Inexpensive RELIABLE video card for AMD Ryzen 2700 mobo

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

hz01938

Proper
Dec 25, 2018
71
0
130
First, I do not play games at all. All I need is a reliable video card.

Here is my system:

3HIdcuH.png


I am suspecting my current GTX 1050 card does not get along with the system. It got along with my previous system (Intel Core i5). I heard Nvidia drivers are relatively buggy. True?

Is Radeon HD 6450 too old?

Probably the most graphics intensive task that my computer does is processing photos from my Canon camera occasionally. I assume any video card made in the past 10 years can do it.
 

hz01938

Proper
Dec 25, 2018
71
0
130

Yes, I did a clean install on a formatted Samsung 970 EVO SSD a few days ago. As I mentioned earlier, BSOD started to happen once every few hours on this clean install vs once every few days for the old install. Isn't it strange.
I am doing Prime95 memory test now. I will let it run overnight.
 
Power off, pull the two memory modules that are in the A1 and B1 slots, leaving only the two modules in the A2 and B2 slots. Use the system for a while to see if the problems continue. Might also download and run ASUS Realbench. Choose the stress test option. Run for 8 hours. Choose 8GB in the memory amount field, since you will have 16GB installed after removing the other two modules.

Not only does Ryzen have a VERY problematic tendency to not like memory modules that use less expensive ICs (Memory chips), it is GREATLY compounded when running four modules vs only two. Try the above, see if you still have issues. I think you have two problems here. One, those are actually somewhat budget memory for what Ryzen prefers to see, not "budget" in the grand scheme of things, but budget for enthusiast memory, and they likely use a lower quality IC on those modules. They do not seem to be USER verified on Ryzen, at the advertised speed, like these are:

https://www.reddit.com/r/Amd/comments/62vp2g/clearing_up_any_samsung_bdie_confusion_eg_on/


So even if they are on the motherboard QVL, that ONLY means they were validated to be compatible with that board using a pair (Not a quad set) of modules and only at the default SPD of 2133mhz. The QVL lists, even though they LIST faster memory modules, are not a validation that they will RUN at the ADVERTISED or profile speed. Only that they "work" on that board, at all.


Secondly, quad modules. That can be a problem for ANY motherboard, on ANY platform, and that isn't something new. Running quad modules on a dual channel board (Usually not a problem on boards with triple, quad, or higher numbered memory channel architecture) can always be problematic. You will never see ANY memory manufacturer state that any of their memory kits will 100% work at the advertised speed with four modules installed on a dual channel board. They certainly CAN, with some amount (sometimes a little, sometimes with only MUCH effort in tuning the memory and CPU configuration) of fiddling with the memory settings in the BIOS, but it is always an iffy proposition with these platforms, and it is DOUBLY so with Ryzen which really doesn't like quad memory configurations at high speed.

Try two modules, if that works, we can move on to running all four modules, but at the default speed of 2133mhz, at which time we can begin moving up the ladder if that works, increasing speed and voltage where necessary until you reach a plateau where instability begins to creep back in again. Honestly, if the board supports it, when 32GB is required this is exactly why I always recommend using the least number of modules necessary, so 2 x16GB for 32GB, which is really no more stressful on the memory controller than 2 x8GB, and less likely to cause these kinds of issues OR to have one or more modules that are not "playing nice" with the others.

Which is the other possibility, that there is one, or two memory modules that have a problem. If using only 2 modules, you still have a problem, try swapping one of them out for one of the ones you removed. If you still do, swap out the other one for the remaining module you removed.

If these did not all come together in a single kit, try to keep the ones that came together, together, if possible.

Lots more on this in my guide here including memory specific test procedures for use in testing memory configurations in Prime95:

*Resolving memory problems and setting up XMP/DOCP/AMP profiles
 

hz01938

Proper
Dec 25, 2018
71
0
130

First, I am sorry for misleading you to believe there are 4 memory modules. I just copied and pasted the information from my order. I should have emphasized that memory configuration is 16GB x 2.
I have run Prime95 overnight, and no issues have been found. For memory speed, I have been using "auto".
Thank you for all the detailed explanation. I will run Realbench next and report back.
 

hz01938

Proper
Dec 25, 2018
71
0
130

I have just done the stress test per your suggestion. It stopped at 11 min. Here are the results:
cUMSbcA.png


The log.

The only content in bmps.crash.txt:
# Blender 2.68 (sub 0), Revision: 58537
I will try to lower the speed of the memory to see if it makes any difference.

 

hz01938

Proper
Dec 25, 2018
71
0
130

Thank you for the information.
I replaced 1050 with Radeon RX 460 more than a week ago, and it has not made any difference.

I finished the stress test with DDR4 speed at 1333 MHz without the unstable problem.
HSgkucv.png

I will do a longer test for 1600 MHz.
 
Seriously, take out two sticks. Then retest and choose the "Up to 8GB" memory option. Do that with the memory at the full speed profile setting. See what happens. Be sure the two sticks you leave in the motherboard are installed in slots A2 and B2. You will be putting the memory back in later, this is just to determine if using four sticks are the bigger problem or if it is the actual choice of memory module and trying to run at full speed that is the problem.

There are a LOT of users out there who cannot ever achieve full speed operation on even two sticks, dependent on WHICH memory modules they have installed, and have had to resort to a lowered speed of around 2933mhz but it's different for everybody. That last test you did, was not a pass, it was a fail. Not sure why Realbench is saying "Stress test passed" when clearly there is an error due to instability earlier in the test.

Regarding the motherboard issue, I really doubt there is any problem with your motherboard, aside from maybe the issue with the audio driver, but problems with memory configuration can certainly be a cause of driver issues since every driver makes use of memory, some more than others, but all in some way or another, so any instability in the memory configuration can definitely cause crashing driver problems although I would usually believe it to be more random and not always the same driver. But it certainly could be.
 
Yes, increasing voltage is one of the steps, and in this case that works as you've skipped the other steps. Fortunate, but not actually the usual "way to get there". In this case though, forgivable. I did mention increasing the voltage earlier but didn't expect to make that suggestion a recommendation until we had determined if it could be stable with two sticks first.

So that's cool anyhow that it now passes. The memory voltage should definitely have been 1.35v from the start though if running the memory at their full advertised profile speed. If you did, or do, not have the memory actually set to it's XMP/DOCP profile in the BIOS, that is usually where you want to start in which case the voltage SHOULD have already been AT 1.35v, but if it's stable now, that's great. There might still be some minor improvements in the primary and secondary timings that could be tightened up by using the profile configuration rather than a manual one unless you've already tightened timings manually but considering the problems you've had already I don't think I'd worry about that at all until you've determined everything else is working correctly, and maybe not even then.

Let's see if you have any further crashes or errors now that you've increased the DRAM voltage. Whole problem might be solved at this point.
 

hz01938

Proper
Dec 25, 2018
71
0
130

Thank you for the elucidation and the suggestion of RealBench that helped to reach this step. I knew nothing about XMP/DOCP until today because that was the only option other than dropping the memory to the bottom regarding RAM. I still have only basic conceptual understanding of it.
There is performance drop, but I am not worried about it all at this point of time. I will see if this is stable first, then come back to it as you suggested.
RealBench benchmark test results:
Original with unstable configuration:
Mon Feb 4 2019
07:15:30

Image Editing: 60076
Time: 84.692

Encoding: 239412
Time: 40.048

OpenCL: 37693
KSamples/sec: 757

Heavy Multitasking: 150977
Time: 64.778

System Score: 114500

With the latest DOCP profile:

Mon Feb 4 2019
12:18:12

Image Editing: 45694
Time: 111.349

Encoding: 245117
Time: 39.116

OpenCL: 37693
KSamples/sec: 751

Heavy Multitasking: 60427
Time: 161.848

System Score: 89694
 
That is because there were probably MUCH tighter memory timings with the memory at the full speed but using the non-DOCP profile timings, thus much lower latency, than with the profile enabled, however, those performance scores are misleading since you could not possibly have maintained that performance AND maintained stability. I guarantee that with the default timings, even if you added voltage, you could almost certainly not remain stable for a full 8 hours of Realbench or 4 passes of Memtest86, and if you can't, you can't run your system like that.

Memory corruption of data due to instability is a SERIOUS and REAL issue, that would result in the whole system's file structure being entirely corrupted at some point. We sure don't want that. So a performance hit is preferably, by miles and miles, plus a stable system that doesn't continuously crash is a good thing too. :)

There are of course to some degree, things you can do to tighten that up a bit later. In some cases you can manually make adjustments to the timing configuration but when and if you do, it is extremely critical to test the system by running both Memtest86 and Realbench, after every change to the primary or secondary timings, to ensure that the memory is still stable at those settings. Sometimes a setting simply cannot be stable. Other times you might be able to retain that setting by increasing voltage by a small amount such as .005-0.20v DRAM voltage increase and retesting.

That's a whole other conversation though. If you can get the memory to run at the DOCP profile advertised and expected for your memory, as outlined as the advertised speed, latency and voltage, and be stable, that is the most you can expect or desire for most systems without a great deal of fiddling around and tens of hours of tuning.
 
I would also download CPU-Z and look at both the memory and SPD tabs. You will need to choose each of the slots, on the SPD tab, or memory tab, can't remember which it is that has the drop down for each DIMM slot, to make sure that ALL memory is running at the expected speed and that all memory is showing to be running in dual channel. I would take screenshots of each tab and each DIMM slot information, and post it so I can look at it to make sure everything is correct. If you were running in dual channel before, and are not now, that would also potentially be a reason for the drop in performance numbers.
 

hz01938

Proper
Dec 25, 2018
71
0
130


It has been trouble free for over 10 hours now. When everything is stabilized, I will report some other minor problems that I thought were unrelated to the BSOD, but now think they were all caused by the same culprit. I will keep this configuration until this weekend barring any new issues.
Here are the screenshots of CPU-Z per your request (only two RAM slots are used currently as described earlier):
jsOeUnl.png

LDCO4Ui.png

BTG3vQi.png
 
So, that shows you ARE running in dual channel, but you are NOT running at the DOCP profile setting. Did you save settings when you exited the BIOS after enabling DOCP? You might look at me like "WTF dude, how dumb do you think I am?" but you'd be surprised how many times I've gone through pages of troubleshooting with somebody only to find out that they were going into the BIOS and making changes, but then not saving the settings.

Regardless, if you DID save the settings before exiting, the configuration was apparently not stable and the system reverted to the default setting of 2133mhz (1058.8mhz x2, for DDR, double data rate) instead of running at 3000mhz (1500mhz x2) as you can see in your first screen shot. Going from 3000mhz down to 2133mhz on Ryzen would certainly account for the performance loss you are seeing.

You may need to reset the BIOS to default settings, save, exit, back into BIOS, set DOCP, save settings, exit, back into BIOS and verify that the settings are sticking. Something definitely isn't right if it's supposed to be at the DOCP profile settings right now, but isn't. Unless you changed things per my earlier instructions and went away from the DOCP setting. With DOCP set, after you save and go back into the bios, you should be showing 3000mhz and 1.35v, if not, it may need to be configured manually leaving all the timings on auto but setting the speed manually to 3000mhz and the voltage to 1.35v.

So, do you have 64GB of memory, or 32GB? Do you have 4 16GB sticks? I realize only two are installed right now, which is why I ask because it seems earlier, before I said to remove two of them, you were showing only 32GB in your screenshot as well, which means maybe two of those sticks were not working earlier.

Or do you only have two 16GB sticks, period, and they are all that has been installed this whole time? And lets just verify that they ARE in the A2 and B2 slots, yes? Sorry, but this thread has gone through several metamorphosis of it's life so it's hard to keep track. LOL.
 

hz01938

Proper
Dec 25, 2018
71
0
130


I am terribly sorry but I should have mentioned that I did not use the default DOCP RAM speed of 3200 MHz. Instead, I selected Auto as my screenshot shows. Yes, I did save the changes before existing BIOS.
My original intention was changing the voltage only, not over-clocking the RAM. I found that there was no way to change the voltage without using DOCP, so I selected DOCP, and accepted all the changes except the RAM speed because I thought high speed would be detrimental to reliability. My current thought is that if everything is stabilized, I will increase the RAM speed gradually.

So far so good. I let the PC run overnight and did another stress test. I will do the 8-hour stress test tonight.

 

hz01938

Proper
Dec 25, 2018
71
0
130

Thank you for the confirmation. It has been BSOD free for over 30 hours now.
To have more certainty about the RAM speed, I changed it from Auto to 2400 MHz this morning.
JQ0EmNM.png

K2lTjJ7.png

It passed the stress test without any problem:
wNTj12F.png

Here are the new benchmark scores:

XML:
Tue Feb 5 2019
18:47:30

Image Editing: 137587
Time: 36.98

Encoding: 241797
Time: 39.653

OpenCL: 37693
KSamples/sec:  757

Heavy Multitasking: 158735
Time: 61.612

System Score: 136414

 
I'd go ahead and set DOCP, set full speed 3000mhz (If it doesn't automatically set the speed, which you might ALSO need to set to DOCP) set the voltage to 1.35v (Also, if it doesn't do it automatically when you set DOCP) and then test like that.

Usually when you set XMP (DOCP on ASUS, since they apparently don't want to pay Intel any royalty fees for using the term XMP), it will set the memory speed, timings and voltage, automatically, but it may not on your platform. As I said much earlier in this thread, I haven't had much opportunity to do any advanced configuration fiddling with the memory settings on Ryzen yet. I'd like to buy a Ryzen setup just to mess around with but it's hard to justify right now, and might be a moot point when Zen 2 comes out anyhow as memory compatibility is supposed to be much improved from Ryzen gens 1 and 2.

I do want to ask again, because I don't think you ever answered on this account, do you actually ONLY have two modules, total, or do you just have two installed right now? I was unsure if you had two 16GB modules or four? If you have four, leave it with only two for now while you are still trying to test and get up to full speed. If you have more, we can worry about those later.
 

hz01938

Proper
Dec 25, 2018
71
0
130

I know this thread has been too long to be clear about everything. Yes, I stated twice before that there are only two modules (32 GB total):
I should have emphasized that memory configuration is 16GB x 2.
only two RAM slots are used currently as described earlier)
 

hz01938

Proper
Dec 25, 2018
71
0
130

I really appreciate your unbelievable tenacity in holding my hands through this journey. I will report back after bringing the RAM speed to the maximum.
 
Man, this is what I do. I'd rather do one thread like this than ten other threads where I just know that the second something doesn't go the way the OP wants it to, they're going to turn venomous and try to crap all over the help you've tried to give them. It's no problem. My pleasure. Glad we've gone from "This whole system is a pile of crap and none of my applications want to work" to "stable for over 30 hours". That's progress anyhow.