DDR5-6000 Memory is the Sweet Spot For AMD Ryzen 7000 Zen 4 CPUs (update)
AMD's Ryzen 7000 Zen 4 CPUs appear to have DDR5-6000 as their sweet spot, allowing for a 1:1 IFC ratio. A 1:1 ratio indicates that the memory operates at the same frequency as the memory controller on the CPU, which should provide the best-case situation.
Intel has divided the memory ranks with Alder into two categories: a 2:1 mode known as Gear 2 that is the default for DDR5, and a 4:1 variant known as Gear 4. A 1:1 ratio has the advantage of allowing for reduced latencies and a balanced speed. In contrast, a greater ratio allows for better overclocking and quicker data transfer rates but also results in worse latencies.
- AMD Ryzen 3000 "Zen 2" Sweet Spot - DDR4-3800
- AMD Ryzen 5000 "Zen 3" Sweet Spot - DDR4-4000
- AMD Ryzen 7000 "Zen 4" Sweet Spot - DDR5-6000
DDR5-6000 for AMD Ryzen 7000 "Zen 4" CPUs already sounds amazing for AM5, with a default to DDR5-5600. Higher frequency DIMMs are supported, however, if you exceed the DDR5-6000 limit, you will be forced to use a 1:2 IFC. DDR5-6400 operating at 1:2 is said to offer poor results and is not advised if you want better gaming performance.
Update:
AMD announced DDR5-6000 as the "sweetspot" memory overclock for their next Ryzen 7000 "Zen 4" processors in its Discord AMA. In AMD language, a sweet spot frequency is an inflection of performance, stability, cost, and ease. AMD's Robert Hallock, who led the Discord AMA, suggested that FClk be left undisturbed at "Auto" for the best results. Memory overclocking yields lesser results with dual-rank DIMMs, or two DIMMs per memory channel, as with both AMD and Intel currently. Some of the AMD EXPO-certified DIMMs launched in recent days use memory clocks higher than DDR5-6000.
Team Group Adds T-Force Delta RGB in DDR5-6600 and DDR5-6000 CL30 to lineup - 05/25/2022 07:55 AM
TEAMGROUP is proud to announce the release of two new specification upgrades of T-FORCE DELTA RGB DDR5 Gaming Memory. The overclocking RAM comes in a high-speed variant clocking in at 6,600 MHz and a ...
DDR5-6000 memory with EPIC-X RGB heatsink from PNY - 05/09/2022 06:30 PM
PNY Technologies introduced the DDR5-6000 memory XLR8 DDR5 6000MHz MAKO RGB with an aluminum heat sink....
G.Skill Launches 2x32GB DDR5-6000 CL30 64GB Memory Kit - 03/30/2022 09:10 AM
SKILL is delighted to announce the launch of a high-capacity, ultra-low-latency DDR5-6000 CL30 64 GB (32 GB x2) overclocked performance memory kit under the Trident Z5 family, designed for the latest ...
Senior Member
Posts: 1261
Joined: 2013-02-22
Gskill used the old B-die chips. Those older Samsung B-die chips could run super low latencies, you can really lower the tRFC a ton. I have these kits= and you can check the B-die revision is some of the older stuff. The new revisions of b-die could hit much higher frequencies.
My daily driver with that kit took the tRFC down to 268 and its rock stable. I can get it tighter, but this is a nice stable compromise without much voltage.
12699
Same kit hits 3866mhz C16 on my 3700x so I dont think the kit is the issue I think my CPU's might have been duds and it's quite well known that zen 1/+ typically maxed out at 3466 on a good sample but I couldn't even hit that
Senior Member
Posts: 8164
Joined: 2020-08-03
It matters sometimes(see from your link). My point was more about latency. Just because memory shows say DDR4-3600 14-15-15 doesn't mean the tRFC isnt super high and latency sucks. Its all about latency for gaming. The right way to do this is to latency test each setup and post the ns latency to rank each ram. Notice how the 4600 is lower FPS than the 3600, sure it could be just the main settings like CAS but without showing latency all these types of tests are for show.
12698
4600 has to be using 1:2 controller mode that's why its lower on Intel 4600 c18 will crush 3600 c14
Senior Member
Posts: 2649
Joined: 2015-06-11
It matters sometimes(see from your link). My point was more about latency. Just because memory shows say DDR4-3600 14-15-15 doesn't mean the tRFC isnt super high and latency sucks. Its all about latency for gaming. The right way to do this is to latency test each setup and post the ns latency to rank each ram. Notice how the 4600 is lower FPS than the 3600, sure it could be just the main settings like CAS but without showing latency all these types of tests are for show.
12698
wasn't this caused mostly by infinity fabric sync rates?
Senior Member
Posts: 769
Joined: 2016-10-04
It matters sometimes(see from your link). My point was more about latency. Just because memory shows say DDR4-3600 14-15-15 doesn't mean the tRFC isnt super high and latency sucks. Its all about latency for gaming. The right way to do this is to latency test each setup and post the ns latency to rank each ram. Notice how the 4600 is lower FPS than the 3600, sure it could be just the main settings like CAS but without showing latency all these types of tests are for show.
12698
We do not know the IF settings especially the 4000Mhz one.
Senior Member
Posts: 1677
Joined: 2017-02-14
Gskill used the old B-die chips. Those older Samsung B-die chips could run super low latencies, you can really lower the tRFC a ton. I have these kits= and you can check the B-die revision is some of the older stuff. The new revisions of b-die could hit much higher frequencies.
My daily driver with that kit took the tRFC down to 268 and its rock stable. I can get it tighter, but this is a nice stable compromise without much voltage.
12699