Ryzen - Return of the Jedi

[chew*]

I have exhausted every option except newer bios which i do not want to use...i know its slower backflashing is a pain and some microcode is not leaving...im seeing this on taichi right now...annoying..

My llc is set the way its set so when i set 1.40 bios i see 1.40 socket with dmm. Same with soc...

I set manual voltages to ensure board is not playing dirty tricks...Auto at mercy of bios..

Fan is ocing...= run cooler

[Yeroon]

Chew* - Out of curiosity, is the voltage difference from the DMM to software a fixed offset, or a percentage? Also, are different boards of the same model consistent in the skew?

[AMD_OCer]

I have exhausted every option except newer bios which i do not want to use...i know its slower backflashing is a pain and some microcode is not leaving...im seeing this on taichi right now...annoying..

My llc is set the way its set so when i set 1.40 bios i see 1.40 socket with dmm. Same with soc...

I set manual voltages to ensure board is not playing dirty tricks...Auto at mercy of bios..

Fan is ocing...= run cooler

We are at the mercy of newer BIOS at this point on the CH6.. or some other weird setting.

Well at least we got the 4x8gb running 3200 CL 14, this is a big feat. Thanks for the hard work.

[lowdog]

I have exhausted every option except newer bios which i do not want to use...i know its slower backflashing is a pain and some microcode is not leaving...im seeing this on taichi right now...annoying..

My llc is set the way its set so when i set 1.40 bios i see 1.40 socket with dmm. Same with soc...

I set manual voltages to ensure board is not playing dirty tricks...Auto at mercy of bios..

Fan is ocing...= run cooler

Hey Chew check this;

Buildzoid did a breakdown of the Taichi https://www.youtube.com/watch?v=_ic5d5qAhBU

Around the 21 minute mark he talks about the LLC on this board and his finding are way off in comparison to what you are finding but he is taking his DM measurements from a completely different place so that probably accounts for the difference.....what do you reckon about his findings please????

[I.nfraR.ed]

It's always best to measure closest to the core (CPU or GPU) and in this case it is on the back of the socket, where chew* is measuring.

[chew*]

Hey Chew check this;

Buildzoid did a breakdown of the Taichi https://www.youtube.com/watch?v=_ic5d5qAhBU

Around the 21 minute mark he talks about the LLC on this board and his finding are way off in comparison to what you are finding but he is taking his DM measurements from a completely different place so that probably accounts for the difference.....what do you reckon about his findings please????

Build zoid can measure where he wants that is his choice.

I was advised by many ( stilt, raja, tin just to name a few ) with a electrical engineering background to only measure at socket so that's where I measure.....

I just did a video and measured both. I also debunked any comments as to the accuracy of my DMM.....

Uploading now........

[lowdog]

Build zoid can measure where he wants that is his choice.

I was advised by many ( stilt, raja, tin just to name a few ) with a electrical engineering background to only measure at socket so that's where I measure.....

I just did a video and measured both. I also debunked any comments as to the accuracy of my DMM.....

Uploading now........

@ Chew ........EXCELLENT!

Was in no way implying your findings were inacurate rather to the contrary was thinking his were off. Thanks so much for the information and your time put into all Ryzen related stuff, much appreciated.

[chew*]

Ryzen - Mythbusters R5/Taichi edition how to 32g dual rank.
https://youtu.be/MwM7HmREc5Y

[AMD_OCer]

Thx for both vids today chew. We really appreciate all your efforts!

So CH6 can do 4x8gb @3200/CL14 (through the 3200@CL14 kit) where ASRock can do 2x16GB (3200@CL14 through the 3400@CL15 kit)... We are almost through the 32GB battle.

Which of the two boards would you buy/keep for yourself at this point after messing with them for a while?

[chew*]

3400 16-16-16 kit

Both i would buy again lol.

Last contestant with ref clock is k7...almost caught up on bills...then i will get it.

[Formula350]

For what it's worth regarding CPUz voltage readings... I haven't the ability to check my setup with a DMM at the socket like you, and the Titanium's test points are all the way below the friggen Chipset, so I can't imagine THEY are very accurate (I haven't even bothered checking them). Nevertheless, just based on what HWiNFO is showing, it looks like CPUz is going off of the Motherboard's VCore.
Running stock (well, 102BClk) and have AIDA64's Cache stress test running since that seems to be the most stressful.
HWiNFO
MB VCore = ~1.248V (1.240 to 1.256V)
CPU CoreVID = 1.150V
CPU VDDCR_CPU = 1.219V (1.212 to 1.225V)

CPUz
Core Voltage = 1.248V (1.240 to 1.256V)

The reading update at different rates but that's just due to each ones refresh being different.
Feels like VDDCR_CPU is the one to go by though, as my MOSFET reading (in HWiNFO) keeps pace with the CPU's reading for Voltage. They're generally within a few thousands of each other's readings. Same for the VDDRCR_SOC and it's MOSFET which both show a pretty solid, unmoving 0.963V, whereas Mobo's is reading 0.976V

Meh I'm curious now, so I'll see what the board's test pins read...
And... to start with. The voltage differential between that GND pin and where I felt comfortable taking the reading from is +0.004V. Meaning they coincide exactly with HWiNFO's readings. CPU reads 1.252V(+0.004V higher) and CPU-NB reads 0.980V (+0.004V).
Oddly, the board has a test pin for CPU_VDDP and yet, the pin reads 0.005V. Which either means, it was only reading 0.001V or margin of error and reads nothing. I'd have to do a continuity test to know if it really isn't doing a thing.

I CAN say that the latest BIOS (official 1.40) didn't re-correct for the +20 temp offset, as tCTL still reads high compared to tDIE. Which made me feel better since I could not for a moment feel that things were actually running at 70C under water.

Also, notes in regards to 1.40 BIOS...
The 1.0.0.4 DDR Blues are not just ASRock related, it's indeed something to do with the AGESA. As soon as I flashed it, I'm also no longer able to run 3200 on my Titanium. >_>
The hidden BCLK on the board only goes up to 103 lol I modded it into view. 103 even isn't stabled, for that matter, but 102 is fine though. I suspect it's just at the ragged edge of it's ty built in capability is all, since it results in the storage controller flaking out. (Must access Ryzen's internal BCLK)
I have not tried flashing back yet to see if the Titanium has the same issues. The AFUWIN flash will let you write over everything if you want, so either that, or the "Emergency Flash" might be the way to go in order to truly flash-back.

[chew*]

Ill get to the b350 soon formula350.

Just been slammed with work...so busy i still have 3 untouched R5 chips but i like this 1400 so much that im not even concerned with any of them except maybe the cheaper 6c sku.

Anyway i'll get more stuff done just need to bear with me.

My time is extremely valuable and everybody needs me and i pretty much need none of them.

Between fixing cars/trucks helping run our shipping business, setting up my buddies srt swap neon i built for the track and overclocking i got alot on my plate....

To be continued.

[Formula350]

Ill get to the b350 soon formula350.

Just been slammed with work...so busy i still have 3 untouched R5 chips but i like this 1400 so much that im not even concerned with any of them except maybe the cheaper 6c sku.

Anyway i'll get more stuff done just need to bear with me.

My time is extremely valuable and everybody needs me and i pretty much need none of them.

Between fixing cars/trucks helping run our shipping business, setting up my buddies srt swap neon i built for the track and overclocking i got alot on my plate....

To be continued.

lol It's all good. The big test was "is the modded BIOS corrupt?" and as aGeoM initially proved, it isn't. It gave myself the confidence to then go ahead and flash my Titanium with a modded, so once again we have more evidence.

Now, whether or not a dumped ASUS that gets modded will flash, is indeed still up in the air. However, rest assured, I'm hanging around because of the good details that are showing up in this thread So take your time heh

As a bit of a followup to my voltage post... Really hard to know quite WHICH one of the two voltage readings to put more faith in... lol Is it VDDCR_CPU or MB's VCore? Cuz when the system is idle, the VDDCR_CPU is 1.481V and VCore is a more realistic 0.520V. And yet the MOSFET reads as indeed cranking out the 1.48ish volts.

*throws papers in the air* lol Ryzen has surely turned everything upside down for me :P

EDIT: haha I just realized... you've overclocked a Neon!

EDIT 2: Alright, so for anyone on at least the MSI Titanium, I can say that flashing BACK from 1.40 to 1.10 (shipping version, I can post if anyone needs it), will reinstate the OLD Microcode. I don't remember what my 1.10 showed before as I was too dumb to make note of it originally, but just now before I flashed back I did and the results:
v1.40: 800111C
v1.10: 8001105

Also it loaded up my RAM at 3200 automatically heh
Now that I've confirmed that, I am going to flash another more heavily modded 1.40 to poke my nose in, then off to bed as it's nearly 3AM!

[chew*]

EDIT: haha I just realized... you've overclocked a Neon!

Its only a stable 100% OC stock does about 135whp with the single cam.

This does 270whp or maybe a tad more with the 2.4 turbo swap with mods

Best mod of all is insurance goes by vin # single cam ATX sure yah that's what it is

Still need to wire up the 2 step WOT box but sneakers are on now....maybe take it to the dyno if he gets me a wideband with a serial port I can plug into the laptop....

Can probably pull a few more ponies out of it but I think the head is due for an overhaul so we really just went to nail a time at the track before something goes poof......

Last leak down test showed something going on with cylinder 4 and it does not seem ring/piston/cylinder related.

You should see my neons

[chew*]

Here is some of my past and present personal stuff,

Old all motor setup I had in a ugly shell.


One of the nicer cars I built and I literally mean built but got rid of.
Before



After



My current race car. yes that is a cage inside it

[Formula350]

lol Well, you know what the women say... "Boys and their toys..."
Some of us get the itch, where more is never enough, and stock DEFINITELY isn't enough. So we free mod it till that isn't good enough. Then we throw money at it for real performance. Then we sacrifice time... and knuckles... right up until about the point it's starting to be so crazy it's a Trailer Queen. (Note to onlookers: not Trailer Park Queen, but a car that doesn't drive itself to and from races lol Also known as a Garage Queen, as its where it sits most of the time).
Looks nice though! Always did like the look of DRs, have a set myself, though they did come with the rims lol My girl is OB1, so I don't have the luxury of tinkering with programming Long long LONG ago, I did tweak a tune for my 1991 Firebird I had in early 2000s and got it burned to a chip, which did seem to help, but I don't honestly know if I ever did hammer one out for my 93, or if I did I don't think I ever had gotten it burned (as I don't have an EEPROM programmer). Alas, that was so long ago I don't even remember the name of the software I had used! Forgotten all I had learned anyways :P
I'm a RWD person myself, but I know those little SRT4s aren't anything to turn my nose up at! Dodge hammered out a nice little 4banger for the tuner crowd with that one. I bet one of those AWD Darts with that would be a friggen blast.
EDIT: hah you posted the followup whilst I was typing this. I'm surprised to see the "Old" Neon body style being used :P Those definitely were Hairdresser models haha


At any rate, got 1.40 back on and perhaps the tighter timings in the AGESA just meant 14-15-15-35 weren't stable, as I let the BIOS apply what it wanted (15-15-15) and it booted up this time. Which since CL15 won't work above 2400, means it's running 16-15-15-35, but it's running at 3200... *shrug* Way too tired to vet these speeds tonight, though. Only thing I'm going to do is reboot, see if BCLK 102 works with 3200 still, and go to bed! lol

[chew*]

Older neon = instant weight reduction.

Add in weight redux ive done....hits the scale at 1900lbs

Power to weight ratio.

Its turbo and fully built.

The all motor fully built sits on the engine stand now collecting dust. Might stick it in my daily. 9k rpm trapping 3rd is nasty.

Burning proms lol. We build our own pcms nowadays

[AMD_OCer]

Ok so I found this table on overclockers.net which has lots of DRAM configurations that users were able to achieve with the CH6 MB.

I exported it to excel and sorted them to get the "best" ones that interest me the most. Here is the results:

Username DIMM Part number BCLK DRAM ratio DRAM freq Timing DRAM v SOC v Link
elmor F4-3200C15-16GVK 2x16GB 116.4 2666 3103 MHz 14-14-14-14-38-1T 1.4 1.15 http://www.overclock.net/content/type/61/id/2986578/
Kriant F4-3200C15D-32GTZ 2x16GB 114.4 2666 3050 MHz 14-13-13-30-2T 1.35 1.15
femfemti F4-3200C15D-32GVK 2x16GB 100 3200 3200 MHz 18-14-14-34-1T 1.35 1 http://valid.x86.fr/6e5xmn
BoMbY F4-3200C14D-32GVK 2x16GB 100 3200 3200 MHz 18-16-16-16-36-2T 1.35 1.18 http://valid.x86.fr/j2ckza

Couple interesting findings:
The top 2 rows are the most interesting, both have a SOC voltage of 1.15.. In Ryzen the mem controller , USB, etc are all handled under that SOC voltage, so that might explain some points.

I see some users on that list using BCLK as high as 135-140s...

[chew*]

Thats pretty much ch6.

It can do 2x16g but not rated.

Im of the mind rated or better or don't bother.
Cas latency is worth alot more than most people think...

It actually completely takes up the slack for the lack of higher fabric speeds.

I would avoid taking any stock in any of the results done on latest betas.

Amd slacked off memory performance alot in latest agesa...

Any gains are merely the fact that timings are way looser which is pretty evident by looking at trc from one verion pre latest agesa to latest agesa.

Also if people disable interleaving etc to get "higher" your going to want to null/void those results.

Once again done by making performance worse in an effort to show a higher #.

I would avoid really high bclk for 24/7...120 is already pushing it for 24/7 use...my 2c..

[AMD_OCer]

chew: which BIOS version are you at with the CH6?

I agree with everything you said, though at the same time, I really like the 3103 MHz, 14-14-14-14-38-1T by elmor, its just 100Mhz shy of the 3200 goal with 14 timing. Maybe his limitation was that he got the 3200C15 kit instead of the 3200C14, which would mean that that is the best he would get.

Am debating between the 4x8gb 3200 CL14 vs the 2x16gb 3200 CL15. Maybe I will grab whichever is cheaper. 100mhz wont even be dramatic in terms of results, given latency is still 14

Might order parts soon.... first AMD build in 4+ years.... I had nothing but AMD before that... and I use to OC the crap out of them. Need to read up more on the Ryzen architecture to get more familiar with what I can do with OCing. Things changed a lot.

[chew*]

Some users complained video was tough to see settings on asrock due to color scheme so here goes.

32g how to

step 1 bios settings voltages ref clock 120 save reboot
PCI/e gen 1/2 is based on your ref clock.
Read stilts strictly technical on the shift points and use accordingly......

step 2 set cpu clock multiplier reboot

step 3 set xmp reboot

step 4 fine tune timings through xmp profile and fine tune voltages 1.0-1.2vddp, vddr vtt SOC if necessary.

Confirm stable in memtest before you boot.......

Here is where video left off last night after raising my vddp to 1.20......

Check video compare time stamp to prime 95.......as you can see vddp made a huge difference from almost killing OS to stable.

[AMD_OCer]

^^ Thanks for that. CH6 still wins because it got the 4x8GB at the rated stock speeds and latency.

With the ASrock, mem kit used is a 3400 CL16.

[chew*]

^^ Thanks for that. CH6 still wins because it got the 4x8GB at the rated stock speeds and latency.

With the ASrock, mem kit used is a 3400 CL16.

3400 cl16 breaks down to a 3200 cl 15 kit which breaks down to a 3000 c14 kit.......

Its actually a lower bin than say a 2x 16g 3200 rated C14 kit which would break down to 3400 c15, 3200 c14, 3000, c13...

Basically the 4x8gb did spec.

the 2x16g exceeded spec

The asrock is a tad slow in 32m pi so I have a feeling it runs a bit slacked vs the asus which is also something that must be factored in

[AMD_OCer]

Yep it exceeded spec with latency.

If you read my previous post, 2x16gb 3100 CL14 is very respectable on the CH6. Take into account that the 3200CL14 kit is cheaper than the 3400@CL16.

I really like the 4x8gb running exactly at spec, its just that 2x16gb allows further upgrades. And those kits seem to be cheaper than getting 2 kits of 2x8gb.

So the CH6 wins in both cases, since I can literally use 4x8gb spec or 2x16gb at 3100@CL14

[chew*]

Misinformation = worse than no information.

Socket vs caps dmm testing on taichi...

https://youtu.be/xuXIT_CHdcs