AMD Ryzen Discussion Thread.

[trparky]

Yeah, I can be a strange one indeed.

What's wrong with ASUS? Their hardware is usually pretty stable.

 

[Rehmanpa]

Yeah, I can be a strange one indeed.

What's wrong with ASUS? Their hardware is usually pretty stable.

Their customer support makes Comcast look like the pcmasterrace vs console peasants.

 

[EarthDog]

Yeah, I can be a strange one indeed.

What's wrong with ASUS? Their hardware is usually pretty stable.

Nothing more than any other... some customer service horror stories, but, never heard more from that camp as opposed to any other.

 

[Aenra]

I just think that it's overated, overhyped and that by now, important this, by now, no longer what it was but instead significantly declined; again, for me.

I think that barring exceptions (and these can only be gleaned after a good amount of time/research and assuming proper reviews are availabe [ie not "presentations" like this site does]), they are a company that shouldn't be number one in people's minds anymore.

And last but not least, the above in mind, its high end stuff is thus highly overpriced compared to those of other brands. At times, you can get both better performance and a lower price if you look elsewhere.
Again, just me.

(and irrelevant, but if only ASRock moved on to a UEFI that wasn't a clone of Asus's [ten thousands clicks and twenty sub-sections just for kids to feel cool], i'd actually rate it higher than Gigabyte, assuming cost was no factor)

 

[Chicken Patty]

So far what I've gotten it stable at. 1.376v and 3.8 GHz. Earlier I had tried 1.360v and 3.9GHz and failed. So I'll try 3.9GHz at this voltage to see if it passes. The water cooling literally dropped my 100% load temperatures from 85ºC at 3.2GHz and 1.200v to 65.5ºC at 3.8GHz and 1.376v. Room is about 72ºF for both tests. Massive difference.

 

[Nuckles56]

So far what I've gotten it stable at. 1.376v and 3.8 GHz. Earlier I had tried 1.360v and 3.9GHz and failed. So I'll try 3.9GHz at this voltage to see if it passes. The water cooling literally dropped my 100% load temperatures from 85º at 3.2GHz and 1.200v to 65.5º at 3.8GHz and 1.376v. Room is about 72ºF for both tests. Massive difference.

Why with the mixed temperature units? I find that really frustrating.

 

[R-T-B]

Why with the mixed temperature units? I find that really frustrating.

Probably because all software reads out in C, and we don't measure in anything but F in the USA.

 

[Nuckles56]

Most software does have options to got to Fahrenheit if you so desire though, which is why I said something

 

[Chicken Patty]

I'm just used to Celsius when it comes to PC readings, seems more of a standard, or maybe not. Clear enough to understand though. I've gone ahead and edited so it's even more understandable

 

[Aenra]

I'm just used to Celsius when it comes to PC readings

Yeah, more practical too. Mention Celsius in relation to PC temps, everyone understands you, because all software defaults to C. Mention Fahrenheits and you're bound to confuse some people

 

[Johan45]

I went through the transition of Imperial to metric in grade school so have learned both and some of the conversions. There's a simple conversion for F° to C° which is (F°-32°)/1.8. Using 2 instead of 1.8 makes this easier and close enough. 72°-32°=40°/2= 20°C

 

[EarthDog]

Why with the mixed temperature units? I find that really frustrating.

Pc in general has used C since i have been involved (2 decades). Anyone (here) saying F when talking pcs is confusing.

 

[Johan45]

I have no idea why some countries are so stubborn and refuse to conform ( resistance is futile), Metric is so damn easy compared to Imperial, that shit used to make my head hurt.

 

[TommyT]

i just notice that mt top xft is top only to 3.4ghz ?

amd ryzen 1600 its not getting to 3.6ghz top?

 

[Johan45]

i just notice that mt top xft is top only to 3.4ghz ?

amd ryzen 1600 its not getting to 3.6ghz top?

When you're in windows start cpuz and then right click on it. You should see a list of all core speeds you might not be seeing the top speed on core one.

 

[HTC]

Finally found a TaiChi board in my neck of the woods, so i made the order and should receive the parts in a few days.

I'll try to get it working with Win 7 but if i can't, already have an Ubuntu USB drive ready: should be a crash course in Linux because i've never used Linux before!

 

[Chicken Patty]

Finally found a TaiChi board in my neck of the woods, so i made the order and should receive the parts in a few days.

I'll try to get it working with Win 7 but if i can't, already have an Ubuntu USB drive ready: should be a crash course in Linux because i've never used Linux before!

If you need any help with that board, let me know. Been messing around with it a bit since I got it

 

[purecain]

Guys i got confused.. spare me a trip to my cousin's or my dad's please ^^

Assuming an 1800X with 4 cores disabled, aka 4 - 0; how much L3 is available to me? Its entirety, or half?

(i'm asking in anticipation of R9s)

the entirety...
I also picked this little gem up which will definatly help all overclocking on the chipset... these settings are from the c6h but should be on other 3rd party boards.
The Stilt 6/16/17

Those who are able to train the memory at high speeds (>=3466MHz), but are unable to stabilize it due to signaling issues, I suggest that you try decreasing the "Command & Address" related drive currents (increasing the resistance).

AMD CBS > UMC Common Options > DDR4 Common Options > CAD Bus Configuration > CAD Bus Drive Strength User Controls:

Clock Drive Strength = 30.0Ohm
Address / Command Drive Strength = 30.0Ohm
CS / ODT Drive Strength = 30.0Ohm
CKE Drive Strength = 30.0Ohm

24.0Ohm is the default value for all of them, at >=2666MHz MEMCLK (regardless of the DRAM configuration).

These values are not very sensitive so anything up to 60Ohms should allow you to train the memory.

At default settings (24.0Ohms) anything above 3466MHz was unstable due to signaling issues (only B2 DIMM slot was able to run 3600MHz stable).

so it looks to me like the important settings for getting all ram working are coming to light...
procODT @ 80ohms was the sweet spot for my b-die @3466mhz.
now I know the limiting factor for setting 3600mhz thanks to the the stilts advice above. I'm looking forward to testing these out. the possibility of higher clocks is a great news!
I seem to be getting better performance in cpu-z aswell.

 

[NdMk2o1o]

I love this latest agesa 1.0.0.6 update, I have 2x8GB dual rank sticks of mismatched RAM and have them both running at 2800mhz (their rated speeds) I couldn't run both sticks above 2133 on 1.0.0.4 and ohhhhh it tastes all the more sweeter knowing that I have them on a $60 Asrock board that can also clock my 1600 non-X to 3.9ghz @1.35v all day long, it's good to see AMD back in the game, though I guess I could of bought an i5 4c/4t for the same as my Ryzen 1600.....

 

[phanbuey]

im having issues... 1.0.0.4 runs at 3200Mhz stable 1.35v but wont even post at that on 1.0.0.6 - :/ gonna have to wait for the non-beta i think...

 

[purecain]

try going through the procodt settings. start at around 40ohms and keep moving it up one each time. I meant to put that 60ohms was the sweet spot for my kit. when ive looked at other threads across the net. I found maybe another 10 people saying they had found stability with that setting at 60ohms.... good luck anyway phanbuey.

 

[Norton]

I love this latest agesa 1.0.0.6 update, I have 2x8GB dual rank sticks of mismatched RAM and have them both running at 2800mhz (their rated speeds) I couldn't run both sticks above 2133 on 1.0.0.4 and ohhhhh it tastes all the more sweeter knowing that I have them on a $60 Asrock board that can also clock my 1600 non-X to 3.9ghz @1.35v all day long, it's good to see AMD back in the game, though I guess I could of bought an i5 4c/4t for the same as my Ryzen 1600.....

+1- Asrock did a nice job on their boards with this update!
My Trident Z sticks (F4-3200C16D-16GTZB) are running 3200 and they loaded their XMP profile first shot with no problem at all

 

[Aenra]

Off topic, but i didn't want to start a Threadripper thread (before they're even out).

Wanted some opinions really.. I'm planning for a second rig and what i'm thinking is a bit.. well, you'll see
- i don't want to go below or over 8 cores
- i want quad band, not dual
- i want it to be AMD
- better freqs obviously a bonus

Which would restrict me to some R9, or would have, if it wasn't for the core count. So what i'm currently having in mind is:

- buy the R9 flagship (16c/32t). I get me the quad band. Why the flagship you ask?
- disable an entire die; i keep it as an 8core. Why not buy a cheaper R9 and keep it 8core? Core configuration, that's one + this way, i only have one complex active no matter what, so i escape the CCX-to-CCX latency 'penalties'. And this is the only way to do that.
- extra bonus? Running it as 4+4 - 0+0 or 0+0 - 4+4 means i get double chance at getting a good OCing result; be it in higher freqs with either setup, or at the least, with a lower voltage. Do one, write down highest freq/voltage, do opposite and see which is best.
- possibly buy a binned one from siliconlottery.com, do the above, hope for a 4.2ish GHz 8core; ie basically an 1800X with quad channel RAM, lol

Now obviously, money for nothing, i know; i pay for the flagship and i end up running it as an 8core. But.. i don't want fewer cores, i don't want the CCX-to-CCX latency and i do want me the quad channel; most of all, i do not want to buy a second Intel, i personally think AMD has earned my support again.
As such, excluding the price issue (they're so much cheaper anyway, it's... O.K.), what do you guys think?

And most importantly (as i haven't disabled cores thus far in any R7, don't know how it goes), would i be able to achieve higher freqs the less cores i've got on? Not sure how it goes with Zens in this aspect, never even tried as neither build was mine.

 

[Johan45]

I'm not certain on the intricacies but the way I understood was 2 channels are handle by each die(2 X CCX) that's why epyc can handle soooo much ram in "8" channel. As for the OC that's just luck of the draw and I haven't seen anywhere that disabling parts of the complex actually get it any higher.
What I can't understand is WHY ? If you need 8 cores/16t, why wouldn't 16c/32t be even better assuming you have the multithreaded SW to begin with. Double cores should more than make up for anf latency or Overclock penalty.

 

[R-T-B]

You won't escape inter-ccx penalties as you still use the l3 cache and memory controller from the disabled die.