it becomes a bit of a slippery slope re. 'sub-ambient / sub-zero cooling' etc. In my case, it eventually just became too time-consuming, what with work, social life etc. There were also just too many folk in other bench-marking threads who turned hostile just because their nose was out of joint after being displaced. In any case, I still bench - but just against myself -  w/ a new setup. Also, I discovered that I like building new systems more than benching them against 3rd parties.

 

All that aside, once you have tuned your RAM, utilize a bench-optimized OS and have a decent CPU and/or GPU chip to compete with along w/ perhaps some custom v/bios, it is all about 'Watt = heat energy'. It doesn't matter if it is rookie or elite league.

 

Most benchmark apps report some sort of average temps, not starting temps, so as long as you're reported temps are not in the single digit C (or lower) temp range, you can experiment with cooling w/o entering the sub-zero leagues.

 

A simple step would be some QDs for your w-cooling loop which can hook in an additional rad that gets submerged in cold water with 1kg or 2kg  of store-bought water ice cubes (making sure that there is actual flow through the submerged rad, such as an extra pump, but obviously NOT fans).

 

Most box store sell those 6+ L annoyingly squeaky Styrofoam boxes which are useful for that...just make sure it has a Rubbermaid underlay as they eventually do leak.  I did a fair amount of multi-GPU HWBot up to 4000 W total where even 2kg of ice water would only last for a single bench run if that, but with single or dual GPU, you probably have enough for two or three runs. 

 

It was always fun reading the HWBot forum comments on what 'sub-ambient' really meant, ie. if you were benching in certain parts of Canada, Russia, Norway and the like during the winter. Some folks had their systems on the patio outside, connected with tubes and wiring via their kitty's trap door  to the inside, but hey, ambient...

 

If there's a Praxair or similar near you, you might even want to pick up a few kilograms of DICE / Dry Ice. You can add it to the aforementioned Styrofoam tub / water ice setup...it will make a huge racket and have hard cubes dance on your submerged rad's fins for a short while, but it will certainly help with benching temps. It will take some experimentation, but as mentioned, you want to make sure to stay within the reasonable parameters at HWBot for given class - a reported 21 C is probably ok, 5 C, not so much...

 

Good luck, hoping to see your additional results soon !

  

  

 

 

 

 

That was a good read, thank you. I had no idea you were vendor sponsored at one point. I do understand the rabbit hole that HWbot can send you down. I would never be upset about being displaced fairly and other than being a bit upset, I wasn't at all mad until @damric mentioned the possible cheating. I have used 1usmus's calc before for Ryzen and will try again for Intel. There has to be something i'm missing on the MOBO. Every setting I set just goes black and won't post (I am waiting for tuning). When I finally get back into BIOS the RAM is set at 3444. It's weird. My test bench in the dungeon will be 100% QDC's as soon as the extra 4 from Koolance arrive. I am in complete agreement with you there. I will keep the DICE in mind!

 

P.S. Don't see an intel CPU option in the RAM calc, what setting should I use?

Edited December 4, 2021 by Avacado

[Avacado]

3 hours ago, Bones said:

I'll have to check that guy out - If there are enough of us seeing what they've done as a cheat they'll get it but good, that I can promise you.
However we all know the bot isn't exactly "Good" about taking care of things properly either, just how it is.
Some can slide all day, others get nickeled and dimed to death.

Be aware ANY form of chilled water automatically places you in the "Apprentice" league, be it slush box or any other - Only ambient temp water setups are considered as "Enthusiast" or lower in class.

Be sure to read up on all the class requirements before you find yourself going up against the apprentice guys because some of those can kick some serious butt when they want to. Also know in time you will move up to "Novice" (6 months) and then "Enthusiast" (1 year) and that's across the board for everyone.

Understood, thank you Bones for chiming in!

[damric]

Ok just refreshed myself on the HWBOT 3Dmark rules.

 

Firestrike: LOD/Tess driver manipulation allowed.

Ice Storm: Not allowed

Cloud Gate: Not allowed

Sky Diver: Not allowed

Time Spy: Not allowed

Vantage: Allowed

3Dm11: Allowed

 

So basically the older stuff feel free to cheat the benchmark by lowering the Tessellation and Level of Detail to nothing via driver manipulation. The 3DMark site will flag it but it does not matter for your HWBOT submission. If I remember correctly, old versions of the benchmarks didn't flag that, so it's probably allowed so that new benches are on the same level playing field.

 

I still don't know why that guy has a weird CPU-Z. Very sketchy.

Edited December 4, 2021 by damric

[J7SC_Orion]

5 hours ago, Avacado said:

That was a good read, thank you. I had no idea you were vendor sponsored at one point. I do understand the rabbit hole that HWbot can send you down. I would never be upset about being displaced fairly and other than being a bit upset, I wasn't at all mad until @damric mentioned the possible cheating. I have used 1usmus's calc before for Ryzen and will try again for Intel. There has to be something i'm missing on the MOBO. Every setting I set just goes black and won't post (I am waiting for tuning). When I finally get back into BIOS the RAM is set at 3444. It's weird. My test bench in the dungeon will be 100% QDC's as soon as the extra 4 from Koolance arrive. I am in complete agreement with you there. I will keep the DICE in mind!

 

P.S. Don't see an intel CPU option in the RAM calc, what setting should I use?

 

I love those Koolance QDs  , even for daily / non-benching builds.

 

On the Ryzen RAM calculator, you won't find an Intel option but it doesn't really matter...just choose the fastest / youngest Ryzen in the top left to 'approximate' your 9900K (most early Ryzens were not very good at mem oc'ers, later ones are much better). That should get you close.  As mentioned, the Samsung B-die kits I have (GSkill GTZR) were originally marketed for Intel Skylake anyways.

 

I have 3866 nominal and 4000 nominal speed kits of that. The former made it all the way up to 4266 - which was really my best Intel Skylake CPU's IMC top speed at reasonable VCCSA & VCCIO voltages; no idea how fast the 4000 nominal kit really is. Alternatively, they would also go to 14-14-14 at up to 3600 while being undervolted. That flexibility is really worth a lot when doing multiple HWBot benches because some benches like tighter timings, others react better to higher bandwidth.

 

That black screen / reboot / reset you mentioned is weird indeed. It does that even on 'stock' XMP ? Should be interesting what your new 4000 kit does on stock XMP. BTW, I always maxed cache / ring, especially for Cinebench and the like - on the other hand, that could also lead to boot trouble or automatic 'F5' resets if it went past a certain stability threshold.

[Avacado]

I don't know. XMP 3600 runs fine. Have even managed to see 3800 in BIOS, but just loops. Apparently my board doesn't do well with MEM OC's at 5.3/5.4 GHz core.

 

Took the top 3070Ti spot though for 9900k/3070Ti. 

 

 

[The Pook]

is your board the Pro Wifi or just the Pro? Try the modified BIOS, I was stuck at 3733 on 4x8 on my Aorus Ultra but the modified BIOS let me run 4133. 

 

Pro Wifi: 

 

OneDrive

ONEDRIVE.LIVE.COM

 

having a hard time finding the non-Wifi BIOS (they might be interchangeable but idk) but if you feel like digging it's probably in here somewhere.

 

[Avacado]

16 hours ago, The Pook said:

is your board the Pro Wifi or just the Pro? Try the modified BIOS, I was stuck at 3733 on 4x8 on my Aorus Ultra but the modified BIOS let me run 4133. 

 

Pro Wifi: 

 

OneDrive

ONEDRIVE.LIVE.COM

 

having a hard time finding the non-Wifi BIOS (they might be interchangeable but idk) but if you feel like digging it's probably in here somewhere.

 

Someone was able to link me the non-wifi bios, thanks POOKIE DOOKIE!

 

 

[Avacado]

[J7SC_Orion]

Nice !  BTW, has your DDR4 4000 arrived yet from Amazon? Specs ? Issues w/ reboot, per above ?

[The Pook]

1 hour ago, Avacado said:

Someone was able to link me the non-wifi bios, thanks POOKIE DOOKIE!

 

 

 

guess you're putting "SPONSORED BY POOKIE DOOKIE" in all your HWBOT submissions from now on

 

[Avacado]

1 hour ago, J7SC_Orion said:

Nice !  BTW, has your DDR4 4000 arrived yet from Amazon? Specs ? Issues w/ reboot, per above ?

Should be here today. It's only 2x8 GB, but it will be fun to play around with for $80 bucks. Even with the modded bios, I still have not been able to get higher than XMP 3600 stable. I know Giga boards are T structure, but I will try the 2 stick anyway. 

[Avacado]

1 hour ago, The Pook said:

 

guess you're putting "SPONSORED BY POOKIE DOOKIE" in all your HWBOT submissions from now on

 

Yes, I might just do that. 

[Avacado]

RAM came it's 4133. Loved Mushkin back in the day. Going to try it again.

[The Pook]

what IC? 19-23-23 doesn't look like B-Die

[Avacado]

8 minutes ago, The Pook said:

what IC? 19-23-23 doesn't look like B-Die

Oh it's not. Doesn't matter. Board sees xmp but won't load, just shuts off and goes back to bios. Same * as with the 3600. I might have to try another board.

 

Really don't need bdie. won't be taking it higher. I bought these to test if it was the board, and it is. 

 

I've got an ASUS z390-a I can try out.

Edited December 5, 2021 by Avacado