Ryzen - Return of the Jedi

[chew*]

Thats typical when stressing amd imc. The more things change the more they stay the same.

[cdawall]

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.

[sdlvx]

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.

[chew*]

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.

[Charles Wirth]

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.

[chew*]

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

[chew*]

Ok Here goes timings versus fabric single rank and dual rank.

Single rank 2933


Single rank 3200


Dual Rank 2666


Dual rank 2933


Conclusion

Single rank speed + fabric wins

Dual rank Application specific tossup.

[cdawall]

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.

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.

[chew*]

As promised.

Bios setting voltages used through all memory testing.

[drmrlordx]

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.

Should be listed as MEM VDDIO or similar.

You will also want to bring MEM VTT up to half your MEM VDDIO value.

[chispy]

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.

[chew*]

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

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.

[chispy]

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

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 , 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 . My Sammies B-die can go ~3300Mhz cas 12 no problem whatsoever

Regards: Chispy

[chew*]

Solder pads. Left SOC +/- Right VCORE +/-



Grounds on right side of pads. Positive on left sides of pads.


Soldered to a spare molex extender for ease of use. YMMV.

[Harney]

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

[chew*]

The outcome of knowing your voltage vs guessing. 1.39-1.41 measured vcore, 1.0 SOC




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.

[techtrancer]

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+

[chew*]

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.

[techtrancer]

Is the Chip who does not newer?

[cdawall]

Found out the ram does CL14, cinebench scores didn't change with the tighter timings

[chew*]

Is the Chip who does not newer?

1800x = 2933 fine

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.

[techtrancer]

Hm
Maybe the CPU multiplier makes Problems?
The PII X6 have Little Bit Problems with 200x10
200x9 ore 200x11 works better than 200x10

[chew*]

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.

[Oberon]

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.

[cdawall]

@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