[SOLVED] Timings for Patriot Viper Steel 16GB 4400Mhz with Ryzen 5600x

kingbowcat

Reputable
Oct 9, 2019
237
14
4,615
Hi,

I am upgrading my processor and Ram being the Viper Steel 4400Mhz rated duel channel memory. I was wondering with the new 5600x what timings I should shoot for as lots of differing opinions online and a lot of talk of infinity fabric 1:1 allowing for tighter timings?

Heard some people have got 3800Mhz on CL 14 with this kit? Or you think this is BS? or Silicone lottery?

Does anyone have this memory kit and tell me what timings they have achieved or should I just experiment with it when it comes?

Thanks alot.
 
Solution
3800Mhz CL14? Possible, but unlikely. It also depends on what size kit it is, what size EACH stick is, and how many DIMMs make up the kit. G.Skill is the only company I know of that even HAS any 3800mhz CL14 kits, and those are limited to only three kits they make.

Unless you are a pure hobbyist when it comes to tuning systems, I'd recommend that you either simply run it at the A-XMP profile configuration, IF it is a kit that is validated for the motherboard model you have, or if it won't run at that configuration then I'd simply set the speed and DRAM voltage and then either leave the timings on auto or use the Ryzen calculator.

I personally don't see there being all that much benefit with Ryzen once you go past 3733mhz, but, I have...
3800Mhz CL14? Possible, but unlikely. It also depends on what size kit it is, what size EACH stick is, and how many DIMMs make up the kit. G.Skill is the only company I know of that even HAS any 3800mhz CL14 kits, and those are limited to only three kits they make.

Unless you are a pure hobbyist when it comes to tuning systems, I'd recommend that you either simply run it at the A-XMP profile configuration, IF it is a kit that is validated for the motherboard model you have, or if it won't run at that configuration then I'd simply set the speed and DRAM voltage and then either leave the timings on auto or use the Ryzen calculator.

I personally don't see there being all that much benefit with Ryzen once you go past 3733mhz, but, I have not yet fiddled with a Ryzen 5000 series memory configuration either so it might be a bit more tolerant but it's very doubtful you'll get 1:1 ratio at anything higher than 3600-3733. We've "heard" of higher speed kits while retaining 1:1, but I don't think it's widespread or common above that speed even with Zen 3 parts.
 
Solution

Karadjgne

Titan
Ambassador
Fclock is reported stable at 1900 (3800MHz ram), for a 1:1 on Zen3. AMD claims fclock should be stable at 2000, but I've yet to see that happen, even Steve at GamersNexus could only get 1900, same with Linus.

So I'd do as Darkbreeze suggested, run the timings at XMP/DOCP, but I'd manually set fclock to 1900 for 1:1 as 2200 and a 2:1 ratio might be a hit to performance. Not sure if there's enough speed to compensate there at 4400MHz.

You could always try Dram Calculator from guru3d.com and see what (if any) gains are to be had from tighter secondary and tertiary sub-timings.

My RipJaws are at stock xmp primary timings, they wouldn't take the shift from cl16 to cl14, but there was a marked difference in synthetic performance (Cinebench R.20) with tightened subs.
 
Yes, tightening down timings AFTER you get a stable configuration, is always a good option for those WILLING to put in the required work to validate the stability of the adjustments. For others, the reality is that the differences in performance for ANY kit at it's profile speed and whatever overclock beyond XMP or whatever changes to tighten down timings might be possible isn't really going to offer any earth shattering performance gains, so unless you are the "squeeze out every ounce of free performance possible" types that is willing to do the somewhat extensive stability testing necessary, it's usually best to simply try to stick to something close to the profile, if possible, or a manual setting with auto configured timings, if not.

But, like we've both said, the Ryzen DRAM calculator is another alternative that can SOMETIMES be helpful too.
 
  • Like
Reactions: Karadjgne

Karadjgne

Titan
Ambassador
I went from just over 3700 points in stock configuration running CB, to just over 4800 points simply by using dram calculator and hard setting the fclock manually. Pretty sure that'll have some impact on higher core count gaming, how much I couldn't say for sure since I hadn't bothered to run the tests beforehand. But thats a considerable bump in a blender workload, especially considering that running ClockTuner2 and lowering temps and voltages while raising clock speeds over 400MHz per core only raised that score from 4800 to 5100.

I swear there are days when I can believe that fine tuning a Ryzen and ram is somewhat more trixy than fine tuning the Quadrajet carb on my Caballero.
 
1 out of 300 visitors would be able to tell the difference without a synthetic benchmark to show them there WAS one. Some timing adjustments I can "feel" it a little "snappier" but overall regardless of bench results the differences for most people are not noticeable.

If you are trying to set a pole position on some benchmark site, then sure. The scenarios where any of this translates into "real world" performance gains are few and far between, and almost entirely dependent on the workload.

I think having at least somewhat more memory than you think you need, at whatever the fastest speed is that makes sense financially when you look at the returns on investment, plus of course running in dual channel in the correct DIMM slots and set at their advertised speed, is more than good enough for 90% of users, especially gamers, to avoid any memory related issues in performance.
 
There are no listing for Patriot Viper beyond 2666Mhz on the MB QVL that are tested and known to work so choose a kit that are tested at 3200MHz preferably CL14 (2x8) with Samsung B-Die chips.

Here is the QVL for Pinnacle ridge: https://www.msi.com/Motherboard/support/B450-TOMAHAWK-MAX#support-mem-14

G.Skill F4-3200C14D-16GVR are ideally suited for your CPU.:sneaky:

With all the hype from above I'm surprised you're not completely confused.
Also no tested kits beyond 3400Mhz so 3800MHz is BS.
 
Last edited:

Karadjgne

Titan
Ambassador
I've not in 40 odd years of building and fixing and tinkering around in pc's ever used ram listed on the QVL of any motherboard. I've come close, but never the exact model. Mostly because the QVL is a Vendor list, not a ram list. It's incomplete, a sampling and seriously misleading.

There's only half a dozen actual ram OEMs give or take a few, and they make all the ram for every vendor. If you've tested corsair lpx, you've tested Patriot, Kingston, Adata, team and many more besides since the SkHynix chips used are all the same. But even then that's no guarantee per model since most vendors can and will change chips according to contract. A Patriot model might be tested with SkHynix chips, but 6 months later be using Micron or Nanya chips.

On top of that, the models themselves are all similar, the model numbers reflecting physical changes. There's over 3000 individual model numbers just for gskill Trident-Z, but if you tested the gray/white, you've tested the black/white, rgb, red, blue, camo etc the only difference being the color of the heatsink or the size of the kit.

On top of that, the vendor tested that particular ram stick on that particular motherboard. There's absolutely no guarantee that the same model ram will work on the same model motherboard bought by someone else.

It's a mess.
 
QVL list isn't quite a joke, in my opinion, but it IS (Again, in MY opinion) more of a fallback if a kit cannot be validated as compatible by means of the memory manufacturer's list, for those that have them. For those that do, they are MUCH more likely to be not only more accurate but far more expansive as well, however, Patriot is not one of those manufacturers that provides full spectrum board compatibility listings like G.Skill, Corsair and Crucial do. So, you are pretty much stuck either referencing the QVL (And there is no such thing as "similar" when it comes to two memory kits with different models. Even DIMMs with identical model numbers can have anything from minor to fundamentally and distinctly different configurations) or trying to find validation that a given memory kit works on a given board on various forums by people who have proven compatibility.

Even then, it's often a roll of the dice. But if a kit is not listed, and there are no other qualified models with similar speed/frequency by that manufacturer for that board on it's QVL, then I'd agree with MeanMachine on that one in that it's a lot less likely that faster or different kits of the same speed are going to be compatible.

What is the ACTUAL model of the memory kit, exactly?
 
Then it is VERY likely to be compatible with any Ryzen platform, but might NOT be able to run at that speed, or at any given "configuration" settings, past 3800mhz, without a great deal of fiddling with speed, timings and voltage. Hard to say really tbh. I haven't seen a lot of people have much success with very fast kits like that on current Gen Ryzen platforms except the folks doing reviews or those who have been tuning and building systems for a long time. I guess it could go either way but if it was my kit I'd likely run it at 3733-3800mhz with whatever timings a similar speed Viper steel kit has advertised or whatever the lowest timings you can get it to run stable at, and by stable I mean running the extensive testing that includes Memtest86 and Prime.
 

kingbowcat

Reputable
Oct 9, 2019
237
14
4,615
Your right seen a lot of memory benchmarks and its minimal. Went for cheaper 3200Mhz 16 cl memory as deal for Patriot memory couldn't be purchased due to shenanigans of online store I tried to purchase it from. (overclockers UK) needless to say I won't be visiting them ever again.
 

Karadjgne

Titan
Ambassador
Don't think any were designed specifically for any of the Ryzens, afaik the fabs were in production long before Ryzen showed up.

But what ended up happening is that Samsung B-die ended up being as good a fit as possible, able to get its XMP values where others failed. Then came Agesa and a bunch of compatibility fixes, so right now things are pretty much ironed out, especially for the major OEMs and consequent Vendors.

That leaves things like OC and timing changes right back where they started, with the silicon lottery. Some just works better than others, and with the 5000 series, 2x 16Gb is the best fit and performance, followed closely by 4x 8Gb, then 2x 8Gb (usually with 3200MHz +) . And that has to do with the 5000 architecture and how it deals with Dual Rank as well as Dual Channel.

That generally means 3800MHz, with a 1900 fclock, 2x 16Gb, Dual channel, and Cas14 or Cas16 Samsung B-die is about as good as it'll probably get for a 5600x.
 
The ONLY memory I know of that was created, and manufactured, specifically for Ryzen platforms, is the G.Skill Flare X and the G.Skill Neos. Those kits ARE specifically made for Ryzen platforms, and are compatible with a majority of motherboards, but you STILL need to validate compatibility because there is NO memory that is compatible with ALL motherboards at advertised speeds, or in some cases, at all.
 

kingbowcat

Reputable
Oct 9, 2019
237
14
4,615
The ONLY memory I know of that was created, and manufactured, specifically for Ryzen platforms, is the G.Skill Flare X and the G.Skill Neos. Those kits ARE specifically made for Ryzen platforms, and are compatible with a majority of motherboards, but you STILL need to validate compatibility because there is NO memory that is compatible with ALL motherboards at advertised speeds, or in some cases, at all.

Is the patriot ram I got compatible with B450 tomahawk max? How do I find out?
 
You can't, not really, because Patriot doesn't offer a compatibility list showing which memory kits they make are compatible with exactly what motherboards, like G.Skill, Corsair and Crucial do. Most of these smaller companies who make memory, don't, which is why I largely avoid them except for occasionally on older systems or some Intel platforms, since most Intel systems are far more accepting of a wider range of memory types, speeds and configurations. Ryzen is particular and is still to this day far more finicky than Intel platforms.

Also, if it's on the motherboard QVL list, then usually that is going to be pretty solid as far as compatibility as well unless the memory manufacturer has CHANGED the makeup of the module since it was tested for QVL, which actually happens quite a bit. So a memory kit that shows as compatible on QVL might not even BE that same memory in reality anymore. There are also some sites out there that offer user validated model listings for Ryzen motherboards.
 

FoxVoxDK

Distinguished
14-15-15-15-30(3800) is doable, but depends on the die's you get and the voltage you're willing to run.
14-14-14-14-28(3800) is also possible, but "rare" and the extreme low latency requires a board that's daisy chained and have the dimm slots very close to the socket.
Sneezing in the vicinity of the RAM with these timings might make them upset with you.
 
14-15-15-15-30 is doable, but depends on the die's you get and the voltage you're willing to run.
14-14-14-14-28 is also possible, but "rare" and the extreme low latency requires a board that's daisy chained and have the dimm slots very close to the socket.
Are you offering suggestions for Ryzen in general, or that specific memory kit, or that specific motherboard, or some combination of all three, or what? You can't just spit out timings or configuration settings and not offer not only some kind of verification to evidence that what you are saying is accurate, but also what hardware it's actually accurate for. Not that you aren't right, I'm not saying that, just that I think in a case like this more information is probably necessary.

Anybody can just say "try this" and occasionally it will work, but more often than not what works under one set of conditions is not going to work under another set or in some cases even under the same conditions based on additional hardware considerations.
 

FoxVoxDK

Distinguished
@Darkbreeze OP asked:

Heard some people have got 3800Mhz on CL 14 with this kit? Or you think this is BS? or Silicone lottery?

I answered it's not BS. OP is clearly looking for input regarding experience and to me it seems like he's going to have a whack at it.

Regardless, the PVS4400 set is going to give him tons of joy trying to squeeze better and better timings from, and hardware wise, apart from what I already said, and the timings he can achieve is unpredictable. I just answered his query.

@kingbowcat If you're interested, Actually Hardcore Overclocking on youtube is person who first made me aware of these sticks, he does some pretty decent(very decent :p) OC's on them, start your journey there.
 
Last edited:
Right. I agree with most of that. Just keep in mind, of course, that as you say it's unpredictable and every set of sticks, on every board, with every CPU, are all likely going to be capable of somewhat different end results because all these components have their own characteristics and even two identical motherboards, CPUs or memory kits might not individually be capable of being fine tuned to the same result as the other one. It will be trial and error, which is why we say you have to put in the work. There are no numbers that just "work" for every configuration regardless of the component in question when it comes to tuning, unless you're just willing to settle for something very loose, which sort of defeats the purpose.

If you decide to custom tune that configuration, do yourself a favor and at least read the section on stability testing, here:

 

kingbowcat

Reputable
Oct 9, 2019
237
14
4,615
Right. I agree with most of that. Just keep in mind, of course, that as you say it's unpredictable and every set of sticks, on every board, with every CPU, are all likely going to be capable of somewhat different end results because all these components have their own characteristics and even two identical motherboards, CPUs or memory kits might not individually be capable of being fine tuned to the same result as the other one. It will be trial and error, which is why we say you have to put in the work. There are no numbers that just "work" for every configuration regardless of the component in question when it comes to tuning, unless you're just willing to settle for something very loose, which sort of defeats the purpose.

If you decide to custom tune that configuration, do yourself a favor and at least read the section on stability testing, here:


Thanks for all the help but I just realized I messed up.. My MOBO isn't even compatible with the memory (not on the list) though the lower MHz memory of the same series is. But I can run everything fine? Or is it not on the list as the clock is so high (higher than MOBO can handle) not accounting for slowing the speed down for lower CL?