i could run b-die @ 1933 mhz on maximus viii hero

[ksateaaa23]

this b die was bought long back. but one problem i have is data corruption beyond 1800. 

1700@12-13-13-28-220-2t
1750@12-14-14-28-220-2t
1780@12-15-15-28-220-2t

around 1.75-1.8 v

it was never stable beyond 1800

but now i was able to run it

1850@12-15-15-28-374-2t

ran x265 1080p, 4k and few 3d benchs. worked perfectecly.  still can't run xtu. data corruption. pfn list is corrupt.................

1933@15-15-15-28-374-2t

ran super pi 32m successfully couple of times with almost idential timings with i7 6700k @ 5ghz. result is not impressive but it could finish the bench successfully.

any suggestions to tweak it further on maximus viii hero. ( officially board can run upto 1900 memory only) though i did 2000 single channel with loose timings)

[ksateaaa23]

Dec 26, 2017 7:47:42 GMT -5 ksateaaa23 said:

this b die was bought long back. but one problem i have is data corruption beyond 1800. 

1700@12-13-13-28-220-2t
1750@12-14-14-28-220-2t
1780@12-15-15-28-220-2t

around 1.75-1.8 v

it was never stable beyond 1800

but now i was able to run it

1850@12-15-15-28-374-2t

ran x265 1080p, 4k and few 3d benchs. worked perfectecly.  still can't run xtu. data corruption. pfn list is corrupt.................

1933@15-15-15-28-374-2t

ran super pi 32m successfully couple of times with almost identical timings with i7 6700k @ 5ghz. result is not impressive but it could finish the bench successfully.

any suggestions to tweak it further on maximus viii hero. ( officially board can run upto 1900 memory only) though i did 2000 single channel with loose timings)

 

  

Attachments:

[Macsbeach98]

I never had a lot of luck clocking memory in the Maximus VIII hero 1866 was about the fastest I could get it to go/

[ksateaaa23]

1945 @ 15-15-15-28-374-2t (1.8v) dual channel

[Shadyreaper]

Dec 26, 2017 8:28:03 GMT -5 ksateaaa23 said:

1945 @ 15-15-15-28-374-2t (1.8v) dual channel

so is the issue it wont post or it wont boot into windows at above that speed? also have you tried 1.9v? mine same mem take 1.9v in MOCF which is like 1.95v real since it OVs also I notice you dont have your maxmem set to boot into win at 4000 and above ya need to limit down to like 2GB and try that 

[ksateaaa23]

Dec 26, 2017 11:42:22 GMT -5 Shadyreaper said:

Dec 26, 2017 8:28:03 GMT -5 ksateaaa23 said:

1945 @ 15-15-15-28-374-2t (1.8v) dual channel

so is the issue it wont post or it wont boot into windows at above that speed? also have you tried 1.9v? mine same mem take 1.9v in MOCF which is like 1.95v real since it OVs also I notice you dont have your maxmem set to boot into win at 4000 and above ya need to limit down to like 2GB and try that 

it did boot and ran super pi 1m successfully . then tried spi 32m. in 6th iteration i got not convergent in square error.
tried geekbench 3. ran for a min. and almost finished but crashed error... pfn list corrupt.

will try upto 1.85v to see it gets stabilised. can't risk too much because it's the best ram for my am4 platform. does 1780 @ 14-15-14-38-56-1t. none of the other rams are completely compatible with am4. also let me see 2gb in maxmem helps.

[Shadyreaper]

Dec 26, 2017 14:07:27 GMT -5 ksateaaa23 said:

Dec 26, 2017 11:42:22 GMT -5 Shadyreaper said:

so is the issue it wont post or it wont boot into windows at above that speed? also have you tried 1.9v? mine same mem take 1.9v in MOCF which is like 1.95v real since it OVs also I notice you dont have your maxmem set to boot into win at 4000 and above ya need to limit down to like 2GB and try that 

it did boot and ran super pi 1m successfully . then tried spi 32m. in 6th iteration i got not convergent in square error.
tried geekbench 3. ran for a min. and almost finished but crashed error... pfn list corrupt.

will try upto 1.85v to see it gets stabilised. can't risk too much because it's the best ram for my am4 platform. does 1780 @ 14-15-14-38-56-1t. none of the other rams are completely compatible with am4. also let me see 2gb in maxmem helps.

man I have put 2v MOCF which is like 2.5v to mine with no heatsinks on them wazza testing thinking it is IMC just needs a little cold on it gonna see what happens with them when I get my 7350k later this week but yea I know how ya feel not wanting to put a bunch of volts to them

[ksateaaa23]

received GALAX HOF ddr4 4000. lot of drama. actually it was held up in customs for three weeks. then i called up customs and they asked cost, i said it is around 247 usd. and accordingly informed him that 19000 inr. he said 42% will be charged as customs duty 7995 inr . i said fine. i have also sent official mail regarding this to customs. don't know what happened they didn't charged anything. released from customs and sent to hyderabad nsh and no tracking for a week. have to go to air port post office, talk to mumbai foreign post office and given the post bag no., date and flight no. so that they could trace it in hyderabad. finally got it with out custom duty.

z370 ultra gaming is having lot of problems with this ram. first day  i could get only single channel working. working only upto 4000mhz and dual channel is a struggle. best i could do is 3650 on frequency & 3550 12-12-12-28-240-2t (booting but unstable during benchs). have to figure it out (1.75 to 1.85v). seems like a struggle on low end boards. 12-15-15-28-374-2t seems stable dual channel up to 3500+.

TIME TO GET Z390 BOARD. 

Attachments:

[Mr.Scott]

Pretty tight customs there. What a hassle.

[georgekokovinis]

I must admit that I am extremely surprised by these results.

Any decent B-die mem set will run at :

1) Quad slot boards ( Hero VIII for example ) at 3866 / 12-11-11-28-1T-220 with very tight seconds and thirds with 1.2V VccIO and VccSA and
at most 1.85V on the mems.
Alternatively it can go 4000 / 12-12-12-28-2T-240 with not so tight seconds and thirds and no more than 2.00V on mems.

2) Dual slot boards ( Apex IX, Apex X ) can easily do 4133 / 12-11-11-28-240-1T with good IMC and good mem sticks at 1.25VccIO and 1,25VccSa
and 2.02V up to 2,05V on mems.

Waza stable too.

The culprit is to know how to tune seconds and thirds and RTL-IOL.
Provided that you have set maxmem in windows at 4096 max as permanent, hours and hours of benching is possible without issues.

[georgekokovinis]

I can run 4000 12-12-12-28-280-1T on Apex VI quad channel with Galax Hall of Fame 2017 with 1.05VccIO and 0.95VccSA at 1.90V any benchmark.
QUAD CHANNEL X299.

[ksateaaa23]

Oct 26, 2018 18:16:08 GMT -5 georgekokovinis said:

I must admit that I am extremely surprised by these results.

Any decent B-die mem set will run at :

1) Quad slot boards ( Hero VIII for example ) at 3866 / 12-11-11-28-1T-220 with very tight seconds and thirds with 1.2V VccIO and VccSA and
at most 1.85V on the mems.
Alternatively it can go 4000 / 12-12-12-28-2T-240 with not so tight seconds and thirds and no more than 2.00V on mems.

2) Dual slot boards ( Apex IX, Apex X ) can easily do 4133 / 12-11-11-28-240-1T with good IMC and good mem sticks at 1.25VccIO and 1,25VccSa
and 2.02V up to 2,05V on mems.

Waza stable too.

The culprit is to know how to tune seconds and thirds and RTL-IOL.
Provided that you have set maxmem in windows at 4096 max as permanent, hours and hours of benching is possible without issues.

admire your knowledge but would like you to analyse few more facts before jumping to conclusion. not all boards will do 12-12-12-28-220-1t/2t 2000+. need better binned b-dies and cpus with very strong imcs.


1. this is an ordinary b-die bought from tapakah/sam ocx(he is our forum member now). got this vey economically for 140 USD. he could do only 12-13-13-220-2t 2000mhz spi 32m but not waza stable, on MAXIMUS VIII IMPACT. he is definitely a skillful overclocker and a good memory tweaker.


2. MAXIMUS VIII HERO is officially listed at 1866 and after a year only they added support upto 1933 and it's not a board you can run 2000mhz consistently.


3. You had MAXIMUS VIII EXTREME which is a much better board than hero. viii hero is a very ordinary board for memory overclocking for your information. adding subs above 1933 mhz for your information. not even 10 people submitted beyond 1933 and total 13 subs. i was the only guy who was consistent doing around 1975-2000.


it's a different ball game when you are playing with lesser hardware.

moreover this viii hero is my best board. other boards are just about 100-150 usd entry level boards. stil i compete.

[ozz]

um.... W9 mathematician ^^^^, our K knows his figures, and can put it to practice, i dunno how he gets clocks like he does on the low voltages he uses, ive talked with pete b4 about this, but he does it

[georgekokovinis]

The purpose of posting is definitely not the creation of " internal competitiveness " or any other kind of unwanted bad impressions.

Thought to share my experience with hundreds and hundreds of hours of trial and error testing - always on air - of DDR4 B-die ( earlier
and later versions ) of this indeed tricky and difficult to stabilize memory platform.

So, please stay with me for a while, while I try to share ( only for good purpose ) my findings only on Intel platforms and not AMD, of which
my experience goes down to almost zero

B-die is a tricky sucker.
No set is same with the next and adjustments are always due in order to make them work at acceptable levels.

OK. First up. Quality of HW and a lot of luck.

Ksa is 110% correct to point out that a 100 USD board - and indeed a 4 mem slots board - is not possible to perform mem wise like an ultra
expensive Apex X or Apex XI or Apex XI.

So, in order to keep things in check and for the sake of wide spread comparison we will take two very popular and cheap boards " to work with ".

Hero VIII and Impact VIII. Both boards can be sourced currently for 100 USD or even less.

IMC is the most decisive factor for the final outcome of mem tuning. A good IMC will provide good to very good results, a bad IMC will bury
even the best mem set and an exceptional IMC will provide exceptional results.

1) Hero VIII. A board with low limits in mem tweaking due to its design. Testing over 30 cpu's on it, I found only five which gave me the basis to proceed with further mem tweaking.
Unfortunately all the rest were limited to 3733 /12-12 or at best 3866 / 12-12-2T.

I had the opportunity ( and luxury ) to buy and resell more than 200 b-die mem sets - G.Skill and Galax.

Out of the pool of 200+ mem sets, I have kept 22 sticks. And have created dual and quad channel sets which IMHO give me the best possible
results when working with the rest of my HW.

2) Impact VIII. This little sucker has true potential. It is cheap, and given a good IMC ( not extra-ordinary ) it can fairly easily do
4000 / 12-11-11-28-220-1T tight with max 2.00V.

Waza is a totally different story. Mem settings for waza are entirely different than for XTU and Geekbench 3.

Third timings and RTL-IOL - regardless the board-cpu- mem set are the field where a lot of work must be done to reach a stable overall setting.

Notorious benchers like Shamino, Frediyama, Alex@ro, Dancop, Sam ( Tapakah ), Firekiller, Roberto, Allan and many more have one common saying - Can't tune thirds ? Quit and move to the next set.

There are more than fifty combinations of settings that someone stubborn and persistent can try in order to find the optimum setting for the given set of HW.

Needs time, lots of time, lots of advice and a pool of HW to work with.

Apropos, since it was mentioned, I never liked Maximus VIII extreme for mems. Prefer Impact.

I am with you KSA. Most ( 99% ) of people benching, are limited by one factor or more in order to reach the optimum.

So, let's leave it here for now, since books and books and ultra extensive articles have been written on DDR4 and not only.

Hobby wise, your results are good, especially on AM4, which is a difficult platform ( and limited - very limited ) for mem tweaking.

I would be extremely happy and obliged if Bullant chimed in the discussion, with his endless knowledge of mem tweaking.

All the best to all

[ksateaaa23]

I will take this.

"1) Hero VIII. A board with low limits in mem tweaking due to its design. Testing over 30 cpu's on it, I found only five which gave me the basis to proceed with further mem tweaking.
Unfortunately all the rest were limited to 3733 /12-12 or at best 3866 / 12-12-2T."

but george didn't seen a single sub on hero viii or couldn't see it in your hardware library. i can see viii extreme, impact and the two ocfs..

VIII IMPACT definitely is the second best to OCFs that era(3 years back but feels like lot of time)

thanks about AM4, hardly have any cpus there. just ryzen 7 1700 and nothing good there. must have done well on intel as well, otherwise might not have many points at hwbot.


even ocf is now available @120 usd at newegg. when i was talking about cost, talking about it when they were launched and not today.

i will say again playing with lesser hardware is a different ball game than having the best of everything and talking big.

i do my best with whatever my limited available hardware.

not keen on big lectures. in my understanding this game is more about money, some knowledge & skill.

i just post here out of enthusiasm and nothing more. not really interested to listen to lectures.

[Mr.Scott]

While on the Hero VIII subject, I'm interested to know what bios's you guys consider the best.

[ksateaaa23]

Oct 27, 2018 7:59:06 GMT -5 Mr.Scott said:

While on the Hero VIII subject, I'm interested to know what bios's you guys consider the best.

1302, 1601 in my experience. 0014 xp bios is also good.

[Mr.Scott]

I will try them. Thank you.
I've been running 0040 xp and 1001.

[georgekokovinis]

Oct 27, 2018 7:15:27 GMT -5 ksateaaa23 said:

I will take this.

"1) Hero VIII. A board with low limits in mem tweaking due to its design. Testing over 30 cpu's on it, I found only five which gave me the basis to proceed with further mem tweaking.
Unfortunately all the rest were limited to 3733 /12-12 or at best 3866 / 12-12-2T."

but george didn't seen a single sub on hero viii or couldn't see it in your hardware library. i can see viii extreme, impact and the two ocfs..

VIII IMPACT definitely is the second best to OCFs that era(3 years back but feels like lot of time)

thanks about AM4, hardly have any cpus there. just ryzen 7 1700 and nothing good there. must have done well on intel as well, otherwise might not have many points at hwbot.


even ocf is now available @120 usd at newegg. when i was talking about cost, talking about it when they were launched and not today.

i will say again playing with lesser hardware is a different ball game than having the best of everything and talking big.

i do my best with whatever my limited available hardware.

not keen on big lectures. in my understanding this game is more about money, some knowledge & skill.

i just post here out of enthusiasm and nothing more. not really interested to listen to lectures.

I am sorry to see such a reaction.

1) It is not a lecture. Tried with some brief words to share a vast experience I have acquired on DDR4.

2) You do not see Hero VIII because I never bothered to submit results with that board. I may make 100 runs and submit 5 that are worthwhile. None on Hero VIII was worthwhile submitting.

Anyway, to each his own. Not worth bothering again to share knowledge acquired thru hard effort, reading and discussing with TOP overclockers asking : " I do not understand - please explain ".

[ShrimpBrime]

He looked to find your subs while visual verification helps with understanding. Of course seeing is believing, and K may have simply misunderstood the direction you wanted to take the conversation. Happens a lot on the internet while we can't hear tone of voice and see facial expressions, we can only imagine feelings and reactions by what we read and that's what makes books interesting but movies straight to the point.

I personally find this thread a very good read. Please don't not share knowledge based solely on the plain words you read which may be taken out of context in some cases. Misunderstanding is human nature. I'll be honest, I wouldn't know some of the stuff if I didn't get to read the shared knowledge.

[ksateaaa23]

Oct 27, 2018 9:33:45 GMT -5 ShrimpBrime said:

He looked to find your subs while visual verification helps with understanding. Of course seeing is believing, and K may have simply misunderstood the direction you wanted to take the conversation. Happens a lot on the internet while we can't hear tone of voice and see facial expressions, we can only imagine feelings and reactions by what we read and that's what makes books interesting but movies straight to the point.

I personally find this thread a very good read. Please don't not share knowledge based solely on the plain words you read which may be taken out of context in some cases. Misunderstanding is human nature. I'll be honest, I wouldn't know some of the stuff if I didn't get to read the shared knowledge.

seeing is believing. one management guru Edward Deming say that "In God we trust, all others should bring data". i believe in that. 

[ksateaaa23]

first you said 4000 possible

"1) Quad slot boards ( Hero VIII for example ) at 3866 / 12-11-11-28-1T-220 with very tight seconds and thirds with 1.2V VccIO and VccSA and
at most 1.85V on the mems.
Alternatively it can go 4000 / 12-12-12-28-2T-240 with not so tight seconds and thirds and no more than 2.00V on mem"

then you said 3866 at best

"1) Hero VIII. A board with low limits in mem tweaking due to its design. Testing over 30 cpu's on it, I found only five which gave me the basis to proceed with further mem tweaking.
Unfortunately all the rest were limited to 3733 /12-12 or at best 3866 / 12-12-2T."

now you say that none of the subs on hero are worth submitting.

"You do not see Hero VIII because I never bothered to submit results with that board. I may make 100 runs and submit 5 that are worthwhile. None on Hero VIII was worthwhile submitting."

[ksateaaa23]

your knowledge is exceptional but may not be useful on my boards. i haven't asked you to share your knowledge, instead you could have started a new thread and shared your enormous knowledge.

i have made 1112 submissions to hwbot on maximus viii hero, because i thought it was best board for me. definitely talking big things are irrelavant to me when i am using low end boards. that's the only point i want to make here.

let me give details of my ddr4 boards.

1. maximus viii hero

2. z170-d3h

3. ax370-gaming k5

4. z370 ultra gaming

[ksateaaa23]

"Hobby wise, your results are good, especially on AM4, which is a difficult platform ( and limited - very limited ) for mem tweaking."

this statement was just given to say that my results are good is indirectly to say that (hardly i have decent subs with am4 platform ) you don't know anything about memory tweaking.

you are saying you do well on platforms which doesn't require mem tweaking. such statements at best can be avoided. keep your knowledge with you who cares. if needed i will learn from whom i want to...............

on my hero, i have run 1975-2000 at least 5+ times. passed geekbench 3 with memory @1945 16-20-20-45

hwbot.org/submission/3295398_ksateaaa23_geekbench3___multi_core_core_i3_6320_11376_points/

on z170-d3h rated at 1733, ran memory at 1866 12-15-15-28-374-2t and could do 1902.8mhz

hwbot.org/submission/3842077_ksateaaa23_memory_frequency_ddr4_sdram_1902.8_mhz

on ax370-gaming k5 was able to what was expected 14-15-15-36-1t upto 1785

on z370 ultra gaming could do 2149 when the board was rated at 2000. tight timings 12-15-15-28-2t upto 1866.

hwbot.org/submission/3875570_ksateaaa23_memory_frequency_ddr4_sdram_2149.5_mhz

i can only do as per my hardware. they are fairly good on those mediocre boards.

B-DIE BINNING IS NOT THE ONLY THING FOR OVERCLOCKING. IT MAY BE YOUR SUBJECT OF INTEREST BUT NOT MINE.(TOO EXPENSIVE FOR ME)

YOU HAVE GIVEN A BIG LIST OF OVERCLOCKERS. I AM A BIG FAN OF MACSBEACH98. HE IS AN ENCYCLOPAEDIA AS FAR AS OVERCLOCKING IS CONCERNED. HAVE GREAT REGARDS FOR THE DEPTH OF KNOWLEDGE HE HAS. ONCE SHANE JR.(AVALANCHE) SAID HE IS THE BEST OVERCLOCKER IN THE WORLD AND I SECOND THAT.

[georgekokovinis]

Truly unbelievable !!!

Wow. You do not have to read or believe mate what I write
Just ignore me and move on.

[Macsbeach98]

Oct 27, 2018 8:16:55 GMT -5 Mr.Scott said:

I will try them. Thank you.
I've been running 0040 xp and 1001.

0001 for baseclocking locked CPUs

[Mr.Scott]

Oct 27, 2018 16:55:07 GMT -5 Macsbeach98 said:

Oct 27, 2018 8:16:55 GMT -5 Mr.Scott said:

I will try them. Thank you.
I've been running 0040 xp and 1001.

0001 for baseclocking locked CPUs

Thank you. I have it. You told me that before.
I listen.

[ozz]

Oct 27, 2018 11:06:34 GMT -5 ksateaaa23 said:

YOU HAVE GIVEN A BIG LIST OF OVERCLOCKERS. I AM A BIG FAN OF MACSBEACH98. HE IS AN ENCYCLOPAEDIA AS FAR AS OVERCLOCKING IS CONCERNED. HAVE GREAT REGARDS FOR THE DEPTH OF KNOWLEDGE HE HAS. ONCE SHANE JR.(AVALANCHE) SAID HE IS THE BEST OVERCLOCKER IN THE WORLD AND I SECOND THAT.

and i 3rd it, ive seen it in real time many times, im the only person here thats had the priviledge of seeing him in action , hes forgotten more than most know and he doesnt forget easily, doesnt matter what board, what ram, what card or cpu he just knows, hes a very talented man in all facets and im not saying this coz hes my friend and mentor , its just plain truth

[Macsbeach98]

K I am chuffed that think that Thankyou

But I am just an average overclocker that has most desktop sockets from socket 7 on and use LN2 to get fast speed out of them.
It makes me a good all rounder having all the different sockets.
A jack of all but a master of none.

[ksateaaa23]

Oct 28, 2018 0:21:33 GMT -5 Macsbeach98 said:

K I am chuffed that think that Thankyou

But I am just an average overclocker that has most desktop sockets from socket 7 on and use LN2 to get fast speed out of them.
It makes me a good all rounder having all the different sockets.
A jack of all but a master of none.

simplicity and humbleness of great man.