Wow! Is ddr5 very overclockable or did I just get lucky?

[iameatingjam]

My old ddr4 3200 kit became very unstable if I were to overclock it by even 50mhz. But I bought some 5600 CL36 kingstom fury beast ddr5 for my new motherboard and cpu (13600kf and gigabyte auros b760), even though I wanted 6000 the guy at the store convinced me 5600 would be more stable so I just went along with it. However I found out, at least on some benchmarks that it actually scored worse than my ddr4, probably because of the latency? Not happy with this I tried pushing the mts up, got all the way to 6000 without even touching timings and it performed way better than before ( on passmark, from 88th percentile to 96th, and also increased cpu performance (well obviously - but it was nice to see nontheless).

Ran a stability test for 8 hours with 0 errors. Thats crazy! It even went up to 6100 but I thought maybe I shouldn't push it and brought it back down to 6000. I tried lowering the latency from 36 to 34 without much visible effect. Possibly because I didn't touch any of the subtimings because I really don't know how they work.

Is this normal? Is it safe? Should I try pushing it even more? Or should l maybe educate myself a bit more on how timings work and take that angle instead? Or maybe I should just leave well enough alone and run at 6000 CL36. According to HWmonitor the ram isn't anywhere near hot though I understand overclocking ram can also have a small affect on cpu temps. However with my Artic Liquid Freezer II 240, I'm able to run at max boost frequency indefinitely - or so it seems. So I have a bit of headroom there. And I can't overclock the cpu frequency above what the boost would have done anyway on this motherboard (didn't really think about that before buying... hmmm).


Anyway I've just never had ram that could so easily run above it advertised clock speed. Thats awesome.

 

[AlwaysHope]

Sounds like you got a kit that is designed with Intel's IMC in mind. In other words, its highly tuned to fit in with Alder lake & Meteor lake IMC. Bet its a different story on AMD platform.
Btw, what app are you using for RAM stability testing?

 

[iameatingjam]

Sounds like you got a kit that is designed with Intel's IMC in mind. In other words, its highly tuned to fit in with Alder lake & Meteor lake IMC. Bet its a different story on AMD platform.
Btw, what app are you using for RAM stability testing?

I used memtest64. Do you think thats adequate or should I run something else?

 

[Keullo-e]

Memory, like every other component is an invidual sample. Maybe you just had bad luck with your DDR4.

My DDR4-3200 32GB kit does 3466MHz fine at 1.38V, but that's their limit. Even with 1.45V, I can't get to 3500 stable.

 

[Dr. Dro]

I used memtest64. Do you think thats adequate or should I run something else?

Not an adequate tool, TPU's memtest64 doesn't work well, or at all. It's one of w1zz's old tools that haven't been updated for modern computers, I'd like to see him work on it again someday.

But you are not pushing difficult clocks or timings for an Intel 13th gen processor or motherboard. I'm doing 6400 C30 and my motherboard (despite being a high-end model) is a complete stinker for memory OC, likely because it's an E-ATX model with a large PCB. You need to run TM5 with the anta777 extreme1 profile, that's definitely the hardest thing to conquer. My system was stable in memtest86 and occt but failed TM5 within the hour... and VST is extremely difficult to pass.

I personally use:

1. bootable Passmark memtest86 or memtest86+ open source (Passmark's is better), run for 4 full passes
2. OCCT in AVX2 mode (run the 30-minute schedule a few times)
3. TM5 with anta777 extreme1 profile for 6 cycles or so (can be replaced with Karhu memory test, 10000% coverage)
4. Y-Cruncher with FFT and VST tests to settle stability once and for all

 

[Keullo-e]

I prefer Testmem5 with extreme preset myself.

 

[iameatingjam]

I tried upping the clock to 6200 and it did appear to work but performance didn't increase at all, and infact decreased, so again I brought it back down to 6000mhz and went back to timings. I put cl as 32 and matched the secondary timings to some other kingston cl32 kit and performance increased again! Now I'm in the 98th percentile and also my cpu score increased by over 1000! (in passmark performance test) Thats crazy. 5600 CL36 ram works as 6000 CL32 ram. This is definitely not the experience I had with ddr4. Guess thats just the silicon lottery for you?

I'll do some more stability testing tonight while I sleep.

 

[avidgamer121]

Memory, like every other component is an invidual sample. Maybe you just had bad luck with your DDR4.

My DDR4-3200 32GB kit does 3466MHz fine at 1.38V, but that's their limit. Even with 1.45V, I can't get to 3500 stable.

i would never above 1.35v, atleast that's what i heard is the highest stable voltage for ddr4

 

[Marcus L]

i would never above 1.35v, atleast that's what i heard is the highest stable voltage for ddr4

Depends on the ic's I've ran mine at 1.45 and up to 1.5v for years, also 1.35 is max jedec voltage for ddr4 though if you oc or run xmp/docp etc that goes out the window

 

[ir_cow]

It's all about the ICs used on the memory. Not surprised your 5600 kit works at 6000. Probably isn't Micron IC though.

 

[AsRock]

Little old now but have a watch of this, some are just more overclocked.

Like mine will do 30CL 6000 or 32CL at 6200

 

[Keullo-e]

i would never above 1.35v, atleast that's what i heard is the highest stable voltage for ddr4

I trust my interwebz buddy who said that these chips are fine with max 1.4V

 

[P4-630]

My kit does 6000Mhz @ 1.25V

 

[Marcus L]

I trust my interwebz buddy who said that these chips are fine with max 1.4V

Samsung B die can be run north of 1.6v it depends on the IC's

 

[Wirko]

I tried upping the clock to 6200 and it did appear to work but performance didn't increase at all, and infact decreased, so again I brought it back down to 6000mhz and went back to timings. I put cl as 32 and matched the secondary timings to some other kingston cl32 kit and performance increased again! Now I'm in the 98th percentile and also my cpu score increased by over 1000! (in passmark performance test) Thats crazy. 5600 CL36 ram works as 6000 CL32 ram. This is definitely not the experience I had with ddr4. Guess thats just the silicon lottery for you?

I'll do some more stability testing tonight while I sleep.

You never mentioned voltages, so I assume you're probably letting the mobo set them. That may be just fine but do use a monitoring tool that shows you the voltages.

 

[Dr. Dro]

I tried upping the clock to 6200 and it did appear to work but performance didn't increase at all, and infact decreased, so again I brought it back down to 6000mhz and went back to timings. I put cl as 32 and matched the secondary timings to some other kingston cl32 kit and performance increased again! Now I'm in the 98th percentile and also my cpu score increased by over 1000! (in passmark performance test) Thats crazy. 5600 CL36 ram works as 6000 CL32 ram. This is definitely not the experience I had with ddr4. Guess thats just the silicon lottery for you?

I'll do some more stability testing tonight while I sleep.

Wooden DDR5 configuration, believe it or not. You're nowhere even close to silicon lottery tier there. That's assuming you're stable at all, mind you, since you're just haphazardly setting your primaries and ignoring the rest without a shred of concern for stability, that's why your performance is decreasing, it's error correction kicking in.

I trust my interwebz buddy who said that these chips are fine with max 1.4V

DDR5 works different because the power delivery is on the sticks themselves, 1.5ish is well within safe range, and temps are 95 or below, problem is high frequencies + slower refresh rates lead to corruption if the memory gets warm (60C+ range depending on your tREFI value).

 

[iameatingjam]

I prefer Testmem5 with extreme preset myself.

Thanks I did just that last night with the new timings with no errors. =D.

You never mentioned voltages, so I assume you're probably letting the mobo set them. That may be just fine but do use a monitoring tool that shows you the voltages.

Thats correct. I just let xmp/mobo set them, they are at 1.25.

Wooden DDR5 configuration, believe it or not.

Huh?

You're nowhere even close to silicon lottery tier there.

I didn't mean to brag, I know there's a lot faster configurations out there. I've just never had this experience with memory before. Every kit I've had, if I change any paramater my computer starts bugging out so I just left them alone. I mean you can lose the silicon lottery can't you? Cause I guess I lost before. Thats all I meant.

That's assuming you're stable at all, mind you, since you're just haphazardly setting your primaries and ignoring the rest without a shred of concern for stability, that's why your performance is decreasing, it's error correction kicking in.

Well I was changing it by the smallest possible degree at a a time. Changing from 36 to 34 didn't yield any results, but it was stable, so when I changed to 32 I copied secondary timings from some sticks I figured probably had similar chips. And then I ran stability testing shortly after. I did testmem5 on extreme plus the 8 hours on memtest64 ( though that was before changing the timings).

And when changing frequency too, I did it in small intervals, checked the result afterwards and moved it up if it worked and down if it didn't. I mean I'm no expert here, but I thought that was the way to do it. And it did yield positive results, infact I was blown away by how much performance ( at least in benchmarks) did change. But I mean, open to constructive criticism here. If I'm doing it haphazardly, please tell me a better way.

 

[P4-630]

Something fishy about this...

If they know at Kingston they'd run at 6000MHz @ 1.25V without errors I think they would bin/sell them as a 6000 kit, no?...

 

[iameatingjam]

Something fishy about this...

If they know at Kingston they'd run at 6000MHz @ 1.25V without errors I think they would bin/sell them as a 6000 kit, no?...

I was thinking the same thing. But who knows, maybe demand for 5600 was high at the time or maybe it was more stable on some other cpu. idk. Why, do you think I should do additional testing? I've done memtest64 and testmem5 extreme now.

Something fishy about this...

If they know at Kingston they'd run at 6000MHz @ 1.25V without errors I think they would bin/sell them as a 6000 kit, no?...

Could be possible I'm reading something wrong. This is what I'm looking at.

 

[Dr. Dro]

I didn't mean to brag, I know there's a lot faster configurations out there. I've just never had this experience with memory before. Every kit I've had, if I change any paramater my computer starts bugging out so I just left them alone. I mean you can lose the silicon lottery can't you? Cause I guess I lost before. Thats all I meant.

Well I was changing it by the smallest possible degree at a a time. Changing from 36 to 34 didn't yield any results, but it was stable, so when I changed to 32 I copied secondary timings from some sticks I figured probably had similar chips. And then I ran stability testing shortly after. I did testmem5 on extreme plus the 8 hours on memtest64 ( though that was before changing the timings).

And when changing frequency too, I did it in small intervals, checked the result afterwards and moved it up if it worked and down if it didn't. I mean I'm no expert here, but I thought that was the way to do it. And it did yield positive results, infact I was blown away by how much performance ( at least in benchmarks) did change. But I mean, open to constructive criticism here. If I'm doing it haphazardly, please tell me a better way.

You can safely disregard TPU's Memtest64 entirely, it's not effective on modern PCs - you could be giga unstable and it would still pass. Did you keep an eye on any potential ECC errors? DDR5 features on-die error correction, so it's more resilient to crashes. You need to keep in mind.

I mentioned haphazardly because in the OP you said you were only changing primaries and you did not understand secondaries and tertiaries... but I don't blame you! I'd be lying if I knew the mechanics behind most of those myself, and I'm a pretty seasoned guy... I just took what I knew from DDR4, looked up what the new timings meant and tried to keep them as low as possible on my setup. Don't take it as a dig at you, it's not

Every setup is different, so I can't just give you a working timing set. It's a lot of trial and error involved. Passing testmem5 extreme1 without it spitting any errors at all is a very good sign, but if you just changed some of your primaries, there is still much to do. You could even leave your primaries completely stock and tweak out tRFC and tREFI, those are responsible for the biggest gains as I understand.

BTW - 1.25V is quite low, you can go up to 1.5V without much problem if your RAM has decent heatsinks and you can keep it below 65C when under stress. That range is where it usually begins to show corruption if your tREFI is relaxed. Just remember that volts = heat. Good luck!

 

[Wirko]

that's why your performance is decreasing, it's error correction kicking in.

What exactly do you mean here? Even the full (server) ECC doesn't correct errors by repeating failed transfers (the way TCP network protocol does). Much less on-die ECC on consumer DDR5 memory.

Performance drop at increased speed may be due to some timings that are automatically increased. Or due to less likely and weirder things: either switching to Gear 4 (??) or IMC that can run synchronised with ring bus at 6000 but not at 6200 (????).

 

[ir_cow]

Something fishy about this...

If they know at Kingston they'd run at 6000MHz @ 1.25V without errors I think they would bin/sell them as a 6000 kit, no?...

Not always. It falls under the generic binning for the midrange. Let's say all Hynix-M can do 6400 @ 1.35V , some don't make the cut and are binned lower. They might do fine at 6000 1.25V though. I have those new 24GB DIMMs that are Micron. Generically binned to 5600 but runs 7000 just fine. (With more voltage)

It's only the top echelon that is rigorously binned, the rest are done as needed to fill the demand. It could be that they don't make the timings cut either at whatever bin it intended for. Say they want to sell you 6000 CL32. Doesn't make it, so it's sold as 5600 CL36.

Realistically most Hynix-M can do 6000 CL30-36-36-50 at 1.5V. I haven't been able to get below 28 though.

 

[Dr. Dro]

What exactly do you mean here? Even the full (server) ECC doesn't correct errors by repeating failed transfers (the way TCP network protocol does). Much less on-die ECC on consumer DDR5 memory.

Performance drop at increased speed may be due to some timings that are automatically increased. Or due to less likely and weirder things: either switching to Gear 4 (??) or IMC that can run synchronised with ring bus at 6000 but not at 6200 (????).

Did not mean the ECC algorithm itself, there are multiple ways a CPU will attempt to correct errors before it goes belly up and all of them imply a performance reduction

 

[Shrek]

Is DDR5 so overclockable because it is madly error correcting? if so I would prefer to run at a speed that has few, if any, errors.

 

[ir_cow]

The error correcting as I know it for DDR5 is only for "resting" bits. Not sure how that translates. But that's whats in the write papers.

Ram has the cells refresh periodically to keep the bits in tact since it's only hold in a state by voltage. I wonder if this just means DDR5 has means to check for these errors.

ECC Regisited server memory has a parody bit while sending data I believe. server ram is outside my knowledge base really.

But I don't think ECC is the reason for high overclocking. The errors will still show up snd I would think things would be slower due to having the data discarded and resent.