Ryzen Owners Zen Garden

[YautjaLord]

After seeing all your SoC related comments, decided to check the safest SoC voltage, best i got was posted in KitGuru 1700X review - 1.25v? Whatdafuk?

Lowered LLCs a tad, from Turbo to High, lower but stabler Cinebench R15 score. SoC is @ 1.0v.

 

[robot zombie]

Alright, I give up. The limitations between the lesser b-die, dated BIOS, and possibly memory controller are unavoidable. This is the best I can do without hitting constant BSODs or just failing stability tests instantly. Like, starts tossing up errors the moment I start prime95. It's a shame. I really feel like I should be able to get just a little more out of this hardware. Maybe if ASUS ever figures out their troubles with AGESA, I can.

I did try 3200/CL14, but the performance is marginally worse, it's a slower 3200/CL14, if that makes sense. Actually just scores worse across the board. Maybe later I'll run it to show. What I'm posting here is basically the timings for that scaled up to what would perform the same at 3400, with that 66mhz speed bump. So I wind up with similar latency, but a little faster speed.

Maybe there's a better sweet spot in there. I can't find it, though. There's shockingly little between this specific config and which combinations of faster speeds/looser timings or slower speeds/tighter timings will actually be stable. A little nudge either way and stability drops off of a cliff :/ And it's always like I'm one step away. But 10 steps later nothing has changed.

There's no way it should ever be that touchy. I'm not exactly pushing for extremes. I'm really thinking it's the mobo. Maybe I just don't know enough yet. But I feel like something has to be missing here. The wall I'm hitting is too brutal. It took me a week to get to this point. :/ I dunno, maybe I'm expecting too much. I know 3600/CL17 b-die isn't the best. Don't get me wrong, either, it still performs great - probably much better than what most people out there have. I'm still happy with it. Just feels like kind of a letdown to be held back by something I just can't pin down, knowing there should be a way to go another rung up. One can only bang his head against the wall so much. I feel like I've done my time reading, tinkering, clearing CMOS, restoring backups.

Maybe one day I'll pick it up again. But right now I am soo over it

EDIT: Just for posterity. 3200/CL14...

No matter what I may do, timings do not get sufficiently tighter. And speeds don't get sufficiently faster at similar timings. I gain nothing past that 3466/CL16 config. If I could do something like a loose 3200/CL12 or even CL13 I might be thinking differently. I also tried for 3400/CL14 and even 3400/CL15, but never managed to get it stable.

Power is another factor... ...this 3466/CL16 config runs at 1.41v and 1.1 SoC. Anything performing better requires significantly more to post, let alone begin to pass stress tests. I played around with termination, secondary voltages, powerdown/geardown, everything. I started with the lowest reasonable voltage for both DRAM and SoC, going up in minimal increments. And I'm telling you, minimum SoC is like 1.15 or even higher, while DRAM needs to be close to or at 1.5v. Even if I could tweak the timings to make those faster, high-power configs stable, it's not worth the extra 5-7C in CPU temps for what is maybe a 2-5% RAM performance boost.

Just goes to show, there's a lot more to it than just speed and CAS latency. Seems the way to get the most out is to pick the specific speed/CL that lands you the best subtimings, with a reasonable balance of all three. It doesn't always make sense to shoot for the fastest speed you can do with the lowest possible primary timings because of sacrifices made elsewhere. There's a sweet spot for every hardware combo. Shooting for the sweet spots other people hit may be a misdirection.

Lesson here is to look at what YOU are seeing, not other people. If I did as people generally recommend and shot for those pretty low primary timings and slower speed, I'd be losing out in terms of overall speed:latency. And that's because the subtimings for my sticks to hit that are necessarily too loose. I can just as easily kick on geardown, up the primaries/speed, and hold tighter subtimings for better performance at absolutely no sacrifice to overall latency.

How this shakes down in specific applications is another question. And a much harder one to answer...

One thing I can say, and if anybody's curious I have my 3200/CL14 profile saved to run some more benchmarks, but CPU performance actually goes down measurably with that profile over my 3466/CL16 one.

 

[YautjaLord]

Don't exactly remember if i put the pic of the beast in this thread before, so....:

EK-CryoFuel & 3+ litres of distilled water taken care of. Don't mind the ASUS GTX 760 (dead unfortunately ) & Samsung 850 Pro SSD on the right side of the pic.

P.S. Shot with Galaxy A6 cam, 16MP, 4:3, 4608x3456 res. Nice lil cam & phone, love it.

 

[AlwaysHope]

Alright, I give up. The limitations between the lesser b-die, dated BIOS, and possibly memory controller are unavoidable. This is the best I can do without hitting constant BSODs or just failing stability tests instantly. Like, starts tossing up errors the moment I start prime95. It's a shame. I really feel like I should be able to get just a little more out of this hardware. Maybe if ASUS ever figures out their troubles with AGESA, I can.

I did try 3200/CL14, but the performance is marginally worse, it's a slower 3200/CL14, if that makes sense. Actually just scores worse across the board. Maybe later I'll run it to show. What I'm posting here is basically the timings for that scaled up to what would perform the same at 3400, with that 66mhz speed bump. So I wind up with similar latency, but a little faster speed.

Maybe there's a better sweet spot in there. I can't find it, though. There's shockingly little between this specific config and which combinations of faster speeds/looser timings or slower speeds/tighter timings will actually be stable. A little nudge either way and stability drops off of a cliff :/ And it's always like I'm one step away. But 10 steps later nothing has changed.

There's no way it should ever be that touchy. I'm not exactly pushing for extremes. I'm really thinking it's the mobo. Maybe I just don't know enough yet. But I feel like something has to be missing here. The wall I'm hitting is too brutal. It took me a week to get to this point. :/ I dunno, maybe I'm expecting too much. I know 3600/CL17 b-die isn't the best. Don't get me wrong, either, it still performs great - probably much better than what most people out there have. I'm still happy with it. Just feels like kind of a letdown to be held back by something I just can't pin down, knowing there should be a way to go another rung up. One can only bang his head against the wall so much. I feel like I've done my time reading, tinkering, clearing CMOS, restoring backups.

Maybe one day I'll pick it up again. But right now I am soo over it

EDIT: Just for posterity. 3200/CL14...

No matter what I may do, timings do not get sufficiently tighter. And speeds don't get sufficiently faster at similar timings. I gain nothing past that 3466/CL16 config. If I could do something like a loose 3200/CL12 or even CL13 I might be thinking differently. I also tried for 3400/CL14 and even 3400/CL15, but never managed to get it stable.

Power is another factor... ...this 3466/CL16 config runs at 1.41v and 1.1 SoC. Anything performing better requires significantly more to post, let alone begin to pass stress tests. I played around with termination, secondary voltages, powerdown/geardown, everything. I started with the lowest reasonable voltage for both DRAM and SoC, going up in minimal increments. And I'm telling you, minimum SoC is like 1.15 or even higher, while DRAM needs to be close to or at 1.5v. Even if I could tweak the timings to make those faster, high-power configs stable, it's not worth the extra 5-7C in CPU temps for what is maybe a 2-5% RAM performance boost.

Just goes to show, there's a lot more to it than just speed and CAS latency. Seems the way to get the most out is to pick the specific speed/CL that lands you the best subtimings, with a reasonable balance of all three. It doesn't always make sense to shoot for the fastest speed you can do with the lowest possible primary timings because of sacrifices made elsewhere. There's a sweet spot for every hardware combo. Shooting for the sweet spots other people hit may be a misdirection.

Lesson here is to look at what YOU are seeing, not other people. If I did as people generally recommend and shot for those pretty low primary timings and slower speed, I'd be losing out in terms of overall speed:latency. And that's because the subtimings for my sticks to hit that are necessarily too loose. I can just as easily kick on geardown, up the primaries/speed, and hold tighter subtimings for better performance at absolutely no sacrifice to overall latency.

How this shakes down in specific applications is another question. And a much harder one to answer...

One thing I can say, and if anybody's curious I have my 3200/CL14 profile saved to run some more benchmarks, but CPU performance actually goes down measurably with that profile over my 3466/CL16 one.

The Vdimm figures I see get bragged about on other OC forums with Ryzen 1 & 2 series are in vicinity of 1.41 - 1.50 to get near or on 3466 / CL14 etc...and they use single rank sticks.

What is not often seen or heard on these platforms is dual ranked sticks hitting 3200, 3466 or higher with tight timings. Not sure about rank format of your sticks though.

I would not give up but if you get chance try it on X470 mobo. Buildzoid from YT suggested not doing more than 1.15 SoC when he was reviewing MSI's X470 version of the Gaming Pro Carbon AC series. I tend to think he knows what he is talking about compared to some of them on there. Just my opinion though...

I take it you've tried Ryzen DRAM Calculator? the dude that authors that, has had several updates since launch, so it's not an abandoned project.

 

[SIGSEGV]

can i join to this club?

 

[robot zombie]

The Vdimm figures I see get bragged about on other OC forums with Ryzen 1 & 2 series are in vicinity of 1.41 - 1.50 to get near or on 3466 / CL14 etc...and they use single rank sticks.

I've noticed they are very power sensitive. Sometimes more power actually makes things worse. I was methodical with each trial, starting with the lowest recommended by the calc and working my way up. LLC settings were crucial too. Though many times after failing that I'd say fuck it and crank it and it'd finally boot. Little bit of both, there. Some work better on the low end of the recommended range. Some have to go well above or not at all.

I also tried everything else on the flow chart. No luck at all. Messing with termination at all means it wont post. No matter what. That makes me question this board a lot. I think its the BIOS. Its wayy behind on AGESA updates...

I messed with powerdown, geardown, even tried disabling both and running 2T command rate. Between all of this, I'm systematically running through standard timings adjustments. I would change one factor, run through all of the acceptable timings, change it back and change the next and repeat. Then comes combos. Even what should be trivial timings are giving me problems at any power level or combinations of settings.

When conventional wisdom failed, I basically went brute force. Continuing at this point is like falling in love with the wrong woman over and over again.

What is not often seen or heard on these platforms is dual ranked sticks hitting 3200, 3466 or higher with tight timings. Not sure about rank format of your sticks though.

They're regular ole single-rank b-die. But the latency is designated higher. They're specced as 3600/CL17. Its a thing. From what ive read they should do all but the most extreme. 3400/14 should be doable. Its not :/ honestly I'd even take CL15. It's just a little lower binned than the 3200/CL14 and 3600/CL16 sticks you see the most of. So it'll never do the super high 3800+ speeds, but it should still at least overclock to the level of the modules binned just above it... you'd think. I dunno maybe I lost the silicon lottery. Maybe that's my tradeoff for getting a 2600 that'll do 4.2ghz at 1.3v

That's something I need to consider, too... ...perhaps trying to take the RAM up is leaving the CPU a little starved. Actually with my old ram it was happy with 1.29v. I upped it for stability. I might try going higher. Temperatures start to go up though, just from the RAM itself being under higher voltages, so I don't want to add too much to it. Maybe if I clock down to 4.15 or 4.1. But it's like, at that point is it even worth it? Hmmm...

I would not give up but if you get chance try it on X470 mobo. Buildzoid from YT suggested not doing more than 1.15 SoC when he was reviewing MSI's X470 version of the Gaming Pro Carbon AC series. I tend to think he knows what he is talking about compared to some of them on there. Just my opinion though...

I really like his content. Ive seen all of his videos on memory and stuff. Taught me alot. Guy knows his stuff.

But yeah... there is nothing left to do but give up otherwise. I have tried everything 3 times over now. Trust me I am very thorough. So many long hours of being almost there...

This ram on this board at the most will do CL14 at 3200. Past that its a minimum of 16. No in between. Short of a new board its not happening. For now im holding out for a BIOS update. I see myself going to 7nm when they come out so I don't want to buy another new board just to be a board behind again later. Im planning a build for a friend. Depending on what he wants to pay me I may toss in this X370 and put the money in a good X470. Well see.

I take it you've tried Ryzen DRAM Calculator? the dude that authors that, has had several updates since launch, so it's not an abandoned project.

Ive been the one pushing it these past couple of pages Unfortunately for me very few of them work, or even have the potential to work for me. Even many of the more conservative options never stood a chance. Ive got stacks of booklets made from printouts of every possible config to shoot for. All marked to hell. 30+ pages of configs from it, all tested and tweaked as much as possible.

Its just... what works is so far off from what it prescribes that if it isnt the calc its hardware or firmware. Probably both of the latter in my case.

 

[lexluthermiester]

can i join to this club?

Welcome!

 

[YautjaLord]

Video coming soon. Will test with Prime95 Small FFTs torture test (2 hours) & CPU-Z bench. Games & synthetic benchmarks (3DMark Time Spy, Unigine Superposition, etc...) will come in next video.

 

[GoldenX]

can i join to this club?

Welcome, try not to bump with the FX pile at the door.

 

[Liquid Cool]

I'd like to join as well....

I put together a Ryzen 3 2200G mini itx setup for a family member about 2 months back and used a Gigabyte board with DDR4 3000 memory. I played with that system for quite a few days...had fun overclocking the cpu and gpu but there didn't seem to be as many FPS gain in games as there was when I ramped up the memory from 2133 to 3000 using the XMP profile. Those gains...we're rather substantial.

I also noticed...the majority of the gains were had from 2133 to 2666, then dropped off in percentages the higher I went from there. So, when I purchased the setup in my "system specs" I didn't hesitate to grab a DDR4 2666 kit when I saw it on a good sale.

Although...here is where the trouble starts...I can't get these sticks to run at the advertised speed. I can manually set them at 2400(stock 1.2v) and the system boots up and runs fine(actually great), but when I use the XMP profile and it sets them to 2666...the system won't even boot. I'm not too familiar with newer systems and haven't owned an AMD computer since 2005, not too mention...I've never owned DDR4 before. So, I didn't want to start pumping voltage past their rated 1.2v until I know what I'm doing.

Frankly, I can't tell if the motherboard is at fault, or if the memory just won't run at the rated speed. I'll return/exchange something if I need to....but I like my set up the way it is....

I'd be appreciative if someone could enlighten me as to what to do here...

Best Regards,

Liquid Cool

 

[robot zombie]

I'd like to join as well....

I put together a Ryzen 3 2200G mini itx setup for a family member about 2 months back and used a Gigabyte board with DDR4 3000 memory. I played with that system for quite a few days...had fun overclocking the cpu and gpu but there didn't seem to be as many FPS gain in games as there was when I ramped up the memory from 2133 to 3000 using the XMP profile. Those gains...we're rather substantial.

I also noticed...the majority of the gains were had from 2133 to 2666, then dropped off in percentages the higher I went from there. So, when I purchased the setup in my "system specs" I didn't hesitate to grab a DDR4 2666 kit when I saw it on a good sale.

Ahh, with the APU's RAM speed matters a lot more, not only because the CPU's infinity fabric's latency/bandwith is tied to it, but because the integrated GPU is allocating ram for VRAM. In order for it to function effectively as VRAM it needs to be pretty fast. Additionally, having a decent chunk of RAM helps too. Say you allocate 4GB to VRAM in the BIOS, you will want to have yet another 4GB free for the allocated space to work with. Same as how with a dedicated GPU, performance can suffer a lot from not having at least as much RAM free as the GPU has VRAM of its own.

One thing worth mentioning when shopping for RAM, in general, but especially with Ryzen. Rated speeds aren't as important as the speed/CAS latency ratio. Different vendors indicate different bin levels differently. Take GSKILL for example. They have multiple designations for the exact same DIMMs. 3200/CL14 and 3600/CL16 will both do either of those settings and perform more or less the same. Make sense? For these RAM vendors it seems to come down to however they decided to test it that day

Basically what I'm saying is tightest possible timings for whatever speed are what you want. If your ram is 2666 with decent timings it can easily do more, or less with even tighter timings. It's not a rigid designation.

Although...here is where the trouble starts...I can't get these sticks to run at the advertised speed. I can manually set them at 2400(stock 1.2v) and the system boots up and runs fine(actually great), but when I use the XMP profile and it sets them to 2666...the system won't even boot. I'm not too familiar with newer systems and haven't owned an AMD computer since 2005, not too mention...I've never owned DDR4 before. So, I didn't want to start pumping voltage past their rated 1.2v until I know what I'm doing.

Frankly, I can't tell if the motherboard is at fault, or if the memory just won't run at the rated speed. I'll return/exchange something if I need to....but I like my set up the way it is....

XMP/DOCP profiles with Ryzen can be a bit iffy. As far as voltage goes, don't worry too much. SoC is good to 1.2v and actually that's meant to be kind of a safe limit, factoring in the fact that some boards overvolt SoC significantly. The latest DDR4 can safely be run as high as 1.5v with adequate cooling, though realistically 1.35 is more common for higher speeds. 1.5 is really getting into crazy overclocking/showing-off/benching territory.

But really, you will need more voltage. I'd probably start with 1v SoC (the max most need is 1.1-1.125) and 1.35v for your RAM. Even 1.4v isn't gonna be an issue. I think the easiest way to get it going is to use the Ryzen DRAM Calculator. You'll have to figure out your DRAM manufacturer/bin - follow the instructions given on the page. You probably won't have to get fancy and manually enter the ns latencies. If you wind up with something common, all you have to do is select the manufacturer/bin and hit the purple "XMP" button and it will plug-in the standard latencies for you.

Once you have the info plugged into the calc, chose profile "v1" and hit "fast" or "safe." Start with fast. it will spit out timings and power ranges that should work for the speed you set. It gives you everything. Adhere to what it provides to a T and you should have no troubles, assuming your RAM/mobo is up for it. The calc also has a section for suggested LLC settings. Highly recommend trying them, too.

Also included with the download is a handy flowchart for troubleshooting issues with bad configs. Will help you through every step from posting to running stable. I also posted it in this thread a page back. Can't miss it.

Worth mentioning... ...messing with RAM on Ryzen can get complicated, but it doesn't have to be. Barring bad luck like I've been having, the calculator should get you there. And if it's not quite there, you can always loosen the timings a hair and call it a day The "safe" timings are generally guaranteed. Most will do "fast" just fine and it is still safe, but stability isn't a given on all hardware.

Almost forgot, do make sure your BIOS is up to date. I can tell you first hand that earlier BIOSes on most boards had trouble with RAM compatibility, especially with regards to running rated speeds.

Have fun! Welcome to The Circle.

 

[biffzinker]

One thing worth mentioning when shopping for RAM, in general, but especially with Ryzen. Rated speeds aren't as important as the speed/CAS latency ratio.

I've found out that it's worth running at a slower MHz for DDR4 with tighten timings vers loosen timings higher MHz for my 2600X. I'm getting better benchmark scores at 3200 MHz with 15-17-17-17-35 and secondary timings for refresh of 63/416. I tried running at 3400 MHz with timings at default but I lose bandwidth I get back at 3200 MHz with tightened timings.

 

[robot zombie]

I've found out that it's worth running at a slower MHz for DDR4 with tighten timings vers loosen timings higher MHz for my 2600X. I'm getting better benchmark scores at 3200 MHz with 15-17-17-17-35 and secondary timings for refresh of 63/416. I tried running at 3400 MHz with timings at default but I lose bandwidth I get back at 3200 MHz with tightened timings.

It's a tricky balance. For me, I get worse performance at 3200mhz 14-14-14-14-28 than I do at 3466mhz 16-16-16-16-32. With the latter, bandwidth goes up significantly and latency is the same as the former. Secondary timings are a big factor. I manage all-around better relative secondary timings for the latter, so the higher-speed config is nothing but a benefit to both latency and bandwidth. No matter what I do, I cannot get a tighter balance at CL14. Go figure. I've actually got the AIDA benchmarks posted on this page to show it.

So I guess it's best to say it's a little bit of everything. Just because somebody else has a certain sweet spot doesn't mean that's your ideal. Depends on where your hardware limitations are.

I do agree, best rule of thumb is to shoot for the lowest speed that Infinity Fabric actually needs and tighten timings as much as possible from there. But I still think it's worth mentioning that primary timings are only part of the picture. Once you hit 2800-3200 it becomes a balancing game between speed, primaries, secondaries, and power delivery. Of course that all is a given but one can't overstate the equal significance of each of those factors.

I wonder how it shakes down BELOW what infinity fabric likes. Do you take the hit in IF latency/throughput to get tighter RAM timings? Or do the looser RAM timing affect things less than worse IF performance?

 

[AlwaysHope]

I've noticed they are very power sensitive. Sometimes more power actually makes things worse. I was methodical with each trial, starting with the lowest recommended by the calc and working my way up. LLC settings were crucial too. Though many times after failing that I'd say fuck it and crank it and it'd finally boot. Little bit of both, there. Some work better on the low end of the recommended range. Some have to go well above or not at all.

I also tried everything else on the flow chart. No luck at all. Messing with termination at all means it wont post. No matter what. That makes me think at this point & in light of that glad I went with MSI on my current gaming rig.

Yeah, sometimes too much voltage can have counterproductive benefits with OC on any component. Have heard b4 about Asus being slack with AGESA updates that makes me think this time around instead of going with them, I'll give MSI a go with my current gaming rig. Another thing to consider is empty DIMM slots, apparently it's been found they can cause an antenna effect & exaggerate EM interference. Kinda makes sense when I think about it.

I messed with powerdown, geardown, even tried disabling both and running 2T command rate. Between all of this, I'm systematically running through standard timings adjustments. I would change one factor, run through all of the acceptable timings, change it back and change the next and repeat. Then comes combos. Even what should be trivial timings are giving me problems at any power level or combinations of settings.

When conventional wisdom failed, I basically went brute force. Continuing at this point is like falling in love with the wrong woman over and over again.

Yeah, been there, done that but not so much on Ryzen systems yet, & I ain't going down that road either. Better things to do with my time on the rig like actually gaming on it....lol... even if cutting back the OC a bit. Stability is heaven on earth in computer world imo.

They're regular ole single-rank b-die. But the latency is designated higher. They're specced as 3600/CL17. Its a thing. From what ive read they should do all but the most extreme. 3400/14 should be doable. Its not :/ honestly I'd even take CL15. It's just a little lower binned than the 3200/CL14 and 3600/CL16 sticks you see the most of. So it'll never do the super high 3800+ speeds, but it should still at least overclock to the level of the modules binned just above it... you'd think. I dunno maybe I lost the silicon lottery. Maybe that's my tradeoff for getting a 2600 that'll do 4.2ghz at 1.3v

Haven't researched those sticks but is that specced for Intel platforms? if it is, herein lies part of the problem imo. Unless you spend countless hours testing every little minor timing on them, one can't be 100% certain. Apart from that, perhaps it is bad luck in the silicon lottery after all?

That's something I need to consider, too... ...perhaps trying to take the RAM up is leaving the CPU a little starved. Actually with my old ram it was happy with 1.29v. I upped it for stability. I might try going higher. Temperatures start to go up though, just from the RAM itself being under higher voltages, so I don't want to add too much to it. Maybe if I clock down to 4.15 or 4.1. But it's like, at that point is it even worth it? Hmmm...

That's a consideration only you can answer at this point. If 100 or 50MHz of CPU speed is warranted or not.

I really like his content. Ive seen all of his videos on memory and stuff. Taught me alot. Guy knows his stuff.

But yeah... there is nothing left to do but give up otherwise. I have tried everything 3 times over now. Trust me I am very thorough. So many long hours of being almost there...

This ram on this board at the most will do CL14 at 3200. Past that its a minimum of 16. No in between. Short of a new board its not happening. For now im holding out for a BIOS update. I see myself going to 7nm when they come out so I don't want to buy another new board just to be a board behind again later. Im planning a build for a friend. Depending on what he wants to pay me I may toss in this X370 and put the money in a good X470. Well see.

B450 & X470 designed for 2933MHz out of the box. Not that the IMC is on the chipset of course but the overall design of the boards accomodates that official AMD ram speed. I still think if you tried it on decent X470 board, you could be surprised. B450 is still very young & bioses haven't matured as much.

I
Ive been the one pushing it these past couple of pages Unfortunately for me very few of them work, or even have the potential to work for me. Even many of the more conservative options never stood a chance. Ive got stacks of booklets made from printouts of every possible config to shoot for. All marked to hell. 30+ pages of configs from it, all tested and tweaked as much as possible.

Its just... what works is so far off from what it prescribes that if it isnt the calc its hardware or firmware. Probably both of the latter in my case.

 

[robot zombie]

Yeah, sometimes too much voltage can have counterproductive benefits with OC on any component. Have heard b4 about Asus being slack with AGESA updates that makes me think this time around instead of going with them, I'll give MSI a go with my current gaming rig. Another thing to consider is empty DIMM slots, apparently it's been found they can cause an antenna effect & exaggerate EM interference. Kinda makes sense when I think about it.

I don't think all Asus boards have that problem. Seems to be more with the older and/or lower-end ones specifically. I doubt it's an issue with the Crosshair or anything. Asus... ...man back in the day I loved Asus boards. These days they seem to be slacking. I actually really like the X370-F. Very high-quality VRM, nice feature set, great audio, and the actual BIOS layout/features are pretty much anything I could want, but the fact that it's so behind would've been a dealbreaker had I known when I bought it. That's the thing about motherboards. Doesn't matter if the VRM is amazing and it has all of the features you want if the BIOS has problems. And often you just can't know beforehand. :/ It's a shame. It really is a perfectly fine board otherwise. It would be one of the better X370's out there if not for crap like this. I'm stuck on AGESA 1.0.0.2!

Yeah, been there, done that but not so much on Ryzen systems yet, & I ain't going down that road either. Better things to do with my time on the rig like actually gaming on it....lol... even if cutting back the OC a bit. Stability is heaven on earth in computer world imo.

Agreed. Instability ruins my zen (lol) completely. It just feels wrong. If something messes up, I have to fix it. I'd rather just know it's always gonna work than have it work a little, unnoticeable bit better *most* of the time. With RAM in particular it can just be so sketchy, just corrupting data without you realizing until that day when something bad and irreversible happens to your system. When I was testing different configurations, I actually held a backup that I made made right before I started and that's the one I restored when I stopped testing unstable RAM configs. That kind of instability sketches me out bad. I'm always afraid something got corrupted and carried over into my backups. They only go back so far, you know?

Haven't researched those sticks but is that specced for Intel platforms? if it is, herein lies part of the problem imo. Unless you spend countless hours testing every little minor timing on them, one can't be 100% certain. Apart from that, perhaps it is bad luck in the silicon lottery after all?

Yeap, they are. Most of the gskill ones are. But looking into it, these are supposed to be nothing more than middle-bin b-die, so it shouldn't matter all that much. I know that Intel and AMD carry different profiles/ratings/so-on but I'm purely comparing apples to apples. And doing that, these are in the middle, ratings-wise.

I nabbed them because the price was right at the time and I knew for a fact they were b-die. Honestly, I feel like they're not performing badly at all, like compared to any other RAM available. I was just hoping for a little more than I managed. I don't know what it is but something's telling me they have more to them than what I'm getting. Maybe silicon lottery, but I really suspect motherboard, given how people always say that they get better RAM overclocks with the newer BIOS updates.

If not, it's like I said, this is the price I pay for lucking out on a vanilla 2600 that hits 4.2ghz completely stable at an almost unheard of, low n' cool 1.29v. Can't have ideal everything for every part in a build.

That's a consideration only you can answer at this point. If 100 or 50MHz of CPU speed is warranted or not.

Yeah, I'll mess with it when I get the itch. I'm betting it won't make much of a difference, but I had some nice scoring 3533 and 3600 configs that just weren't quite stable. If a few little things here and there can get me there it might be worth it. Voltage climbs a lot between 4.1 and 4.2... I think I was at 1.25ish to run 4.1 stable. Dunno, man. Eventually I'll get around to messing with them more.

B450 & X470 designed for 2933MHz out of the box. Not that the IMC is on the chipset of course but the overall design of the boards accomodates that official AMD ram speed. I still think if you tried it on decent X470 board, you could be surprised. B450 is still very young & bioses haven't matured as much.

B450 isn't even a consideration for me. Not into their power sections or overall feature sets. But yes, I assume the X470's do much better. I just can't justify it for one bad thing. The board is great save for this minor issue. If my buddy wants to put up a good chunk of what a top-end X470 costs for me to go all-in on this build for him, which with the time I'm going to put in he seems to be content to, then maybe I'll nab one. Less guilt when somebody else is funding your upgrade

 

[YautjaLord]

Done. How do i upload a video here beside typing in URL? It's offline file off of my Galaxy A6, haven't uploaded it yet to YT.

P.S. It stopped been bubble-y, all installed, will run it for few more hours tomorrow to get rid off of all the air bubbles left & then Prime95 & rest of the tests.

 

[GoldenX]

Mmm, tomato sauce. Joking aside, that looks cool AF.
Noob question, you can get any water colour, right?

 

[YautjaLord]

@GoldenX:
More like blood red, but the cam (A6, 16MP, 4:3, 4608x3456) made it look like tomato sauce yeah, also there was extra light from bathroom bulb.
Distilled water, coolant is CryoFuel from EKWB, concentrate, comes in all color variants, i.e. - red, blue, green, etc...., as long as you have money cash for either of this, yeah, you can get it in any color you desire.

P.S. Tomato sauce cause it was all bubble-y & stuff, it was still in process of getting rid off of air bubbles when i took that shot.

 

[SIGSEGV]

using XMP setting, is it normal 2700X has vcore around 1.4 - 1.48?

 

[robot zombie]

using XMP setting, is it normal 2700X has vcore around 1.4 - 1.48?

Yeah, it'll do that at the top end of its boost. Usually only on a core or two at a time, and briefly. Put it under load and watch individual core voltages and speeds.

 

[SIGSEGV]

Yeah, it'll do that at the top end of its boost. Usually only on a core or two at a time, and briefly. Put it under load and watch individual core voltages and speeds.

thanks,
I think it'd be better to use offset -0.15v of vcore to compensate the XFR boost

 

[robot zombie]

thanks,
I think it'd be better to use offset -0.15v of vcore to compensate the XFR boost

Absolutely. A lot of people do just that and it seems to work well for them.

 

[AlwaysHope]

Absolutely. A lot of people do just that and it seems to work well for them.

Except those on MSI platforms. But even without offset, no problems here on my system.

 

[Mussels]

so annoyed by the lack of offset on my MSI board -.- everything else about it works great, but its static voltages and no clock/volt down at idle

 

[lexluthermiester]

Ahh, with the APU's RAM speed matters a lot more, not only because the CPU's infinity fabric's latency/bandwidth is tied to it, but because the integrated GPU is allocating ram for VRAM. In order for it to function effectively as VRAM it needs to be pretty fast.

You are very correct on this point. The faster the RAM the better!

Say you allocate 4GB to VRAM in the BIOS

With an APU, 4GB is overkill. 2GB is all you need because an APU GPU isn't going to push anything that will need 4GB or more.