Overclock.net banner
1 - 15 of 15 Posts

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #1 ·
Done on the MSI Edge. I thought it was impossible for Z690, but it apparently wasn't. Just wanted to share my configuration in case anybody else wants to attempt the same.

The timings are the tightest they can be, although I haven't tried tightening tRP, tRAS, or tCWL further. RTLs are mismatched due to mixed kits (two 2x8 GB Single Rank kits).

Trained with ~1.40V VCCSA and ~1.49V VDIMM (BIOS), on the Memory Try It! "4,400 MHz Gear 2" profile, on BIOS version 1.22 (7D31v12).

This was done with a binned DDR4 IMC on a 12900KF that can boot up to 4,300 MHz Gear 1, which is quite rare for 12900s. So expect difficulty with any IMC that's even marginally weaker. 3,800 MHz max might be the realistic max for most chips. (Which would be even lower on a 12700 or worse.) And the board could be T-Topology? Dunno.

The same 4x8 GB DIMMs could boot up to 4,400 MHz on Gear 2. 4,533 MHz only POSTs. With 2x8 GB, I could boot 5,000+ MHz.

I'll test out my 4x16 GB SR Micron B-die kit at some point in the future, when I'm not lazy and busy.
 

Attachments

·
Registered
Joined
·
757 Posts
Done on the MSI Edge. I thought it was impossible for Z690, but it apparently wasn't. Just wanted to share my configuration in case anybody else wants to attempt the same.

The timings are the tightest they can be, although I haven't tried tightening tRP, tRAS, or tCWL further. RTLs are mismatched due to mixed kits (two 2x8 GB Single Rank kits).

Trained with ~1.40V VCCSA and ~1.49V VDIMM (BIOS), on the Memory Try It! "4,400 MHz Gear 2" profile, on BIOS version 1.22 (7D31v12).

This was done with a binned DDR4 IMC on a 12900KF that can boot up to 4,300 MHz Gear 1, which is quite rare for 12900s. So expect difficulty with any IMC that's even marginally weaker. 3,800 MHz max might be the realistic max for most chips. (Which would be even lower on a 12700 or worse.) And the board could be T-Topology? Dunno.

The same 4x8 GB DIMMs could boot up to 4,400 MHz on Gear 2. 4,533 MHz only POSTs. With 2x8 GB, I could boot 5,000+ MHz.

I'll test out my 4x16 GB SR Micron B-die kit at some point in the future, when I'm not lazy and busy.
Impressive :) Do the IMC prevent you from tightening timings further or is binning beliw avg? Since RP defines RC (important for performance) I would try lowering it, 16 should work fine, maybe lower. Surprised if RFC can't run at 312, 304 or 296.
 

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #3 · (Edited)
Impressive :) Do the IMC prevent you from tightening timings further or is binning beliw avg? Since RP defines RC (important for performance) I would try lowering it, 16 should work fine, maybe lower. Surprised if RFC can't run at 312, 304 or 296.
Aside from the three timings I mentioned above, the rest are as tight as they can be. However, I would say that it is in part due to the IMC and in part due to the thermals.
The heatspreaders on these two kits basically don't allow any air to pass through, so that's extremely limiting on cooling potential. I have two dedicated fans blowing on it already.

When both kits can handle up to 1.68V with super tight timings just fine as 2x8 GB but can't do more than 1.49V together with looser timings, that's a clear indication of a thermal issue. The RAM sticks are just too deeply squashed together.

I don't experience this issue with Micron B-die 4x16 GB; they can handle 1.63V just fine, super tight.
Sometime later, I will try and see whether or not I can go beyond 4,000 MHz. I didn't actually try to do so yet, since I'm already pushing on 1.40V VCCSA.
 

·
Registered
Joined
·
757 Posts
Aside from the three timings I mentioned above, the rest are as tight as they can be. However, I would say that it is in part due to the IMC and in part due to the thermals.
The heatspreaders on these two kits basically don't allow any air to pass through, so that's extremely limiting on cooling potential. I have two dedicated fans blowing on it already.

When both kits can handle up to 1.68V with super tight timings just fine as 2x8 GB but can't do more than 1.49V together with looser timings, that's a clear indication of a thermal issue. The RAM sticks are just too deeply squashed together.

I don't experience this issue with Micron B-die 4x16 GB; they can handle 1.63V just fine, super tight.
Sometime later, I will try and see whether or not I can go beyond 4,000 MHz. I didn't actually try to do so yet, since I'm already pushing on 1.40V VCCSA.
You may end up getting rev B that outperforms B-die due to heat if they are very good bin :p Best rev Bs can run like 15-17-17, RAS 36, FAW 16 at 4000, but RFC below 560 is unlikely so B-die may win after all.
 

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #5 · (Edited)
You may end up getting rev B that outperforms B-die due to heat if they are very good bin :p Best rev Bs can run like 15-17-17, RAS 36, FAW 16 at 4000, but RFC below 560 is unlikely so B-die may win after all.
One thing I am 100% certain of is that tCL is linked to IMC quality. The rest, not so much. But many other major timings instantly throw errors if tightened further and tested.

That is also one thing I am curious about. I want to see whether or not 4x16 GB Micron B-die will run at 4,000 MHz to begin with. That'll depend on how strong this IMC is.
Since the dies are different, there is a chance that Micron might be able to do better than Samsung, if it can achieve tight enough timings with the increased thermal headroom, as you say.
16-18-XX-XX-2T weighs on Samsung B-die heavily, as it makes it perform more like an above-average Micron E-die kit. Looser secondaries don't help, either.

I don't think that tRCD of 17 is possible at 4,000 MHz. 18, perhaps. But I haven't tried that yet, so I can't say that for sure. At 4,174 MHz, tRCD 18 will train but not POST on Z390.
Since I needed the bandwidth, I didn't really test thoroughly as to what I could achieve on 4,000 MHz. I just quickly raised it to 4,174 MHz (the max my CPU IMC would allow).

I will test this 4x8 GB Samsung B-die setup at 4,100+ MHz, and if that fails, I'll move onto Micron B-die.
I didn't bother tightening tRP, tRAS, and tCWL because I only did this to prove that it is possible to run 4x8 GB at 4,000 MHz, with a golden IMC.
I know exactly what my Micron B-die kit is capable of (see below), so whether or not I can push the same timings would depend entirely on the IMC. (If it boots, of course.)
 

Attachments

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #6 · (Edited)
@Taraquin @Raphie @bscool
4,266 MHz at 4x8 GB booted (1.40V VCCSA). Holy cow.
I'm honestly shocked. The MSI Edge might actually be T-Topology!

Since it's just a boot test, I haven't tried to tighten yet. Timings are loose on purpose. Will report back.
One thing I did notice was that upon restarting on 4,266, Explorer crashed. So there might be corruption or IMC instability at 1.40V VCCSA. Just something to note.

4,300 POSTs, but fails to load Windows. And it seems I've corrupted the Windows installation in the meanwhile, lol.
 

Attachments

·
Registered
Joined
·
309 Posts
You’re on to something. Curious how this will develop.
 
  • Rep+
Reactions: Ichirou

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #8 · (Edited)
You’re on to something. Curious how this will develop.
Windows is freshly reinstalled.

4,266 MHz is borderline bootable (desktop only). IMC limit.
It seems even 4,100 MHz is impossible to stabilize as well, although it can survive on the desktop just fine up to 4,200 MHz.
Nothing will let me run TM5 for more than a few seconds without an instant BSOD. I've tried loosening everything. No dice.

(Also, it seems that the CPU really hates being at 1.50V VCCSA. Gets even more unstable than 1.45V or less.)

I think 4,000 MHz might really be it. I managed to get y-cruncher stable with 16-18-16-30-2T and tCWL at 14. @Taraquin
Didn't bother to TM5 test it, since it's not of huge importance to me.

Oh well. Hopefully Micron B-die can do 4,000 MHz as well. A bit skeptic now, but the dies are different, so one cannot be sure.
At least I've found out that the motherboard might be T-Topology. Or it's just strong enough that the RAM is no longer an issue but the CPU is.
It might be worth upgrading to Raptor Lake once that is released, since I would imagine that the IMC is stronger.
 

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #9 ·
@Raphie @Taraquin @bscool
Minor update: It seems my CPU's IMC might've already degraded at 1.40V. I have to raise it from 1.40V to ~1.42V to pass TM5 again, with the same config in the first post.
Maybe 1.40V VCCSA isn't safe for Z690?
 

·
Premium Member
Joined
·
2,306 Posts
@Raphie @Taraquin @bscool
Minor update: It seems my CPU's IMC might've already degraded at 1.40V. I have to raise it from 1.40V to ~1.42V to pass TM5 again, with the same config in the first post.
Maybe 1.40V VCCSA isn't safe for Z690?
I ran 1.45 for quite a while on sa. Never notice any issue.

Edit it was the first few month Strix needed sa in the 1.4 to 1.45v range to run 4133 to 4266 with DR. With newer bios I could run 1.35v to 1.4v.
 

·
Registered
Joined
·
309 Posts
I got 3900 10.000% coverage Karhu stable, could boot into windows with 4K, but always BSOD under load.
There is the law of diminishing returns. Where the amount of effort seriously outweighs the reward. I think for z960 4 sticks, 3900 is that point. It's the start of the hockeystick angle where demand increases exponentially.
 

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #12 ·
I ran 1.45 for quite a while on sa. Never notice any issue.

Edit it was the first few month Strix needed sa in the 1.4 to 1.45v range to run 4133 to 4266 with DR. With newer bios I could run 1.35v to 1.4v.
Yeah, I think I might need to test their new BIOS that came out recently. Just been a bit skeptical because of the possible microcode change.
I'll give it a try and see whether it allows me to go further.
 

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #13 ·
Leaving this here. Let's see how far I can push these bad boys.
 

Attachments

·
Registered
Joined
·
757 Posts
Leaving this here. Let's see how far I can push these bad boys.
Impressive. My bet is RFC can go way lower. 608 maybe (16 ticks increments). Best rev E/B can do around 560 at 4000. RP may go lower aswell, but you may have to increase RAS then since rev E/B seldom likes 50 RC or lower.
 

·
Registered
Joined
·
2,003 Posts
Discussion Starter · #15 ·
Impressive. My bet is RFC can go way lower. 608 maybe (16 ticks increments). Best rev E/B can do around 560 at 4000. RP may go lower aswell, but you may have to increase RAS then since rev E/B seldom likes 50 RC or lower.
The kit isn't fully tightened yet; I'm still playing around. Probably going to loosen and raise frequency first before retightening.
It was just an initial test, of which I am absolutely shocked how easy it was to stable with minimal effort required :)
I might even be able to beat the settings I got on my Z390 with less VDIMM, lol.
 
1 - 15 of 15 Posts
Top