Ryzen - Return of the Jedi

[chew*]

Let's see 32m

Working through imc coldbug and i have better times saved on usb stick but gives you a general idea....







Bench session went very smooth ch6 cooperated 100%.

As long as your aware that 95% of chips are going to have imc coldbug and avoid high mem speeds it just works.

I had ln2 mode enabled for entire session. Never had a boot issue till i tried cas 9...my sticks are not good enough for c9 it seems.

Big to runmc (ron mcall) if he still lurks the forums. 5 years plus the phase units he built just continue to work flawlessly....never even needed a recharge.

Bios was stuck @ 5c,ryzen master 0c. Even under cinebench lowest it hit was -34c, under pi -38 and idling @ -44c with 1.6v

[StAndrew]

Are you running AGESA 1.0.0.4 or the latest AGESA 1.0.0.4a?

[chew*]

I am running the fastest possible PI config.

This chip is flaky i took screens to show.

Typical ryzen dramatic black screen in cinebench when not stable..
This a a 3.9 stable air chip....
Just testing stability not bothering with score...its w10 anyway so does not count...

1.4v



1.4 still


1.45


Where i left off on pi. Tuning diff strap and speed currently..


At the above effeciency...im 7:30s @ 4.8....
Its far from effecient for 4.6 but very fast for these memory speeds...

[FlanK3r]

these quick blacks I got today several times. But how to fix it? I can run 4100 MHz R15 with 1.42V in BIOS + LLC3. So I tried 4150 MHz in steps 1.44V to 1.47V and the same, after few seconds of rendering black window and code simply 8 .

[chew*]

these quick blacks I got today several times. But how to fix it? I can run 4100 MHz R15 with 1.42V in BIOS + LLC3. So I tried 4150 MHz in steps 1.44V to 1.47V and the same, after few seconds of rendering black window and code simply 8 .

Bin to win......no way to fix black screen. its the limit of your chip at x volts and x temps.........

Anyway I just wanted to show there is a ton of room on the table still.

this run was not tuned properly and with just a slight increase in memory speed completely smashed my above run......

busy week again and its certainly killing my ability to tune around the imc bug.....phone is ringing non stop......I do best with 0 distractions....

7:3x is more than possible at 4.6......that's why I am running 4.6...

Of course you need an imc that will cooperate for that 7:3x range but I am 100% certain without a good cold imc I can pull this down to 7:4x range

[FlanK3r]

Any tips for LN2 and IMC coldbug? I will test 1500X and maybe also 1700X this weekend with LN2...

[Charles Wirth]

Start off with a profile 2666Mhz CL14, ignore cpu fan, and see what the cores can do.

Fighting with ram and LN2 is fun, I had the luxury of the cascade @ -120c to see what the IMC wont do. In the guide Elmor has a graph that is pretty accurate to all chips and how they respond cold, just factor in he is basing that graph upon the 1800X and the lesser bins are in fact lesser bins thus lower frequencies to be expected.

Dropping SMT will get you a little extra speed if you are seeking 5Ghz without threading.

[chew*]

Like I said 7:4x definitely possible even fighting a wicked bad cold bug.

I'm really stretching it running this chip @ 4.6 so yet to get an actually tuned run to complete.

I think but not positive 7:3x might be a reality if I can get it to cooperate for a tuned run.

If you notice I pulled 3 secs off time with considerably less memory speed without software tweaking

Think I am going to break the rig down for the meantime however as I just picked up a K7 to test.

32g testing is top priority so give me a few days to work through it and see if its possible.....

[FlanK3r]

Thank you guys for few tips Keep fingers cross After I will lost my nervoues, I will have second PC for LN2 benhcing (Phenom II x6 1090T BE)

[techtrancer]

Thank you guys for few tips Keep fingers cross After I will lost my nervoues, I will have second PC for LN2 benhcing (Phenom II x6 1090T BE)

To Late
My best X6 is sold
Would have been a good processor to bench under LN2

[chew*]

Most good air = bad ln2.

[Johan45]

Was that last run from 1866 divider?

[FlanK3r]

Chew:not always true. I tested x6 1100T, this chip was good on AIR and good at LN2 too

[ZenEffect]

Was that last run from 1866 divider?

Trc gives away what divider is used

[I.nfraR.ed]

Or you can just see the FSB : DRAM ratio on the screen. (FSBRAM - damn smiley...)

[chew*]

Use whats fast...not what you think is fast

[Aerou]

My initial thrill of Ryzens performance for a workstation is now being scarred by random instabilities and crashes when running long-term, on NVMe storage, virtual machines, encoding and BOINC projects when not in use. My feelings from trying to use Ryzen for a workstation - it's not production-ready. Hopefully AMD irons out the kinks in a better BIOS/AGESA/chip revision. There are definitely some I/O issues.

[techtrancer]

Most good air = bad ln2.

Was wathercooling
Don't know How the X6 perform at higher clocks under ln2
My cooling was only enough for 4.2ghz @ Prime
But Need realy less v-Core for 4.2

[chew*]

My initial thrill of Ryzens performance for a workstation is now being scarred by random instabilities and crashes when running long-term, on NVMe storage, virtual machines, encoding and BOINC projects when not in use. My feelings from trying to use Ryzen for a workstation - it's not production-ready. Hopefully AMD irons out the kinks in a better BIOS/AGESA/chip revision. There are definitely some I/O issues.

what is you SOC voltage?

SOC preety much effects usb pci etc etc.

[chew*]

Ok so gigabyte K7........board can be a chore to OC ref clock wise.....

The basic stuff is easy with 2x8g......but it ends there.

BUG
ref clock is rendered useless at least with this chip once you deviate from stock multi ( auto setting )

Pretty much anything over 33.50x multi unless you have a golden 4.1g chip is going to be near impossible to use refclock on.

Will post e6 if you change multi when using ref clock.

I found a workaround to get my 1700 up to 3.9 using auto via ryzen master.....great if you need to clock up....not so great if you need to boot a 1800x that can not do 4.3gig with 36x ( auto )





It can do over 3600 after about 2 minutes of boot looping.

Ok maybe a slight exaggeration but yah at least 5 boot loops I would guess for each timing and AM4 training...

Basically it has potential but bios needs some serious polishing.



32g operation in its current state at optimal timings is not happening.

I honestly will not bother with super slacked timings, I find it a rather pointless sacrifice when the competition can do it...hopefully its possible after a few updates.

good news is it is actual capable of stability at 120 ref clock......

The other good news is that this is the chip I have been running at 1.6v for 3 days on phase change......looks like its not degrading....yet.

[chew*]

Ok got a little time in on the R5 1400.

did 4150 cine air @ 1.4

does 4.4 phase 1.4 not much scaling only tried 1.45....

whooping that 4770Ks azz though.....



fooled around with 32m on it diff strap diff tune faster time

[techtrancer]

How much does the R5 1400 @ 24/7 usable Voltage and 24/7 Memory in Cinebench?

[chew*]

High 800s @ 4050 @ 1.4v 24/7 3200 memory....even works with 32g..

Of course that is this chip...seems r5 is quite the lottery as far as overclocking goes..

[techtrancer]

Thx
With My FX i am around 740 with My 24/7 Setting
~4.7GHz & 930 @ CL11
Think will wait for Ryzen+

[chew*]

since the thread got quite long I am cleaning up usefull info and sticking it on page 1 for easy reference

will be adding K7 info soon