FEATURED AMD ZEN Discussion (Previous Rumor Thread)

DaveB · Mar 17, 2017

Nagorak said:
I'm not sure which Asus X370 you have, but I'm not having any problems unzipping on my X370 Prime.

Do the AIDA tests read/write from your drive? The zip problem to sounds to me like a read or write problem with your SSD (or the board's interface with it). Do you have another one you can try out to compare? I definitely wouldn't just ignore that problem because it could be a sign of something a lot more serious, which could lead to serious data corruption.

I had three boards, MSI B350 Tomahawk, ASRock Fatal1ty Gaming K4 and the Asus Prime X370 Pro, and had three different CPUs, two 1700s and one 1700X. The large file unzip with 3DMark was a problem on all of them. So, yeah, I thought there's some issue with the NVMe/PCIe X4 interface or something in Ryzen itself. So I moved the file to a SATA SSD and tried to unzip there and got the same failure. It was no big deal with 7-ZIP doing the unzip and I had no other such issues with any other programs or normal functions. Smaller files unzip normally.

I went through all the AIDA64 Engineer benchmarks on all motherboards/CPUs with no issues, even overclocked. I posted the results here earlier in this thread. All run on the NVME drive with no data corruption at any time.

Bluefalcon13 · Mar 17, 2017

Nagorak said:
I probably wouldn't bother pushing much higher than 1.4v, and preferably under that. The last 100 MHz on these chips takes so much extra voltage that it's questionable whether it's worth the extra heat, power and likely reduced longevity of the chip. You could try for 4 GHz at 1.4, but if that's not stable I'd probably just stick to 3.9.

Also, keep in mind that the LLC settings on the boards basically increases voltage under load, to try to compensate for it dropping too low. So, it may run up to .025 volts higher than what you're set to (1.375 would actually be as high as 1.4).

- - - Updated - - -

I'm not sure which Asus X370 you have, but I'm not having any problems unzipping on my X370 Prime.

Do the AIDA tests read/write from your drive? The zip problem to sounds to me like a read or write problem with your SSD (or the board's interface with it). Do you have another one you can try out to compare? I definitely wouldn't just ignore that problem because it could be a sign of something a lot more serious, which could lead to serious data corruption.

SSD is fine in my case. Was unzipping Samsung SSD magician ironically. My aegis is rated for ddr4-3000 @1.35V. was running at ddr4-2933 @1.35v.

The boot issues described in my post were odd, and do not occur when running @ddr4-2133. Also had some issues installing applications. They would hang, and never complete. This also was resolved by going back to 2133. I'm positive it's isolated to a ram stability issue, as the symptoms all disappear @2133. I just don't know what knobbies to turn to make it right. Also, I cannot find anything from gskill with regards to the max safe voltage on the sticks. Merely 1.35v @3000 and 1.2v @2133 according to SPD.

DaveB · Mar 17, 2017

EarthDog said:
Worked fine after raising the voltage... is that fine for its stock speeds?

I'm not sure who you're responding to, but my post was the next one up. The G.Skill Aegis ran fine at the Asus Prime X370 Pro DOCP auto setting at 2400 and 2666. It only needed the voltage bump at 2933. It also ran at 2400 and 2666 at the A-XMP auto setting on the MSI B350 Tomahawk, but would only run at 2400 on the ASRock Falal1ty Gaming K4. Extra voltage did not help on either of the B350 boards.

- - - Updated - - -

Bluefalcon13 said:
SSD is fine in my case. Was unzipping Samsung SSD magician ironically. My aegis is rated for ddr4-3000 @1.35V. was running at ddr4-2933 @1.35v.

The boot issues described in my post were odd, and do not occur when running @ddr4-2133. Also had some issues installing applications. They would hang, and never complete. This also was resolved by going back to 2133. I'm positive it's isolated to a ram stability issue, as the symptoms all disappear @2133. I just don't know what knobbies to turn to make it right. Also, I cannot find anything from gskill with regards to the max safe voltage on the sticks. Merely 1.35v @3000 and 1.2v @2133 according to SPD.

I think you need to bump the voltage a bit. Anything up to 1.4V is fine, in fact, you could even go a little more but I had no memory-related instability after that small voltage bump.

Nagorak · Mar 17, 2017

Yeah, I think going up to 1.4v is necessary to get the boards to boot at higher memory speeds, unfortunately. For whatever reason just running at the stock 1.35v hangs on boot, but that small increase to 1.4 fixes it.

Johan45 · Mar 17, 2017

Does your board have a section for boot voltages? A boost there will likely get you over the hang and leave running volts at 1.35

wingman99 · Mar 17, 2017

Nagorak said:
It looks like my rebooting problem in Prime95 might be related to the Gigabyte Gaming 3. I'm not seeing it with the Asus X370 Prime. The power delivery on the Gaming 3 might just not be up to the task. It's unfortunate because I really like the Gaming 3 otherwise.

I also got some Flare X 3200, however I'm not actually convinced it's any different than the Trident Z (other than costing a lot more). I'll have to do some more comparison, but I think that high end Flare X/Trident Z/Ripjaws are all basically the same modules with different heat spreaders. We know they're all Samsung B-Die, and I'm not sure there's any more cherry picking of chips beyond that.

How is your memory working on the Asus X370 Prime? I would RMA the Gigabyte Gaming 3 to the place of purchase, Woomack is not having that problem with his.

Bluefalcon13 · Mar 17, 2017

Johan45 said:
Does your board have a section for boot voltages? A boost there will likely get you over the hang and leave running volts at 1.35

I'll have to look when I get a chance. School is whooping my rear right now. What would it be labeled as? Woomack was saying to bump vttddr some which I did, but probably not enough tbh.

Bluefalcon13 · Mar 17, 2017

So I was looking into the memory issue I am having: the gskill aegis kit is rated for 3000 @ command rate 2N. It appears the mobos currently only support 1N/1T right now. So I need to feed the beast some voltage seems the way to go.

Bluefalcon13 · Mar 17, 2017

So stupid question. Everyone is saying prime isn't the good stress test anymore, so what should I use to check stability?

Jarhead7236 · Mar 17, 2017

Bluefalcon13 said:
So stupid question. Everyone is saying prime isn't the good stress test anymore, so what should I use to check stability?

OCCT ?

Bluefalcon13 · Mar 17, 2017

Jarhead7236 said:
OCCT ?

Haven't tried that yet. For a LONG time I've just moved this 27.9 P95 folder (portable zip extraction) from one windows install to the next. I just wasn't sure what I should be using to stress test at this point cause everyone keeps saying P95 is overkill and too heavy of a load to test with...

Jarhead7236 · Mar 18, 2017

Bluefalcon13 said:
Haven't tried that yet. For a LONG time I've just moved this 27.9 P95 folder (portable zip extraction) from one windows install to the next. I just wasn't sure what I should be using to stress test at this point cause everyone keeps saying P95 is overkill and too heavy of a load to test with...

i've always used P95 as well...even "back in the day"...but i keep hearing about OCCT.

wingman99 · Mar 18, 2017

Bluefalcon13 said:
Haven't tried that yet. For a LONG time I've just moved this 27.9 P95 folder (portable zip extraction) from one windows install to the next. I just wasn't sure what I should be using to stress test at this point cause everyone keeps saying P95 is overkill and too heavy of a load to test with...

I think Prime95 v28.10 is fine my temperature only goes up to 85c with air cooling. Prime95 is a good test, it does not use system memory that much, mostly fits in cache because it is doing math calculations so the Integer and Floating point transistors switch as fast as the clock speed. Other programs don't feed the CPU data that fast do to system memory overhead, unless you are running math calculation program with AVX/FMA3.

mackerel · Mar 18, 2017

Bluefalcon13 said:
So stupid question. Everyone is saying prime isn't the good stress test anymore, so what should I use to check stability?

For CPUs which are FMA3 weak, which includes Ryzen, I've found Realbench to provide a good alternate test at finding instability quickly and effectively. However I'd use it in addition to P95, not instead of. They cover different cases.

wingman99 said:
I think Prime95 v28.10 is fine my temperature only goes up to 85c with air cooling. Prime95 is a good test, it does not use system memory that much, mostly fits in cache because it is doing math calculations so the Integer and Floating point transistors switch as fast as the clock speed. Other programs don't feed the CPU data that fast do to system memory overhead, unless you are running math calculation program with AVX/FMA3.

If the workload fits in the cache or not depends on the FFT size. For "large" FFT it certainly doesn't fit in the cache, and will stress the ram. For "small" FFT it does fit in the cache, and stresses the execution units more, but not the ram.

Also note that the release versions of P95 (28.xx) don't detect Ryzen correctly, and doesn't use FMA3. There is a test release of 29.1 at link below if you want to try an updated one. The CPU detection is updated so it handles SMT now, and it can use FMA3 transform. It seems to give slightly higher benchmark numbers than Bulldozer/K10 it was using in 28.xx.
http://www.mersenneforum.org/showpost.php?p=454570&postcount=93

MNMadman · Mar 18, 2017

DaveB said:
I had three boards, MSI B350 Tomahawk, ASRock Fatal1ty Gaming K4 and the Asus Prime X370 Pro, and had three different CPUs, two 1700s and one 1700X. The large file unzip with 3DMark was a problem on all of them. So, yeah, I thought there's some issue with the NVMe/PCIe X4 interface or something in Ryzen itself. So I moved the file to a SATA SSD and tried to unzip there and got the same failure. It was no big deal with 7-ZIP doing the unzip and I had no other such issues with any other programs or normal functions. Smaller files unzip normally.

I went through all the AIDA64 Engineer benchmarks on all motherboards/CPUs with no issues, even overclocked. I posted the results here earlier in this thread. All run on the NVME drive with no data corruption at any time.

Just found out my system has the same damn problem. Wanted to do some comparative benches. Running on what I thought was a fully stable overclock -- tested with several multiple-hour runs of AIDA64 (CPU, FPU, Cache, System Memory, GPU), RealBench (16GB), Prime95 (Blend), Y-Cruncher (defaults), and some actual gaming -- I can't get the damn 3DMark or PCMark files to unzip. Frustration ensues. Even at full Auto settings and freshly-downloaded files it still gives an "unspecified" error. And this isn't an NVMe problem -- it's unzipping from one folder on the SATA HD to another folder on the same SATA HD. No problems with other zip files either.

Not only do I have a dud overclocker, but the damn thing can't even unzip some files correctly.

Will have to try 7-Zip after I have calmed down and gotten some sleep...

DaveB · Mar 18, 2017

MNMadman said:
Not only do I have a dud overclocker, but the damn thing can't even unzip some files correctly.

Will have to try 7-Zip after I have calmed down and gotten some sleep...

I think it's pretty clear that the Ryzen 7 series just do not OC very much - 4 GHz to 4.1 GHz seems to be the top, while 3.8 GHz to 3.9 GHz is the norm. On the upside, the performance at 3.8 GHz to 3.9 GHz is excellent and we just need Asus, ASRock, Gigabyte, MSI to get their acts together with memory compatibility. That's assuming, it isn't the Ryzen IMC that's the issue.

MNMadman · Mar 18, 2017

DaveB said:
I think it's pretty clear that the Ryzen 7 series just do not OC very much - 4 GHz to 4.1 GHz seems to be the top, while 3.8 GHz to 3.9 GHz is the norm.

Oh yeah, I know. The reason I call mine a dud is the voltage. Most that hit 3.8GHz can do it on 1.2v or so while mine requires 1.351v. Mine can only do 3.65GHz on 1.2v.

I got 3DMark to install, but PCMark failed the install. Re-downloading again.

Edit: PCMark re-downloaded, 7-Zip unzipped with no errors (there were two the first time), installed correctly. I'm going to try re-downloading 3DMark and see if I can get an error-free 7-Zip unzip. It installed fine, but I'm curious...

Edit2: 3DMark re-downloaded ... again. This time got an error-free 7-Zip unzip (three the first time). Uninstalled existing 3DMark and re-installed because of that.

Bluefalcon13 · Mar 18, 2017

MNMadman said:
Oh yeah, I know. The reason I call mine a dud is the voltage. Most that hit 3.8GHz can do it on 1.2v or so while mine requires 1.351v. Mine can only do 3.65GHz on 1.2v.

I got 3DMark to install, but PCMark failed the install. Re-downloading again.

Edit: PCMark re-downloaded, 7-Zip unzipped with no errors (there were two the first time), installed correctly. I'm going to try re-downloading 3DMark and see if I can get an error-free 7-Zip unzip. It installed fine, but I'm curious...

Edit2: 3DMark re-downloaded ... again. This time got an error-free 7-Zip unzip (three the first time). Uninstalled existing 3DMark and re-installed because of that.

What's your RAM at and what's its current timings vs rated speed/timings? It seems to me due to the forced 1T command rate, most modules are pushed harder because of that.

All my problems with memory went away after I bumped my Vsoc to 1.1V, and gave my RAM a smidge more voltage than it was asking for in the XMP profile. Was very stable when I started OCing my CPU. For me, my temps are my limit. The wraith spire is barely handling the 3.8ghz OC. I'm just waiting on my 212 adapter, but it should be in by Monday.

MNMadman · Mar 18, 2017

Bluefalcon13 said:
What's your RAM at and what's its current timings vs rated speed/timings? It seems to me due to the forced 1T command rate, most modules are pushed harder because of that.

All my problems with memory went away after I bumped my Vsoc to 1.1V, and gave my RAM a smidge more voltage than it was asking for in the XMP profile. Was very stable when I started OCing my CPU. For me, my temps are my limit. The wraith spire is barely handling the 3.8ghz OC. I'm just waiting on my 212 adapter, but it should be in by Monday.

All of this happened on the 0038 BIOS, which is the 2T Command Rate version (they do have the 1001 BIOS which is the 1T Command Rate version, but I haven't used that yet). 0038 and 1001 are identical, except for the Command Rate.

And it happened even at all Auto settings (2133, 3.0GHz, etc).

Bluefalcon13 · Mar 18, 2017

MNMadman said:
All of this happened on the 0038 BIOS, which is the 2T Command Rate version (they do have the 1001 BIOS which is the 1T Command Rate version, but I haven't used that yet). 0038 and 1001 are identical, except for the Command Rate.

And it happened even at all Auto settings (2133, 3.0GHz, etc).

Ahhh. Crosshair bios with 2T. Hopefully they get that over to us X370-Pro owners soon enough.

FEATURED AMD ZEN Discussion (Previous Rumor Thread)

Senior Member

Member

Senior Member

Senior Member

Benching Team Leader Super Moderator

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Senior Member

Member

Member

Member

Member

Similar threads