Its always worth the investment in time if you can do it.I manually set my memory speed and timings. It can take you down a rabbit hole and be a big time sink. It paid off for me since now I can tweak my own memory (Hynix A-die) to much better performance compared to out of the box EXPO memory.
well, my $400 (at the time) 6600 cl34 after setting docp I to 6000 and it already did the voltages, and without tightening the timings would not even post, even at 5600. But works fine at 4800 default. docp I did set the voltage to 1.4 on 3 places. I also tried docp II and it did change some timings, but again, no post.At worst you can use Buildzoid's timings for Hynix M and A die memory.
It's a solid upgrade over XMP and EXPO.
Even my worst-binned Hynix will happily do 6000 MT/s without issue.
Or go Gigabyte premium mobo model in future. Their DDR5 support is supposed to be better than the rest, from what I've read.If I ever go Intel on something I will have 32 gig for it. A waste of $400 (the memory started this thread)
gigabyte screwed me on an RMA back in 2008. I will never do business with them again. They lave lost thousands due to screwing me for $200 years ago. I have a memory like an elephant.Or go Gigabyte premium mobo model in future. Their DDR5 support is supposed to be better than the rest, from what I've read.
That WAS the year of the financial crisis and maybe whoever was in charge at the time, was doing crisis management? That person probably screwed over a lot of other customers and not just you, just to show better short term financials. Hopefully, that person is no longer there so why not give them another chance?gigabyte screwed me on an RMA back in 2008.
If you're using ASUS boardswell, my $400 (at the time) 6600 cl34 after setting docp I to 6000 and it already did the voltages, and without tightening the timings would not even post, even at 5600. But works fine at 4800 default. docp I did set the voltage to 1.4 on 3 places. I also tried docp II and it did change some timings, but again, no post.
Its only $138 for 6000 CL30 EXPO, 32 gig, so thats what I ordered. If I ever go Intel on something I will have 32 gig for it. A waste of $400 (the memory started this thread)
The ram SAYS 1.4 right on it, and thats what DOCP used. Before I found DOCP I manually set the one voltage to 1.4, same result.If you're using ASUS boards
vSOC 1.25V
voltage for mem controller / ram 1.35V-1.36V
Overly high vSOC and vMC will actually be less stable
Rest of timings per this post from Buildzoid:
Low effort Single rank Hynix 16Gb A/M-die timings for Ryzen 7000 at 6000Mbps | Actually Hardcore Overclocking (Buildzoid)
Get more from Actually Hardcore Overclocking (Buildzoid) on Patreonwww.patreon.com
HUB review of the ASRock mATX B650 board is up:
Short summary: "Actually really good [for the price]"
-Stock settings enforce a 85C limit for CPU which makes a slight difference in MT performance and VRM temps. But VRMs are still ok with this limit removed and performance within 1% of top boards for the most part. VRM adequate for 7950X with stock limits removed.
-Power on to Windows boot time of 16 seconds
-Echoing my experience with ASRock boards... default XMP/EXPO timings suck. So you will give up gaming performance (few %) here unless you manually tune to at least Buildzoid timings. And their BIOS is bugged for the memclk divider currently (at least on X670E board in my sig) so forget going faster than 6000 MT/s.
-Overall, a solid budget board. Once I can buy a 7800X3D my 7700X in my main rig will be going into a mATX build using this board
One of the things I found interesting on this video was the boot time graph. MSI was CLEARLY the slowest boards to boot during testing. Mine isn't as slow as he shows on the chart, but I haven't had any luck speeding it up either like others with ASUS, ASRock or Gigabyte boards have. Then again, it's not that big of a deal really. How often are we booting up anyway? A handful of times a week?HUB review of the ASRock mATX B650 board is up:
Short summary: "Actually really good [for the price]"
-Stock settings enforce a 85C limit for CPU which makes a slight difference in MT performance and VRM temps. But VRMs are still ok with this limit removed and performance within 1% of top boards for the most part. VRM adequate for 7950X with stock limits removed.
-Power on to Windows boot time of 16 seconds
-Echoing my experience with ASRock boards... default XMP/EXPO timings suck. So you will give up gaming performance (few %) here unless you manually tune to at least Buildzoid timings. And their BIOS is bugged for the memclk divider currently (at least on X670E board in my sig) so forget going faster than 6000 MT/s.
-Overall, a solid budget board. Once I can buy a 7800X3D my 7700X in my main rig will be going into a mATX build using this board
That's for the rated 6600 CL34. But that's overkill for 6000 MT/s and particularly if tuning for timings more voltage can actually negatively impact stability, particularly if airflow over your DIMMs is low/non-existent.The ram SAYS 1.4 right on it, and thats what DOCP used. Before I found DOCP I manually set the one voltage to 1.4, same result.
I feel obligated to say something about the Buildzoid timings that are so frequently referenced. I can't speak with personal experience with his Zen 4 timings but I do have relevant experience on the Intel side.If you're using ASUS boards
vSOC 1.25V
voltage for mem controller / ram 1.35V-1.36V
Overly high vSOC and vMC will actually be less stable
Rest of timings per this post from Buildzoid:
Low effort Single rank Hynix 16Gb A/M-die timings for Ryzen 7000 at 6000Mbps | Actually Hardcore Overclocking (Buildzoid)
Get more from Actually Hardcore Overclocking (Buildzoid) on Patreonwww.patreon.com
The best DDR5 kit you can buy for Ryzen (by far) is this kit, it's basically overclocked with the EXPO settings out of the box.
I feel obligated to say something about the Buildzoid timings that are so frequently referenced. I can't speak with personal experience with his Zen 4 timings but I do have relevant experience on the Intel side.
Don't use those timings, seriously. They're overly aggressive, I would say it's overly aggressive even for a 'power' user that wants every possible frame. Certain timings he has listed such as tRefi at 50000 are just asking for data corruption. His listed timing for tRas at 30 is insanely tight for literally zero benefit. You will never notice a difference between a default tRas of 96 and his setting of tRas with a value of 30.
The best DDR5 kit you can buy for Ryzen (by far) is this kit, it's basically overclocked with the EXPO settings out of the box. If you want a 64Gb kit, this is the best you can get, as a bonus it will also perform slightly better than the aforementioned 32Gb kit.
The referenced Zen4 timings are just as bad, many values are set lower without benefit and likely got autocorrected due to many of us having GDM=enabled. This was said by someone at oc forum that show values derived from formula that has some basis to it than the lower is better screenshot crowd when actual benches don't show any improvement. My result with tRAS=30 is not as consistent as tRAS=tRP+tRTP=49 and I tried many times on various tests. I kept my tRDRDSCL=tWRWRSCL=8, rather than the 4 that was referenced for example, some prefer 6, but I see no difference on this setting. Tightening tRTP=12 vs tRTP>=18 benefit some, hurt others, Intel MLC latency test being one I can consistently reproduce this, so less is not more.I feel obligated to say something about the Buildzoid timings that are so frequently referenced. I can't speak with personal experience with his Zen 4 timings but I do have relevant experience on the Intel side.
Don't use those timings, seriously. They're overly aggressive, I would say it's overly aggressive even for a 'power' user that wants every possible frame. Certain timings he has listed such as tRefi at 50000 are just asking for data corruption. His listed timing for tRas at 30 is insanely tight for literally zero benefit. You will never notice a difference between a default tRas of 96 and his setting of tRas with a value of 30.
Zen 4 IMC is not Alder/Raptor Lake IMC.
Depending on your motherboard vendor, you may have worse-than-expected performance using EXPO out of the box due to unoptimized subtimings.
I have fully tested my memory settings and Buildzoid's timings for Zen 4 have worked even on my worst-binned Hynix IC kit (5600 CL36). Even non-EXPO kits. YMMV with different motherboards and bins of RAM.
We have forum members who have used more aggressive timings than Buildzoid and passed multiple different kinds of rigorous memory testing. In my opinion and experience, his Zen 4 Hynix memory timings are predictably reliable on latest AGESA at 6000 MT/s, and provided your memory controller permits it, you can actually push more aggressive timings to 6400+ MT/s. Obviously each user should do stress testing for themselves, but at least with Zen 4 IMC you will quickly know whether it is going to pass or not. Failures are typically quite quick.
Strange Asrock, if you look a little closer, the motherboard does not have any classic connectors for Led lighting.HUB review of the ASRock mATX B650 board is up:
Short summary: "Actually really good [for the price]"
-Stock settings enforce a 85C limit for CPU which makes a slight difference in MT performance and VRM temps. But VRMs are still ok with this limit removed and performance within 1% of top boards for the most part. VRM adequate for 7950X with stock limits removed.
-Power on to Windows boot time of 16 seconds
-Echoing my experience with ASRock boards... default XMP/EXPO timings suck. So you will give up gaming performance (few %) here unless you manually tune to at least Buildzoid timings. And their BIOS is bugged for the memclk divider currently (at least on X670E board in my sig) so forget going faster than 6000 MT/s.
-Overall, a solid budget board. Once I can buy a 7800X3D my 7700X in my main rig will be going into a mATX build using this board
I feel obligated to say something about the Buildzoid timings that are so frequently referenced. I can't speak with personal experience with his Zen 4 timings but I do have relevant experience on the Intel side.
Don't use those timings, seriously. They're overly aggressive, I would say it's overly aggressive even for a 'power' user that wants every possible frame. Certain timings he has listed such as tRefi at 50000 are just asking for data corruption. His listed timing for tRas at 30 is insanely tight for literally zero benefit. You will never notice a difference between a default tRas of 96 and his setting of tRas with a value of 30.
The best DDR5 kit you can buy for Ryzen (by far) is this kit, it's basically overclocked with the EXPO settings out of the box. If you want a 64Gb kit, this is the best you can get, as a bonus it will also perform slightly better than the aforementioned 32Gb kit.
Yours is likely a dual rank kit whereas his timings are specifically for single rank Hynix. All of my (3) kits I've tested are single rank M-die FWIW.I'll pipe up that on my 6000C30 2x32gb hynix a die kit buildzoid timings caused me some weird system crashes and reboots.
I went back to my own personal timing blend at 6200C28. I never saw any real gain from using his timings in benchmarks or games anyway.
Yeah, that might play a role in it. I left the specific dual rank subtimings unchanged but the memory controller might have been unhappy about some of the rest.Yours is likely a dual rank kit whereas his timings are specifically for single rank Hynix. All of my (3) kits I've tested are single rank M-die FWIW.
I'm anticipating prices for memory and flash will keep dropping through at least the rest of Q2 if not into Q3 so waiting to buy a 64GB (or 96GB) kit in a few months. DDR5-6000 CL30 is already substantially more reasonably priced vs even 1 year ago so DDR5 early adopter tax is fairly minimal at this point.
There are also single rank 64gb kits. I have one, but I haven’t tried his timings yet. I plan to do so when i have time to spare.Yeah, that might play a role in it. I left the specific dual rank subtimings unchanged but the memory controller might have been unhappy about some of the rest.
I never bothered on my single rank 32gb kit because I planned all along to get a larger kit when it was available. Now it looks like there are 96gb kits of hynix coming out from Gskill, but I don't see any EXPO kits on the list yet.
Is there a reliable way to tell what my kit is? I kind of just assumed it was dual rank to make 32GB dimms.There are also single rank 64gb kits. I have one, but I haven’t tried his timings yet. I plan to do so when i have time to spare.