Ryzen Owners Zen Garden

[Johan45]

I would guess flare X would clock better, the Rgb on these does not play nice with my asus motherboard either and is stuck in rainbow cascade or something, I'm not bothered too much but worth mentioning.

Does it not get along with the Aura SW from ASUS?

 

[TheoneandonlyMrK]

Does it not get along with the Aura SW from ASUS?

No I've had no luck with gigabytes own software either.

 

[lexluthermiester]

and is stuck in rainbow cascade or something

I believe you can turn it off or unplug the RGB.

 

[TheoneandonlyMrK]

I believe you can turn it off or unplug the RGB.

No switch to adjust or wire to remove, its ok i quite like the omnicron crystal look.

 

[Vya Domus]

Oh yes LPX, not Pro. These modules are notoriously bad with Ryzens

Wish you good luck, I managed absolutely nothing above 2400 with my kit.

Quick question. The idle clocks seldom go below 3.00 GHz, and never below 2.20 GHz.

Don't know what's up with the clocks either. My 1700X clearly seems to idle just fine as the temperature goes down to ~25C and so does the power consumption and voltages, however no program seems to properly read the clocks as they all report that I am at 4000mhz constantly.

 

[Johan45]

Wish you good luck, I managed absolutely nothing above 2400 with my kit.



Don't know what's up with the clocks either. My 1700X clearly seems to idle just fine as the temperature goes down to ~25C and so does the power consumption and voltages, however no program seems to properly read the clocks as they all report that I am at 4000mhz constantly.

Like I posted earlier a lot of that is dependant on BIOS and which AGESA is used.
I assume you checked the Windows settings

 

[Vya Domus]

I assume you checked the Windows settings

Which settings ? You mean the power plan ? That has always been left on balanced, anyway like I said the idle states do work properly it's just that the clocks aren't displayed properly.

 

[Johan45]

Which settings ? You mean the power plan ? That has always been left on balanced, anyway like I said the idle states do work properly it's just that the clocks aren't displayed properly.

Or there's just no load which will bring the temps/power usage down

 

[Vya Domus]

Or there's just no load which will bring the temps/power usage down

Unlikely since the voltage does drop.

 

[YautjaLord]

I would guess flare X would clock better, the Rgb on these does not play nice with my asus motherboard either and is stuck in rainbow cascade or something, I'm not bothered too much but worth mentioning.

It would only work with GB mobos cause of all that GB's RGB Fusion stuff.

Flare X was stable @ 3333MHz, froze @ 3400MHz on mine, but i also had 2700X @ 4.2Ghz. I'm still working on getting 4.3GHz on CPU/3466MHz on RAM running stable when testing those with Cinebench R15 & Prime95/AIDA64 FPU tests.

 

[TheoneandonlyMrK]

It would only work with GB mobos cause of all that GB's RGB Fusion stuff.

Flare X was stable @ 3333MHz, froze @ 3400MHz on mine, but i also had 2700X @ 4.2Ghz. I'm still working on getting 4.3GHz on CPU/3466MHz on RAM running stable when testing those with Cinebench R15 & Prime95/AIDA64 FPU tests.

I found a few comments on forum's saying gigabyte had added support in a update but I think your right.

 

[robot zombie]

There is a different type of B-die out there and I'm not sure it could be confused with the "good" B-Die. https://www.overclockers.com/adata-xpg-spectrix-d40-ddr4-3000-review/

Yeah, I saw that earlier. I don't think mine is remotely the same. That's a really different creature. I think what we're seeing now is yes, the actual DRAM chips on the sticks are legit b-die, but the controllers vary quite a lot in quality.

The way I see it, there seems to be a small range of controllers from tighter to more relaxed, but within this set, all will do 3200 CL14. I see it best represented at 3600 in the CL15, CL16, and CL17 sets. All b-die with similar controllers that will actually run a lot of the same profiles, though subtimings get looser as you go. They all give the performance associated with b-die. Mine will still do the fabled 3200 CL14 easily. Works great at ~1.35v and under 1v SoC. But it likes slightly more relaxed timings to do higher speeds, even though it will still do them. Even among classic b-die there is a fair bit of variance. Not all b-die is necessarily good for overclocking, but these are the ones people are always talking about.

And then there are a few out there that use completely different controllers. It almost doesn't seem right to call them b-die. But maybe that's where we all have steered each other wrong. We missed an important distinction. It's like... ...you can buy a car with a nice engine but if the transmission sucks then... ...just because car A has the same engine as car B doesn't make it able to put the same power down. And similarly, just because you buy a pair of sticks known to be b-die doesn't magically make them all-around HQ RAM. It just means they have high quality DRAM chips tacked on the PCB. Maybe there was more of it floating around than we've been lead to believe. Or maybe manufacturers, realizing that people were seeking out b-die and actually checking, started tacking on inferior controllers and charging the same price to cash in on the obsession, just based on availability.

And then there are the crazy high speed ones that actually have pretty crummy IC's tacked on... dunno if I'd put those in the same class, either.


I'm pretty sure with the calc, the V1 and V2 profiles are for the 2 originally most common controllers. Could be wrong. Just makes sense to me. Too bad it's not that simple anymore. Maybe it never was.


Sometimes I wonder how much of a role my mobo plays in my case. I don't have some important settings. And something just messing around tells me I'm missing something with this board. It's just not a great memory OCer. Probably a little bit of both for me. Oh well. At least I got 3200 CL14 going.

Shame, really. I still feel like it shouldn't take all of this just to find RAM that works how you want it to. It's not silicon lottery. It's more like a silicon easter egg hunt. Obnoxious to know that what you want is out there, but you just can't find it. Nerve wracking to pull the trigger on expensive RAM only to find it's not what you hoped it was.

 

[Johan45]

Call me a fan boy but I have always had good luck with G.Skill. The pics I posted earlier are a 3600 CL17-18-18 kit @ 3333, I also have some 4266 which is a bit better binned, flare X, and etc.. they all seem to behave similarly on Ryzen. I also have a quad kit of 3600 corsair that runs surprisingly well on TR. I haven't noticed this "different" type yet I guess

 

[TheoneandonlyMrK]

Call me a fan boy but I have always had good luck with G.Skill. The pics I posted earlier are a 3600 CL17-18-18 kit @ 3333, I also have some 4266 which is a bit better binned, flare X, and etc.. they all seem to behave similarly on Ryzen. I also have a quad kit of 3600 corsair that runs surprisingly well on TR. I haven't noticed this "different" type yet I guess

I think you pay for what you get largely ,Gskill do have good binning and memory tuning skills, a fact i will remember ,next time.

 

[YautjaLord]

Didn't change my R15 score though & actually (true story) was a tad (10 or 20cb) lower for some f**ked up reason than with stock 3200Mhz frequency*. Still scores above i7 4770K in single-threaded R15 score. Does it's job fine, no quirks & shit, rock solid, plus aesthetically pleasing, has a Predator 2 left arm gauntlet of sorts appearance to it. (Yeah i know, i am a massive Predator & Predator 2 geek & i think i proved it. )

*Frequency was 3333Mhz, voltage - 1.35v. Lower but still stable R15 score.

 

[robot zombie]

Call me a fan boy but I have always had good luck with G.Skill. The pics I posted earlier are a 3600 CL17-18-18 kit @ 3333,

Hah! Probably the same kit I got, or very similar. TridentZ RGB 3600 17-18-18. I've been playing around with 3333 on it myself, with some promising results. It's a great kit. Just really doesn't do higher speeds well. Does a pretty tight 3200 CL14, though. I'm with you on G.Skill. They're the go-to right now and even the b-die with lesser timings specced OC great.

But I think those belong to the "true" b-die set. High quality chips on high quality PCB's with good designs and good components. I don't think the same can be said for all b-die out there. It's just like any other part you buy. Core components are important but the whole design matters too. Think of the variance in SSD's with the same memory... ...or especially the difference there can be between two different video cards with the same GPU under the cooler. The difference with RAM is that it's often not as easy to know unless you really do your homework.

But then, mobo compatibility is all over the place, too. That too can be hard to shop for. There's not much that says "this mobo is a good RAM OC'er." Makes things even more complicated.

Big ole pandoras box of "not all things are created equal," imo. Though I think generally if it's confirmed b-die, the timings are in line, and the manufacturer has it together, you're still safe.

I think the truest thing anyone can say about it is that it's a complicated situation with a lot of things being unseen on the consumer end and like mrk said, you get what you pay for. My one hope going forward is better compatibility/performance and a little more "Ryzen-specific" stuff. AMD removing some limiting factors and RAM manufacturers hashing things out more on their end. It's one of those things where we probably don't even need better hardware to get better performance. It just needs to be set up to work together better. Lotta ground to cover there.

I also have some 4266 which is a bit better binned, flare X, and etc.. they all seem to behave similarly on Ryzen. I also have a quad kit of 3600 corsair that runs surprisingly well on TR. I haven't noticed this "different" type yet I guess

I sometimes wonder if the real limitations are elsewhere than the RAM's spec. Like up to a certain point it doesn't matter what it's rated for because you're gonna hit a wall with the CPU/mobo before the RAM even gets close to the limits of what it can do. Maybe that's why it doesn't always matter what you get and why "worse" RAM often performs close to or just as well as "better" RAM. Hence, "as long as it's b-die."

 

[lexluthermiester]

Call me a fan boy but I have always had good luck with G.Skill. The pics I posted earlier are a 3600 CL17-18-18 kit @ 3333, I also have some 4266 which is a bit better binned, flare X, and etc.. they all seem to behave similarly on Ryzen. I also have a quad kit of 3600 corsair that runs surprisingly well on TR. I haven't noticed this "different" type yet I guess

I wouldn't call that being a fanboy. You like a company because you have had consistently good experiences with their products.

 

[Johan45]

Hah! Probably the same kit I got, or very similar. TridentZ RGB 3600 17-18-18. I've been playing around with 3333 on it myself, with some promising results. It's a great kit. Just really doesn't do higher speeds well. Does a pretty tight 3200 CL14, though. I'm with you on G.Skill. They're the go-to right now and even the b-die with lesser timings specced OC great.

But I think those belong to the "true" b-die set. High quality chips on high quality PCB's with good designs and good components. I don't think the same can be said for all b-die out there. It's just like any other part you buy. Core components are important but the whole design matters too. Think of the variance in SSD's with the same memory... ...or especially the difference there can be between two different video cards with the same GPU under the cooler. The difference with RAM is that it's often not as easy to know unless you really do your homework.

But then, mobo compatibility is all over the place, too. That too can be hard to shop for. There's not much that says "this mobo is a good RAM OC'er." Makes things even more complicated.

Big ole pandoras box of "not all things are created equal," imo. Though I think generally if it's confirmed b-die, the timings are in line, and the manufacturer has it together, you're still safe.

I think the truest thing anyone can say about it is that it's a complicated situation with a lot of things being unseen on the consumer end and like mrk said, you get what you pay for. My one hope going forward is better compatibility/performance and a little more "Ryzen-specific" stuff. AMD removing some limiting factors and RAM manufacturers hashing things out more on their end. It's one of those things where we probably don't even need better hardware to get better performance. It just needs to be set up to work together better. Lotta ground to cover there.


I sometimes wonder if the real limitations are elsewhere than the RAM's spec. Like up to a certain point it doesn't matter what it's rated for because you're gonna hit a wall with the CPU/mobo before the RAM even gets close to the limits of what it can do. Maybe that's why it doesn't always matter what you get and why "worse" RAM often performs close to or just as well as "better" RAM. Hence, "as long as it's b-die."

Here's my timings if you're interested and shows the memory model number as I do think it's the same. 1.37V in BIOS

 

[lexluthermiester]

1.37V in BIOS

Is it just me or does that seem a bit high?

 

[Johan45]

That's DRAM voltage, stock is 1.35V

 

[robot zombie]

Here's my timings if you're interested and shows the memory model number as I do think it's the same. 1.37V in BIOS

Yep, model numbers match Thanks for the timings man, I'll play with them next time I sit down with it.

Is it just me or does that seem a bit high?

Yes and no. JEDEC says 1.2. But XMP specs 1.35 for most "factory" overclocks. A lot of DDR4 will run that way out of the box. So not too far off. 1.35 is what most significant, but reachable overclocks ask for. I'd say 1.4v is still reasonable, if not starting to get up there. Many people go much higher. And it seems like DDR4 can take it. The general consensus is that DDR4 can actually take some pretty insane-looking overvolts if you really wanna push the speed. I believe the max for XMP 2.0 cert is 1.5v. Gotta be something to that.

 

[TheoneandonlyMrK]

Well im presently back at 3200,16,17,17,17,17(,38,56) with the best bandwidth I've yet got ,i use maxmemm2 though so I dunno it's reliably shown.

But i put in brackets my main two unlowerable settings, any lower and i get boot loops or corrupt os issues put it back it boots.

Scratch this whole post started crashing 3dnark

 

[btarunr]

Zen 2 IPC is 29% higher than Zen1. Whatcha gonna do Intel?

https://www.techpowerup.com/249450/amd-zen-2-ipc-29-percent-higher-than-zen

 

[ne6togadno]

Zen 2 IPC is 29% higher than Zen1. Whatcha gonna do Intel?

https://www.techpowerup.com/249450/amd-zen-2-ipc-29-percent-higher-than-zen

Raja will demo their dGPU with zen2 based system

 

[robot zombie]

Zen 2 IPC is 29% higher than Zen1. Whatcha gonna do Intel?

https://www.techpowerup.com/249450/amd-zen-2-ipc-29-percent-higher-than-zen

Im cautiously optimistic. Waiting to see actual performance numbers.

I got a 2600 w/ second hand x370 because I knew if Zen 2 flopped I'd still be happy with what I already had. OC it and live my life, enjoying what the savings afforded me elsewhere.

On the flipside I held back on the 2700x because if Zen 2 was super amazeballz I would feel silly upgrading but I'd still want to