SOME RYZEN 7 1700 OVERCLOCKING!

Matrix · 1 May 2017 at 17:59

gavinh87 said:
Glad to help. They are some of the best I've used and they outperform the corsair LED's costing twice as much.
I used to like the arctic cooling F12's but they don't last long enough. I've got some 140mm noctua sat here but they are ugly as hell so I wont use them, quiet though.

Yeah there brill on my AIO no humming noise i could hear with the last ones, might get another one to replace the front 120mm fan ( out take) thought at the moment i only have the 2 AIO cooler fans on temps around 28/30C browsing and running Vm's

haszek · 1 May 2017 at 22:20

With the latest asus ch6 bios (1107) I can run 4x8gb at 3200

(using bclk and 18-16-16) which is pretty cool and I don't even have B-die (I have E-die)....the only minus is that it requires 1.44 DRAM voltage:/ (which seems kinda high for my taste)

When I tried 3000 (16-16-16) only required 1.37V and looks pretty stable as well, so I'll leave it at that and wait for next update.

ozzyo99 · 2 May 2017 at 00:25

haszek said:
With the latest asus ch6 bios (1107) I can run 4x8gb at 3200 (using bclk and 18-16-16) which is pretty cool and I don't even have B-die (I have E-die)....the only minus is that it requires 1.44 DRAM voltage:/ (which seems kinda high for my taste)

When I tried 3000 (16-16-16) only required 1.37V and looks pretty stable as well, so I'll leave it at that and wait for next update.

Nothing wrong with 1.44v DRAM. Up to 1.5 is safe 24/7.

Did you try setting 1.4 in DRAM and 1.45 DRAM Boot?

haszek · 2 May 2017 at 05:40

ozzyo99 said:
Nothing wrong with 1.44v DRAM. Up to 1.5 is safe 24/7.

Did you try setting 1.4 in DRAM and 1.45 DRAM Boot?

Yeah I tried :/ it was booting but once in Windows and I hit restart PC, whole computer was switching off instead, then it was starting up after a second but without pstates settings.

edit;

Also this gets me thinking, when Skylake came out with DDR4 all guides were saying do not exceed 1.35V for memory for normal daily usage (it was in the asus raja guide, in the TT guide). Why the change now that people wouldn't worry for anything below 1.5V?

Jokester · 2 May 2017 at 16:43

It was believed that the Skylake IMC couldn't handle the higher voltage.

haszek · 2 May 2017 at 17:00

ah right, cpu related, not ddr4.
thx

antonusklesk · 2 May 2017 at 21:33

So after a bit more tinkering, I believe it may be using XMP that causes the failed POST from cold boot - I've set RAM manually and fingers crossed, it booted at the first time of asking.

Also, pretty pleased with my little 1700, happily stable on 3.8GHz @ 1.25v.

NeoWave · 3 May 2017 at 10:01

antonusklesk said:
So after a bit more tinkering, I believe it may be using XMP that causes the failed POST from cold boot

Don't know what mobo you have but on the MSI B350 Tomahawk I have the latest Bios update causes the board to fail to boot from cold using XMP. Imputing the exact same values manually results in no problem.

HubbleTrouble · 3 May 2017 at 10:56

I have found a curious stability issue that I would be interested to see if others can replicate.

When testing stability I can get my system to pass several hours of IBT/prime95/OCCT/RB. Happy that I have a stable OC I then run a few benchmarks. Everything is fine until I run PCMark 8 where it freezes every time. If run at stock it passes every time. Experimenting further I find that at higher vcore levels it fails, or at a lower vcore without LLC.

For example: 3825Mhz at 1.38750V passes PCMark8, 3825Mhz at 1.4V fails. All other stability tests pass at both vcore levels. At 3850Mhz in need 1.45v to pass normal stability tests, but it fails PCMark 8.

To replicate use PCMark 8 basic edition, "run conventional" - the test will freeze the system when running one of the three passes in the home / photo editing section - changing bright/contrast levels on the foliage. https://www.futuremark.com/benchmarks/pcmark8

Now you might think this is just an aberration for a single R1700, but unhappy with the silicon lottery I did spin the wheel twice. My second R1700 can clock higher while avoiding this issue, but it still does suffer (3.7Ghz safe on original, 3.825Ghz on current). System is Taichi x370, 32GB HyperX 2400/C15 OC to 2666/C16 (Issue occurs with and without mem OC). Tdie doesn't exceed 65C.

It seems that conventional stress testing may be OK for testing the system under load, but perhaps this doesn't fully exercise other aspects that can fail when OCing - perhaps such as rapid dynamic clock/vcore changes that this particular test triggers?

I would be interested if other people can replicate this test on their setups. 2 CPUs on a single mobo/mem combo isn't exactly a good sample to draw conclusions from. If others can replicate this issue then at least we have found a straightforward easily repeatable test for a particular type of OC instability on Ryzen.

antonusklesk · 3 May 2017 at 12:37

NeoWave said:
Don't know what mobo you have but on the MSI B350 Tomahawk I have the latest Bios update causes the board to fail to boot from cold using XMP. Imputing the exact same values manually results in no problem.

Same board here, and yep that seems to be the case. Can't really think of a reason why though. Roll on more BIOS updates

eyetrip · 3 May 2017 at 12:48

HubbleTrouble said:
I have found a curious stability issue that I would be interested to see if others can replicate.

When testing stability I can get my system to pass several hours of IBT/prime95/OCCT/RB. Happy that I have a stable OC I then run a few benchmarks. Everything is fine until I run PCMark 8 where it freezes every time. If run at stock it passes every time. Experimenting further I find that at higher vcore levels it fails, or at a lower vcore without LLC.

For example: 3825Mhz at 1.38750V passes PCMark8, 3825Mhz at 1.4V fails. All other stability tests pass at both vcore levels. At 3850Mhz in need 1.45v to pass normal stability tests, but it fails PCMark 8.

To replicate use PCMark 8 basic edition, "run conventional" - the test will freeze the system when running one of the three passes in the home / photo editing section - changing bright/contrast levels on the foliage. https://www.futuremark.com/benchmarks/pcmark8

Now you might think this is just an aberration for a single R1700, but unhappy with the silicon lottery I did spin the wheel twice. My second R1700 can clock higher while avoiding this issue, but it still does suffer (3.7Ghz safe on original, 3.825Ghz on current). System is Taichi x370, 32GB HyperX 2400/C15 OC to 2666/C16 (Issue occurs with and without mem OC). Tdie doesn't exceed 65C.

It seems that conventional stress testing may be OK for testing the system under load, but perhaps this doesn't fully exercise other aspects that can fail when OCing - perhaps such as rapid dynamic clock/vcore changes that this particular test triggers?

I would be interested if other people can replicate this test on their setups. 2 CPUs on a single mobo/mem combo isn't exactly a good sample to draw conclusions from. If others can replicate this issue then at least we have found a straightforward easily repeatable test for a particular type of OC instability on Ryzen.

That particular test may be enough to push board/cpu to power limit - does it shut down?

I used to have similar when I was doing occt testing.

HubbleTrouble · 3 May 2017 at 13:07

eyetrip said:
That particular test may be enough to push board/cpu to power limit - does it shut down?

I used to have similar when I was doing occt testing.

No shutdown, it actually locks up the PC with the display still visible. This is definitely a different type of crash from the typical "black screen of death" when you push your OC too far. Looking at HWInfo the CPU power draw of PCM8 is very low compared with OCCT and the like.

gavinh87 · 3 May 2017 at 15:19

HubbleTrouble said:
No shutdown, it actually locks up the PC with the display still visible. This is definitely a different type of crash from the typical "black screen of death" when you push your OC too far. Looking at HWInfo the CPU power draw of PCM8 is very low compared with OCCT and the like.

Are you using the ryzen power plan?
If so change to one of the windows ones.

gavinh87 · 3 May 2017 at 17:01

I've found an AVX version of IBT. The output from this is much higher than the one we have been using. It also taxes the CPU more and thus produces more heat, you have been warned!

I've uploaded it to mediafire for those interested. http://www.mediafire.com/file/bm6g387ht7wbtdm/IBT_AVX.zip

Also note that this is my 2nd 1700 and neither can handle this above 3.8ghz no matter the voltage.

HubbleTrouble · 3 May 2017 at 23:34

gavinh87 said:
Are you using the ryzen power plan?
If so change to one of the windows ones.

I was using Ryzen balanced. I tried both high performance and balanced windows plans with the higher voltage and it crashed just the same. With the lower voltage and windows balanced it continued to work as before. So this doesn't seem to be a factor. Would love to see people replicate the test.

gavinh87 · 4 May 2017 at 00:36

HubbleTrouble said:
I was using Ryzen balanced. I tried both high performance and balanced windows plans with the higher voltage and it crashed just the same. With the lower voltage and windows balanced it continued to work as before. So this doesn't seem to be a factor. Would love to see people replicate the test.

I'll give it a go on Sunday when I have a bit more time. Are you overclocking your memory?

muon · 4 May 2017 at 00:58

Ryzen Balanced doesn't perform as well as High Performance for me.

The place it is very evident is the Timespy CPU test in 3dmark.

HubbleTrouble · 4 May 2017 at 02:04

gavinh87 said:
I'll give it a go on Sunday when I have a bit more time. Are you overclocking your memory?

Yes, but have tried with and without OC and it makes no difference to pass or fail. Dual rank 32GB 2400/15 OC 2666/16.

LtMatt · 4 May 2017 at 08:57

muon said:
Ryzen Balanced doesn't perform as well as High Performance for me.

The place it is very evident is the Timespy CPU test in 3dmark.

I get a lower score with High Performance Vs Ryzen Balanced, but it is within margin of error.

muon · 4 May 2017 at 09:49

AMDMatt said:
I get a lower score with High Performance Vs Ryzen Balanced, but it is within margin of error.

I get thousands lower under Ryzen balanced.