Hi MateQuote:
Originally Posted by chew*
Can you link to them I cant find them?
Printable View
Hi MateQuote:
Originally Posted by chew*
Can you link to them I cant find them?
Quote:
Originally Posted by techtrancer
What are these results based on? Cheers
That's silicon lottery results from binning.Quote:
Originally Posted by easyrider
Its only on page 1 like first result in review ;) look harderQuote:
Originally Posted by easyrider
Prime 95 2hours in temp readings, I think now you can start to see one of the limitations. Oddly enough heat out of the heatsink feels like 40-45c. My 290x's in crossfire on water dump 50C loaded after a few hours. I'm very familiar with what the temp feels like. This feels cooler.
http://i1253.photobucket.com/albums/...psqawjaewu.jpg
Highest ive seen 77C if testing 24/7 stability good cooling good paste good mount.
Ive seen much worse bad mount and or excessive voltage.
Did 4ghz require a big bump in vcore then?Quote:
Originally Posted by chew*
Quote:
Originally Posted by chew*
Linky? Forgive me if this is a stupid question. :)
preety much will most likely for all chips.Quote:
Originally Posted by easyrider
I would not exceed 1.50v for 24/7 stability testing. I personally don't feel the gains are worth it plus the chips just really don't like it.
I'm noticing that the l3 cache write speeds in aida64 are up and down alot with what I'm seeing with the pics you guys are posting.
I'm not sure if there's something to that or not.
Also, I don't know if you guys heard, but there's supposedly some kind of sleep bug going on with these boards.
The one dude that found it, has his rig at 4ghz, after sleep and wakeup, it changes to 4.2ghz and "seems" stable.
I wonder what would happen with a much higher clock on ln2, maybe possible to get 5.5ghz?
Or dead parts lol, who knows...
One more thing, if there is truly a rom of sorts on the cpu...
Getting it super cold could cause the chip to keep data when it's supposed to be changed.
The only reason I know is because I saw some goofy video some years back on spying, people would take out the ram and put it in cold ice or something, and replace the ram with there own, so that they could grab the mem contents.
Kinda of a weird thing but just saying, rom's and ram, when they get cold, they retain data to the extreme I guess.
So if for example, you bsod and have to reset the cmos on a ln2 session, the reset might not apply correctly.
I don't know for sure, just saying this could be a possibility.
The l3 is buggy in version we are using. Only l3 is fixed in new version. Non issues. Does not effect read write copy. Plus we have tested same version.
We are both on asus. Sleep bug non existant not to mention hpet is on and in performance mode on sleep does not happen.
@Chew - you are a bad man :). I had put a Ryzen build off for a couple of months but have found pricing for the B350 to be practically irresistible @ roughly €100, €375 for a Rz1700 and some memory and I can start playing at under €600 + I have a couple of Sabertooth (R2 and R3) and a couple of FX's to that €600 to bring it down even further.
Interesting to note that Asus have a message on the spec page for the board reading like this
"1*Due to the limitation of AMD Ryzen™ processors, memory frequencies higher than DDR4 2666 MHz are supported only when using one DIMM."
Looks like you have blown that away with the TridentZ you are using.
Read the qvl for the board. Quite a few sets of over 3200 work even shows dual channel. Im betting thats a CYOA move as it seems some cpus imc are not great.Quote:
Originally Posted by robbicfc
Its preety odd to say the least but when shopping for low end x370 i was seeing not so good qvl listings.
2933 at best.
3 b350 boards in 3 different brands qvl 3200 np
People with contact at MB makers should make a case with mb manufacturers, for a B350 mobo with some quality components ,good PWM and more sata ports.Quote:
Originally Posted by chew*
I dont need leds or fancy looking stuff, i also dont really need x370 or sli .A mobo which you could recommend to anyone not doing world records on phase or Ln2 but with solid Oc on air and water.As ryzen is multimedia oriented, there should be 8/10 sata ports 2m2, dts .This kind of mobo would sell well .Wonder if there are more models in the making, or for what we see is what we get.
I can try and ask but obviously the company will want a profitable product.
A beefy pwm simple short and sweet b350 would probably be easy enough but it would put its cost closer to low end x370.
I kinda have one upset with me right now so i think they are out of the question for the time being...
Its not easy being brutally honest, someones bound to take offense but thats just me being me.
How they choose to interpret it can be critical. They can take it offensively or constructively.
The latter of which can only bring better products to us.
Checked memory compatibility on ASUS b350-plus today.
As I suspected QVL being biased towards Samsung has its costs......Avoid Hynix based Ic's on this board.
Luckily the Samsung based IC's offer better timings @ better speeds anyway so really not a big deal unless your trying to be super cheap about using this board.
Fyi there is a clue as to what happened. I'm going to look into it further. PS/2 drop is to much of a coincidence.
Laymans terms PS/2 operates off 5v and ground.
Time to get technical :D
https://www.youtube.com/watch?v=DSRuV1g7X18
EDIT Ok so confirmed 5v and ground on ps2 so that was a dead end. Tired dig deeper after some sleep.
Ok insomnia Wins but made some progress. Found a hot spot.
The culprit is in this area. I'll poke around with the dmm tomorrow.
http://i1253.photobucket.com/albums/...pshgpt8pcp.jpg
http://i1253.photobucket.com/albums/...psywz6ng9w.jpg
Ok, I now know how this aida bug happens.
Here I've booted with 22.5 multiplier and then raised in Windows.
http://www.xtremesystems.org/forums/...1&d=1489782875
On the other hand, I can share info about 4x8GB B-die (the screenshot above is also with 4 slots populated).
On Asrock AB350 Fatal1ty K4 I couldn't boot above 2400 divider.
On Crosshair it was possible to boot 2666 with standard dividers.
Raising bclk I could boot 3133MHz with lower dividers. To get 3200 I had to raise it in Windows.
Can't set lower than CAS14 with 4 slots populated.
RAM is 2x8GB 3600C17 and 2x8GB 3733C17.
Yep i had charged do via same method. Seems msi is not affected like the asus by that bug.
In other news i preety much pinned down the issue with my gaming 5. I want to give them sometime to look into it however.
I have a theory on how as well but lets see what they say first.
Im preety confident it would be impossible to hard mod mine back to life or i would have done it already. :D
So I have two questions for those in the know.....:
1. Am I understanding it correctly that as we pack the boards with more memory the maximum speed at which we can run the memory decreases? I mean, generally speaking....
I'm am still cautious about a new purchase since I actually have to make money off of the next machine I get, and this is an issue that might concern me. I'd ideally want no less than 8GB per stick and at least 16GB to start. But now I'm sort of fearing that I'd be stuck at significantly lower speeds if I eventually move to 32GB RAM.
2. I get that the fabric that connects the CCXs is tied to memory speed and also PCIe. What I'm still a bit uncertain of is under just what condition PCIe 3.0 throttles down to 2.0. (?)
If anyone could either briefly explain it or point me in a good direction I'd be happy...
Ok infra you got 4x4 covered. I will put in work with double sided B die 2x16g.
Got a good deal on these and since this seems to be a bigger topic than I expected......
Keep in mind this is a little B350 so....
No post 2933, no post 3200.
No post 2666 14-14-14
No post 2666 12-12-12
But at least i'm maxxing out AMDs recently announced spec for this config. Anyway i'm looking into it Mattias but as I feared double sided will be a tad troublesome.
http://i1253.photobucket.com/albums/...psd3xijyzy.jpg
Question: are you doing all your RAM tweaking in UEFI, or are you using Ryzen Master?
Bios currently,
Ok so there is room for improvement in bios updates.
This one is rather flaky USB wise (beta) but there is a performance gain so they did not just slack something to run tighter.
Still can't run higher, that could be cpu specific however.
2933 16-16-16 :down: DS 32g
2666 12-12-12 :down: DS 32g
2666 14-14-14 :up: DS 32g
http://i1253.photobucket.com/albums/...psldcvq7kf.jpg
Has anyone seen this? Prime95 for Ryzen.
https://www.reddit.com/r/Amd/comment...ed_link_fixed/
Thx for the heads up. Maybe i can get it to trigger imc/memory errors now instead of just bsods.
Some interesting RAM clocks :) Things are getting better...
3733 is the maximum I can get at the moment, regardless of timings. Mem has XMP at 3733, not sure if there's any connection.
Going to try a different multiplier soon.
https://i.imgur.com/GLkyafC.jpg
https://i.imgur.com/kD7RCWM.png
That is the highest I was able to boot into windows at with my 3600 kit (albeit at looser timings than yours), I was not stable for usage so didn't bother posting when I did my run through. Have you had a chance to 32m those settings? I am curious if my chip sucked and what BIOS are you on?Quote:
Originally Posted by I.nfraR.ed
That's impressive. Are you running any SATA devices at that bclk? What did you have to tweak to get it running that high?
Bios 1001. Running SATA3 SSD (Adata S511).
Turned off everything I could in bios and also had to set maxmem in windows (Just like on Kaby), otherwise it was unstable.
VGA is proven to work with really high PCI-E clocks, so no problem here. I can max out PCI-E on my 790FXB-M2RSH and the card is still stable.
This is a "known" fact that 8xxx nvidia series handle this very well. Not sure about other series, haven't tested much when there's a proven solution.
On the other hand I could not find any single secondary voltage that helps with mem clocks.
I didn't check 32M, only ran R15 and aida64.
Things are definitely getting better, Asus being ahead of competition. Hopefully it is possible to add higher mem dividers and unlock the subtimings, so we can tweak on low bclk as well.
My ch6 is on the way. See if i can duplicate this with DR sticks. Being dual rank im not expecting as high clocks....rated speed @ rated voltage would be nice.
@ I.nfraR.ed
Realistically (stable) speaking , how far someone with 4x8gb B-die can expect to get ? Freq wise and Cas wise ?
@3000 Cas 14 may be doable ? Or even 3200 ?
Hey infra can you do me a favor. No ref clock. Baby steps and boot. Try 14-14-14-32 default ram speed. Reboot then 2666. Reboot baby step down 14-12-12-12 reboot then 14-11-11-28. Keep stock volts. Once it posts reliably baby stepping down to those timings set your cpu clocks and voltages. Lmk how you make out.
Ok so I did some investigating.
I am NOT CLAIMING STABILITY yet. I am just merely testing the limits of the IMC with Dual Rank dimms.
http://i1253.photobucket.com/albums/...pshoryydcv.jpg
http://i1253.photobucket.com/albums/...ps5tx7yhti.jpg
http://i1253.photobucket.com/albums/...psacmgk1id.jpg
http://i1253.photobucket.com/albums/...psp44ntbim.jpg
As we can see in this screenshot I have either found the limit of ram @ 1.35v or a IMC limitation.
http://i1253.photobucket.com/albums/...pspduysuzz.jpg
I got a new board in. Turns out I still hate it. Even using 1001 for a BIOS it is still junk IMO. I have never seen shenanigans like this even with ES chips on prerelease AM3 boards.
Ill trade you a slightly alive gaming 5 :DQuote:
Originally Posted by cdawall
I am up to two dead CHVI's lolQuote:
Originally Posted by chew*
I love my Tomahawk! :P
:up:Quote:
Originally Posted by charged3800z24
Isn't that one of the ones that can't handle full wattage of an overclocked chip?
Just about any B350 is weak overclocked but I can kick peoples azz on it with memory tuning, fun stuff.Quote:
Originally Posted by cdawall
Ok here goes Its ugly cas latency but working is better than not working.
ONCE AGAIN IMC EXPLORATION ONLY. NOT CLAIMING STABILITY
There appears to be a CL hole between 14-17 not to mention odds are not even applying with certain dividers currently.
I see no reason why these sticks can not do 2933 at cas 14-17 other than a hole.
I skipped 18-16-16 testing as I saw no point but I can revisit it later.
http://www.overclock.net/content/type/61/id/2989937/
http://www.overclock.net/content/type/61/id/2989938/
http://www.overclock.net/content/type/61/id/2989939/
once again 18-10-10 was problematic.
Based on these findings and the above findings @ 2600 speeds I had a funny feeling that AMD killed the 10 subtiming support lvl.
I had a set of corsair platinums that after recent agesa update in bios refused to load at XMP 10-12-12-28 @ 2400. Hmmmmm...
I tested the theory loaded defaults shoved them in set timings 12-12-12-28. Boom posted right up just fine.
http://www.overclock.net/content/type/61/id/2989940/
Next up on the testing is to run 32m PI again at the best 2666 timings and the best 2933 timings to see if raw speed plus fabric beats better timings with less speed and less fabric.
Oh and pc3200 with dual rank is on the list when i can figure out how to do it.
I have this funny feeling it will require the cpu slightly sub 0 but i could be wrong.
Nice to see you got the 2933 divider working.
There are many illogical things, certain combinations not working at all, a lot of "holes" in memory overclocking, not just the one in frequency.
I still can't get a certain pattern though. Sometimes retry button on CH6 does wonders after failed training.
I also get the feeling the CPU is "learning", lol. Something that didn't work a day ago, now works.
The memory speed holes are interesting, but it doesn't surprise me. We had holes like this with Piledrivers' base clock. It's a good find, I wouldn't have expected it to show up with memory timings.
Pretty much right now I'm just fooling around with locked ref clock board.
Seeing as many do not have ref clock it is relevant to those with Dual Rank dimms.
I'm fairly certain with a ref clock board I can "get around" certain barriers but am unsure if getting around them will be viable for 24/7 real world usage.
I may need to take a ride this weekend, go visit an old buddy, bring back an old single stage out of retirement.
So that basically confirms what I went through when I was getting my dual rank 2x16g 3200 tridantZ cl14 kit setup on my 1800x. I have a gigabyte gaming 5 mobo. I ended up with the exact same cl timing hole. So anything below cl 18 wouldn't post (I only tried down to 14) but 18 was fine. Kept the timings at 18 16 16 35 instead.
https://cdn.discordapp.com/attachmen...375424/yay.png
Cool you just answered a question for me.Quote:
Originally Posted by starheap
The next step was to find someone with your bin and flash that SPD to my sticks to see if it was somehow XMP related.
Now I don't have to do that.
Another question for you. Did it fail post trying 3200 or did it get by memory checks but have no video?
This board has no port 80 so I went caveman and installed a speaker.
3 beeps memory failure.
1 beep it boots
0 beeps your screwed clear cmos...
With pc3200 I got 1 beep but no video....so I am close but I am missing something.
I am almost 100% positive these timings and holes effect all vendors and is not vendor specific leading me to believe it is definitely in the AGESA and not vendor bios tuning.
Anyone with a "fixed ref clock board" is going to hit these same "walls" and "holes" although a "unicorn IMC" cpu may break this trend.
I just remember it not posting at 3200 at pretty much any timings and I tried some pretty insanely loose timings. I don't remember the exact details atm since I haven't tried 3200 since sometime last week.
Like it wouldn't evenpost at 3200 24 24 24 50 lol
Yah would not surprise me. I think it is an IMC limitation which is why I want to try like -30 to 0 C temps.Quote:
Originally Posted by starheap
If you do get bored and try again stay at same timings.
Watch the 4 leds near port 80. LMK if it gets past the mem check LED and goes to boot.
Mine is definitely passing memory checks.
Interesting I'll have to take a closer look at that when I get home.Quote:
Originally Posted by chew*
I was starting to get the hang of memory tweaking the board yesterday I was able to do some testing with the memory at 3800, but the chip was severely hampered for overclock. I was only able to push it to 3.7-3.8 with the memory pushed that far.Quote:
Originally Posted by chew*
Thats typical when stressing amd imc. The more things change the more they stay the same.
Oh I know I was just posting my findings for the masses.Quote:
Originally Posted by chew*
Do we have control over how much voltage IMC gets? Half the extra voltages on my ASRock don't even have names beyond "1.8v control thing", so I personally have no idea.
I will show a screenshot soon with voltages i used for all testing.
You may be able to correlate names to voltages then. I think this board has some voltages locked out that ch6 has.
I really have not looked to see a screenshot of there voltage page in bios.
Maybe stilt would know better.
I would assume 1.35v into memory goes into cpu imc however.
Seen anything on reviving a Asus C6H board?
Tonight I am set to fire up the 1700+ and hope for 5Ghz with memory speed and timing.
Try soldering a 12v bulb to leads with alligator clips.
Pull battery stick a jumper on clr cmos and the bulb setup on cmos battery terminals. It may speed up process.
Either that or let it sit a day or two unplugged no battery
Ok Here goes timings versus fabric single rank and dual rank.
Single rank 2933
http://i1253.photobucket.com/albums/...psiln8k7gh.jpg
Single rank 3200
http://i1253.photobucket.com/albums/...psjwvl8ad2.jpg
Dual Rank 2666
http://i1253.photobucket.com/albums/...ps4ttbgqiq.jpg
Dual rank 2933
http://i1253.photobucket.com/albums/...ps40miwlja.jpg
Conclusion
Single rank speed + fabric wins
Dual rank Application specific tossup.
I have not and I have had no luck using crashfree reflash, new CPU's or anything else along those lines. It seems once it fails that flash it is dead.Quote:
Originally Posted by FUGGER
As promised.
Bios setting voltages used through all memory testing.
http://i1253.photobucket.com/albums/...psn8tvbjga.png
Should be listed as MEM VDDIO or similar.Quote:
Originally Posted by sdlvx
You will also want to bring MEM VTT up to half your MEM VDDIO value.
Chew i'm glad to see that you got Hynix mem dual rank memory to work at 2933 divider . I have tried many combinations on my Hynix ram for testing purposes. The best i can do is 2666 at 18-18-18-36 2T. I have left Cas alone at 18 as anything lower does not work at all , so that confirms there is some kind of hole at anything below for that , same findings as infrared and you. I cannot get any other timing changed as soon as i start to do baby steps starting to tight latencies it's a no go for me and my ram. Tried 18-17-18-36 did not boot , it hangs. tried 18-18-17-36 same , 18-17-17-36 same and so on ...
Testing last night for a few hours and i end it up with a completely trashed and corrupted windows 10 due to many blue screens. I need to re install windows today to keep on memory testing on Hynix Dual Rank.
http://i67.tinypic.com/52na6g.jpg
Chipsy.
I would not bother trying to tighten up crazy like i did. I was on a "qualification" bios for that 14-10-10 and 18-11-11 stuff. Its quite buggy also dropping devices alot. I did my testing and jumped back off it...its that bad.
I was using it to explore imc to see what the real hard limits of imc is versus what we are locked out of.
The 16-14-14 2666 is public bios as was the 18-16-16 with hard locked timings. Cant go tighter.
All this was B die.
The high bin hynix DR i have are super unstable but at least i know how to get them to post now.
For corrupted OS i suggest bubblegum and paper clips. That is holding mine together :D
Anyway i have been on a trip yet again...gonna take a nap then flash a new bios with hopefully the fix i asked for and see if it got better or worse.
Thank you for the feedback Chew :up: , yep Hynix are very , very unstable and almost impossible to make them run at decent speeds or timings :/ , no worries i bought a full box of bubblegum and paper clips :D. My Sammies B-die can go ~3300Mhz cas 12 no problem whatsoever :yepp:Quote:
Originally Posted by chew*
Regards: Chispy
Solder pads. Left SOC +/- Right VCORE +/-
http://i1253.photobucket.com/albums/...ps0p0ybv7d.jpg
Grounds on right side of pads. Positive on left sides of pads.
http://www.overclock.net/content/type/61/id/2992766/
Soldered to a spare molex extender for ease of use. YMMV.
http://www.overclock.net/content/type/61/id/2992767/
https://www.youtube.com/watch?v=_kWFhO7rnAg
Thx for all the info in this thread very intriguing
Does anybody know if this is Samsung or Hynix it comes up as single rank but can not find what ic brand it is without taking off the HS ..
G.SKILL F4-3200C14D-16GTZ
Just found out here Its Samsung :)
http://www.overclock.net/t/1624603/r...locking-thread
The outcome of knowing your voltage vs guessing. 1.39-1.41 measured vcore, 1.0 SOC
http://www.overclock.net/content/type/61/id/2993344/
Now I need to test some other chips to see what they changed memory wise in 0511 or if this IMC is a dud as I can not get 3200 to not toss errors.
Dont know, but Maybe will help
tRAS=CL + tRCD + tRP
Also:
Write to Read Delay + CAS Lese zu Precharge Zeitspanne = Zeitspanne der Schreibewiederherstellung
https://abload.de/img/speichert2jii.png
Works well on AM3+
Been there done that dropped all timings dropped all voltages raised all voltages. Changed bios.
Conclusion = imc
Why IMC?
2 chips exhibit same behavior.
1 does not.
Is the Chip who does not newer?
http://img.techpowerup.org/170325/18...003200cl16.png
http://img.techpowerup.org/170325/cb...powerlimit.png
Found out the ram does CL14, cinebench scores didn't change with the tighter timings
http://img.techpowerup.org/170325/39...y4100-2100.png
http://img.techpowerup.org/170325/3dmark6748.png
1800x = 2933 fineQuote:
Originally Posted by techtrancer
1700x = 2933 fine
1700 = 3200 fine
I would say...the 3200 capable cpus are not as common.
Im running custom blend 13,312 mb allocated to prime 95...it takes maybe 5 mins tops then you start to see cores start dropping threads randomly. Not the same cores if you repeat.
Tested single stick pass. Tested other stick pass. Explored all voltages up and down. Slacked timings. Tried lowering cpu speed.
Found one particular voltage that gave me a little more stability got me through 20 mins... Then same outcome.
Hm
Maybe the CPU multiplier makes Problems?
The PII X6 have Little Bit Problems with 200x10
200x9 ore 200x11 works better than 200x10
I thought about that to so played with even multis versus .25 .50 .75 etc also tried 39 and 40.
The good thing is i can get it to happen quick so easy to test stuff and move on to the next.
Got my waterblock in yesterday and just got my X370-Pro set up. Couldn't even install windows at 2133 MHz DRAM with the shipping BIOS, but 0511 seems to have fixed that. Doesn't give me great hope for the chip if IMC quality seems to vary as much as chew has seen.
*EDIT* Has AIDA64 been updated to properly report Ryzen 1*00X temperatures? The release notes mention "improved support" for Summit Ridge CPUs and I'm seeing load temps in the mid 40s using its stress test. Seems about right to me.
@chew* so I have been playing with my IMC some. I am not claiming stability on this, but it runs this 32M and 3dmark so far
http://img.techpowerup.org/170326/18...093600cl14.png
1800 @ CL14 :shocked:
Quote:
Originally Posted by cdawall
Yah thats normal for ref clock boards :)
Prime 95 custom blend with 13312 allocated out of 16g shows ryzens weakness though with 3200 divider.
http://i1253.photobucket.com/albums/...psw3fvx9nu.jpg
Thats not throttling...thats rounding errors. Drop to 2933...fine..back up to 3200 not fine.
Sure it wont effect benching but im helping 24/7 guys atm.
I think they will still go lower I am not even pushing voltage yet that was like 1.45v on the ram. They are some good damn sticks. It is actually set to 14-14-14, but the bios didn't save correctly yet again.Quote:
Originally Posted by techtrancer
That's curious I hate p95 (it's an old hatred don't ask) for me 24/7 use is all the effort I am willing to put into it. Honestly the best ram killer I have dealt with is chrome with YouTube streaming. I have had memtest pass and that still have errors.Quote:
Originally Posted by chew*
Maybe normal?
I habe 4x8gb Dual ranked DDR3 in My AM3+ Board
Can run them by ref clock with up to ~970 in Prime Blend Custom
Also Boot into win around 1000
https://abload.de/thumb/968mhz11-11-11-23-56-vwslt.png
But can't boot with the 1866 multiplier
For those of you performing bclk OC, are you using any SATA devices at all? What are you using for storage?
Yes it would be very helpful to find out how the system behaves with different bclk`s, what voltages to raise in order to stabilize etc.Quote:
Originally Posted by drmrlordx
Also , im with chew on the stability front.I ve found out through empirical testing, if it "seems" stable, even with everything youre doing, but isnt stable in prime blend, it will surely crap out on you sooner or later, or for example, write or copy a file with a crc in it, and you are screwed, and dont even know about it :-/ . I run my pc 24/7, sometimes encoding AND playing a game and running a vm at the same time.Stability and most of all ram/IMC stability is very important.Theres also a trap, even tested and confirmed setting with fixed frequency, CAN become unstable when p states are active.
Agreed.
I will give another example.
"My pc is stable for gaming so i don't need to stress test"
Ok so a new game comes out or a new driver.
You install. You think an error when writing to hd will not happen?
http://img.techpowerup.org/170326/170326101259.jpg
http://img.techpowerup.org/170326/170326101308.jpg
http://img.techpowerup.org/170326/170326101313.jpg
http://img.techpowerup.org/170326/170326101318.jpg
http://img.techpowerup.org/170326/170326101324.jpg
These aren't necessarily the exact settings I ran for the runs above, but the voltages are the same as what I have kept for all of the runs. I do not recommend using that much SOC voltage, but my chip seems to eat it up when going for high ram clocks.
Progress with 32G dual rank on a non ref clock board. 3200mhz booted and validated :D
https://valid.x86.fr/cache/banner/1dngd6-5.png
My Friend Post
3733 CL14 on C6H (Bios 1002)
http://upic.me/i/bl/11capture.jpg
Yep once again thats faily easy on a ref clock board.
We are discussing dual rank on non ref clock boards.
Totally different animal.
I wouldn't exactly call it easy...Quote:
Originally Posted by chew*
Plenty of people are doing it if you check the net or bot.
Just saying.
Infrared who has done it tried 4x8 g ( mimic of dual rank ) struggled to hit 3200 and only clocking in windows not at 3200 divider.
I struggled to boot it 3200 @ 3200 divider ( dual rank ) since i can't bump ref clock in windows.
Try doing that and in comparison SR 2x8g 3733 on ref clock board = easy.
Perspective matters.
I got 3200 with 4x8GB on my board using the 3200 divider fine when I tested using 0702, I have not tried with 1002 yet. 3733 has been a pain to keep stable I will say that, but that could also be my specific CPU.Quote:
Originally Posted by chew*
i'm seeing weird stuff myself. 3200 has been a pain in prime blend but I think I may have it sorted. less is more type scenario for whatever reason. Maybe my chip is dumb.
I really think that these are setting stupidly tight tertiary timings, but that is a guess based on my lack of care to look into it further.Quote:
Originally Posted by chew*
New bios out for msi boards with new agesa.
Damn, that IS a lot of SoC voltage. I also see you've cranked that PLL voltage up there . . . and the SB voltage.Quote:
Originally Posted by cdawall
It's a shame the ASRock boards don't have things like current capability controls and power duty controls, but eh what can you do?
I have a habit of going a bit high with voltages. This particular chip and board isn't mine nor do I really care if it died so it got what made it stable.Quote:
Originally Posted by drmrlordx
Voltages is one of those tweaks chew* really makes me look silly. I am a bit rushed and brute force with them he definitely spends more time getting it just right.
If you stability tested you would know what buttons to push and what you did not have to push :p:Quote:
Originally Posted by cdawall
This board is pushing my patience though with 3200 + IMC brutality. I think it may be agesa that i'm battling....
CH6 is on 800110e
Msi is on 80111c
I'm on 800105
Quote:
Originally Posted by chew*
Lol that was learned from pushing benchmarks I test enough to at least know a general vicinity to shoot for.
Guys, silly question
What do you mean by 'ref clock board' in the context of Zen ?
Cheers
A ref clock board has fine grain ref clock tuning in .00 grain.Quote:
Originally Posted by mongoled
A non ref clock board will have depending on bios adjustable bclock in 1.+ resolution.
So non ref clock 100>101>102
Ref clock 99.80>100.00>100.20
Currently the majority of non ref clock are "locked" at 99.80.
Have never seen this before, are these production boards or engineering samples?Quote:
Originally Posted by chew*
If production, have not come across any sort of list that defines which AM4 boards are ref or non ref.
Never saw such feature in a production board in all my time as an overclocker, from epox, abit, dfi, gigabyte, asrock, asus, something in the back of my mind re MSI not sure if its was to do with ref clocks.
Learn something new everyday!
Production and its kinda new to ryzen am4.
Seeing as a .25 multi gives you 25 mhz and that can make or break stability on an oc with these chips ( yes its that fine of a line) 101x40 gets you 4040 mhz.. So even a multiplier is finer grain than Zens ref clock.
So 100.20 is much finer grain control and one point of having a chip.