AMD RYZEN Builders Thread

[IRobot23]

Okay,
just doing some overclocking with my R7 and Crosshair 6 hero.
So settings were 3,8GHz, 1.3vcore = unstable (2 sec and crash) same settings expect LLC from 1 to 4 and VRM current power from 110% to 130%. I was paying attention to vcore, VID and there is no change but CPU runs "stable" (at least 20 min in prime95).

I do have this question.
What did I exactly change?
I mean I was expecting higher voltages (vcore, VID), but nothing change ( I didn't pay attention to core current) ... well even power per core is almost same (HWinfo64), but be bugged?

Thanks in advance.

 

[.vodka]

LLC changed, voltage under load is now higher than it was before, and your CPU is now stable at 3.8GHz. There's some serious vdroop with this motherboard, 1.375v LLC auto (basically, LLC off) droops down to 1.3v on my 3.8GHz 1700 for P95 28.10 128k in place. Check the LLC section in this very useful post. According to The Stilt you shouldn't mess with LLC on this platform and Raja@ASUS recommends LLC1-2 and no more.

Keep an eye on SVI2 TFN voltages shown in HWinfo, those are taken from internal telemetry and are the most accurate ones.

 

[Reinvented]

LLC changed, voltage under load is now higher than it was before, and your CPU is now stable at 3.8GHz. There's some serious vdroop with this motherboard, 1.375v LLC auto (basically, LLC off) droops down to 1.3v on my 3.8GHz 1700 for P95 28.10 128k in place. Check the LLC section in this very useful post. According to The Stilt you shouldn't mess with LLC on this platform and Raja@ASUS recommends LLC1-2 and no more.

Keep an eye on SVI2 TFN voltages shown in HWinfo, those are taken from internal telemetry and are the most accurate ones.

I wouldn't trust the voltages at all that are given by HWMonitor or ANY software monitoring program. DMM or bust.

Okay,
just doing some overclocking with my R7 and Crosshair 6 hero.
So settings were 3,8GHz, 1.3vcore = unstable (2 sec and crash) same settings expect LLC from 1 to 4 and VRM current power from 110% to 130%. I was paying attention to vcore, VID and there is no change but CPU runs "stable" (at least 20 min in prime95).

I do have this question.
What did I exactly change?
I mean I was expecting higher voltages (vcore, VID), but nothing change ( I didn't pay attention to core current) ... well even power per core is almost same (HWinfo64), but be bugged?

Thanks in advance.

The ONLY thing that you changed is the amount of voltage that gets applied when there is a load on your CPU. As I mentioned before, do NOT trust software in what it reports back for how much voltage is applied. If you want to know exactly what it is, then you need to get a multimeter and probe the correct spots around the socket, including VRM's.

 

[.vodka]

I wouldn't trust the voltages at all that are given by HWMonitor or ANY software monitoring program. DMM or bust.

Of course.

Still, check that post, there's a section quoting people who have actually checked with a DMM on the back of the socket and the SVI2 TFN values are closest to what the DMM reported.

 

[IRobot23]

So I tried 1.35V on MB for 3,8GHz on all cores.. Failed. Now trying 1,4V (LLC - auto, VRM current power -auto)

Higher voltage shown, same power as before ( HWinfo64) Unstable.

I have had most of Gigas MB, now with ASUS I am just confused

1.4V on MB would expect around 1.35 (low), 1.375 (mid) 1.395 (high)..., but 1.334V on AVG after ~ 15 min of prime95? And this is shown from ASUS, ryzen shows even lower (1.32V).

Basically few more watts was added (115-125W from previous and current is also few % higher).
EDIT:
Little over 1h stable, under 70C, should be fine.

What would be maximum safe vcore? 1.4V?

 

[krumme]

So I tried 1.35V on MB for 3,8GHz on all cores.. Failed. Now trying 1,4V (LLC - auto, VRM current power -auto)

Higher voltage shown, same power as before ( HWinfo64) Unstable.

I have had most of Gigas MB, now with ASUS I am just confused

1.4V on MB would expect around 1.35 (low), 1.375 (mid) 1.395 (high)..., but 1.334V on AVG after ~ 15 min of prime95? And this is shown from ASUS, ryzen shows even lower (1.32V).

Basically few more watts was added (115-125W from previous and current is also few % higher).
EDIT:
Little over 1h stable, under 70C, should be fine.

What would be maximum safe vcore? 1.4V?

1.4 is perfectly safe. If i remember correctly xfr on the 1800x uses like 1.45V

 

[krumme]

AT ram test on zen
https://www.anandtech.com/show/11857/memory-scaling-on-ryzen-7-with-team-groups-night-hawk-rgb

Pretty weak test imo and nothing new. Pretty nice scaling just from 2400 to 2933.
I just bought 2 different cheap corsair 3000 lpx c15. One runs xmp 2933c15 the other 2933c14. Different chips.

And its cheap stuff not remotely b die quality. Imo c16 used in the test is to slow even for so called midrange.
We need c14 c15 c16 test for it to be interesting.

And btw. Give me that samsung b kit in the test and i will eat my hat if i cant get it to run 3200c14

 

[IEC]

AT ram test on zen
https://www.anandtech.com/show/11857/memory-scaling-on-ryzen-7-with-team-groups-night-hawk-rgb

Pretty weak test imo and nothing new. Pretty nice scaling just from 2400 to 2933.
I just bought 2 different cheap corsair 3000 lpx c15. One runs xmp 2933c15 the other 2933c14. Different chips.

And its cheap stuff not remotely b die quality. Imo c16 used in the test is to slow even for so called midrange.
We need c14 c15 c16 test for it to be interesting.

And btw. Give me that samsung b kit in the test and i will eat my hat if i cant get it to run 3200c14

Ian apparently doesn't realize that AMD/mobo mfgrs set their own subtimings for memory. That's why he's seeing higher MT/s CL16 RAM actually score *lower* than lower MT/s RAM at the "same" timings (e.g. in gaming tests) - it's because the strap is setting worse subtimings.

There's a lot of room for improvement in the default subtimings used for compatibility purposes. Tech reviewers still haven't caught up to the likes of The Stilt and enthusiasts in getting the most out of the AM4 platform.

I bet you DDR-3200 CL14 even at default XMP settings will beat his fastest CL16 set, simply because of the better latency and better subtimings.

 

[.vodka]

Ian apparently doesn't realize that AMD/mobo mfgrs set their own subtimings for memory. That's why he's seeing higher MT/s CL16 RAM actually score *lower* than lower MT/s RAM at the "same" timings (e.g. in gaming tests) - it's because the strap is setting worse subtimings.

There's a lot of room for improvement in the default subtimings used for compatibility purposes. Tech reviewers still haven't caught up to the likes of The Stilt and enthusiasts in getting the most out of the AM4 platform.

I bet you DDR-3200 CL14 even at default XMP settings will beat his fastest CL16 set, simply because of the better latency and better subtimings.

I was surprised by his response in the comments.

Hopefully there will be a followup article that shows the entire picture. This one as it is, is worthless.

 

[IRobot23]

AT ram test on zen
https://www.anandtech.com/show/11857/memory-scaling-on-ryzen-7-with-team-groups-night-hawk-rgb

Pretty weak test imo and nothing new. Pretty nice scaling just from 2400 to 2933.
I just bought 2 different cheap corsair 3000 lpx c15. One runs xmp 2933c15 the other 2933c14. Different chips.

And its cheap stuff not remotely b die quality. Imo c16 used in the test is to slow even for so called midrange.
We need c14 c15 c16 test for it to be interesting.

And btw. Give me that samsung b kit in the test and i will eat my hat if i cant get it to run 3200c14

Well that test is not done well.
https://www.youtube.com/watch?v=87M3QdEzRFk

 

[krumme]

Ian apparently doesn't realize that AMD/mobo mfgrs set their own subtimings for memory. That's why he's seeing higher MT/s CL16 RAM actually score *lower* than lower MT/s RAM at the "same" timings (e.g. in gaming tests) - it's because the strap is setting worse subtimings.

There's a lot of room for improvement in the default subtimings used for compatibility purposes. Tech reviewers still haven't caught up to the likes of The Stilt and enthusiasts in getting the most out of the AM4 platform.

I bet you DDR-3200 CL14 even at default XMP settings will beat his fastest CL16 set, simply because of the better latency and better subtimings.

Yeaa i mean in this case its as easy to just copy The Stil b die settings. How hard can it be? I will bet that works 3200c14 with the strict sub timings.

Its not the technical level i expect from an AT article. Seems like some sponsered stuff. Dont know?..

Next time they can enter here or the threadripper builders thread to get some advice.

 

[el etro]

Hey guys, do the Gigabyte GA-AB350M-HD3 can support the Samsung 960 evo 500gb at Nvme mode, at it's full speeds?

 

[kallisX]

i see AMD has chipset drivers for win7, even though they said it were officially not supported.
but only 64bit, what would happen on 32bit? with my spare office computers i always just clone the drive from the old machine to the new. using windows repair utility or "boot with dissimilar hardware" it has worked in the past eventually. Athlon XP--> AM2-->FM2 --> AM4

 

[Malogeek]

i see AMD has chipset drivers for win7, even though they said it were officially not supported.
but only 64bit, what would happen on 32bit? with my spare office computers i always just clone the drive from the old machine to the new. using windows repair utility or "boot with dissimilar hardware" it has worked in the past eventually. Athlon XP--> AM2-->FM2 --> AM4

Break out of your routine and move into the new millenium.

 

[kallisX]

well not sure i can get all the client correspondance, accounting software info etc. ported painless to a new OS. not even sure it the software supports it. its not uncommen to find 10-15 year old systems in even major companies. because it works

 

[Reinvented]

well not sure i can get all the client correspondance, accounting software info etc. ported painless to a new OS. not even sure it the software supports it. its not uncommen to find 10-15 year old systems in even major companies. because it works

It is becoming more and more uncommon to see that stuff. So, somewhere down the line you have to either 1) upgrade your system, 2) upgrade your software, or 3) really get with the times and upgrade both. While I agree that newer isn't always better, there is no point to really having a faster system if your software is going to be holding you back in every scenario (32-bit os with an accounting software that's dated).

 

[tamz_msc]

well not sure i can get all the client correspondance, accounting software info etc. ported painless to a new OS. not even sure it the software supports it. its not uncommen to find 10-15 year old systems in even major companies. because it works

Old win32 applications *should* work on 64-bit Windows 7 or 8/8.1, but since Windows 8/8.1 isn't supported by new motherboard chipsets and M$FT is a d*** about updates on new processors, Windows 10 is your only option. But given how Windows 10 breaks everything, your best option would be to run a Windows XP/7 VM.

 

[Malogeek]

well not sure i can get all the client correspondance, accounting software info etc. ported painless to a new OS. not even sure it the software supports it. its not uncommen to find 10-15 year old systems in even major companies. because it works

So ignore the new drivers then? If you're stuck with legacy for your applications, then you're stuck with legacy for driver support as well.

 

[PhonakV30]

so Pinnacle ridge CPU (12nm Ryzen) won't work with X370 ?

http://wccftech.com/amd-12nm-ryzen-...unching-february-2018-alongside-x470-chipset/

 

[Malogeek]

so Pinnacle ridge CPU (12nm Ryzen) won't work with X370 ?

Because there's a new chipset? Not necessarily.

 

[Pick2]

Towards the end of your liked article , the chart shows Pinnacle ridge CPU on AM4 socket.

 

[IRobot23]

so Pinnacle ridge CPU (12nm Ryzen) won't work with X370 ?

http://wccftech.com/amd-12nm-ryzen-...unching-february-2018-alongside-x470-chipset/

Is this for real? I mean its says "report", but I do not see why would AMD push AM4 400 lineup before ZEN2-7nm. There still small amount of AM4 300 lineup...

 

[Topweasel]

Towards the end of your liked article , the chart shows Pinnacle ridge CPU on AM4 socket.

Well not to get nitpicky about it you can have a new chipset, one that might be required for a newer CPU, on the same socket. Look at coffee lake. It's the same 1151, but the 3xx chipset will be required for Coffee Lake even though it will support Sky Lake and Kaby Lake CPUs. That said AMD has never played that game unless they absolutely had to and specifically said the wanted the the 3xx platform to last till at least 2020. The only reason they will possibly stop there or release a new socket before that depends on whether they need to update to PCIe 4/5 or DDR5. The X470 may support newer features (more USB 3.1 is my guess) or maybe add PCIe lanes to the chipset ala Intel or maybe just be a straight up rebrand. But chances are it will not be required for Pinnacle Ridge/

 

[Insert_Nickname]

The X470 may support newer features (more USB 3.1 is my guess) or maybe add PCIe lanes to the chipset ala Intel or maybe just be a straight up rebrand. But chances are it will not be required for Pinnacle Ridge/

Given Ryzen "chipsets" are essentially just I/O break-out boxes, I'd be very surprised if there are any major changes aside from a few more 10Gbit USB, SATA3 ports or perhaps FCH PCIe 3.0 support.

All the important stuff is happening indside the Ryzen CPU.

 

[psolord]

Hello.

Saw this on another forum and I am not sure where to post this. I mean I don't know if it's real.

Someone unlocked the extra two cores of the 1600X?

https://gigglehd.com/gg/hard/1797571

What the f? Is this possible?

Anyone of you guys that built ryzen systems had any success doing that? And how?