DDR5 might bring Raptor Lake CPU 20% more performance
Click here to post a comment for DDR5 might bring Raptor Lake CPU 20% more performance on our message forum
asturur
i would love to see the difference with similarly clocked memories. i would imagine single core loss is more, and multi core perf is very small
Undying
I told you buying a z690 dd4 board was a stupid idea. :P
lukas_1987_dion
Why not test with DDR4 3600MHz CL14 or 4000MHz? Changes are too small? Aw.. xD
Horus-Anhur
Geekbench loves fast memory, and scores usually increase a lot with fast ram.
But it doesn't translate to real world performance. Unless it's an application that is very memory bound.
Undying
Like games? 😉
Horus-Anhur
Games like fast memory, but not as much as Geekbench.
So don't expect 20% faster performance as seen in Geekbench scores.
schmidtbag
Most (but definitely not all) games on modern CPUs see very little difference in faster RAM. More often than not, you're going to be bottlenecked by your GPU before your RAM.
VRAM, on the other hand, makes a huge difference.
TLD LARS
It does not make sense for me that single is the same but multi is much better.
Does single run in cache or why does mem speed not matter with single?
To me this looks more like a reviewer that did not set the power limits correctly, so the DDR5 board boost higher in multi then the cheaper DDR4 board does.
nizzen
Pretty funny now that 4000c14 2x16 ddr4 is double the price of 2x16 hynix ddr5 that can oc to 7000+ mhz 😛
Imagine buying new ddr4 with AlderLake or Raptor Lake 😛
If you have 4000c14 laying around from an "old" build, use it.
nizzen
Well.... I god 35% more min fps and average fps from 4800mhz xmp green dell hynix, to 7000c30 tweaked in Battlefield 2142.
Using 12900k on Apex.
Horus-Anhur
I also got dome nice gains by tweaking memory in SOTR and a few more games.
But I also got a few that got little to no gain.
lukas_1987_dion
Well I would expect 7000 freaking MHz and 30CL to make a difference over 4000-4800Mhz lol
And buying new platform this year ofc I would choose DDR5 with crappy latency but high bandwidth, as they also are more efficient.
Will wait for ZEN 4 tho, 12900K with DDR4 CL14 is plenty for me in 1440p for now. (my CB23 single core score is 2032 and 28010 for multi core)
Kool64
Multicore isn't even really a thing to be too worried about for games but it's interesting that faster speeds on memory clocks affect the cores like that. One would think latency would be a bigger issue for multicore communications.
nizzen
Yep, that game loves memory performance. My record is ~360fps average in the benchmark
Stock memory is maybe 220-250 or so. Don't remember.
TLD LARS
There are some cheap DDR4 kits out there, If you can get DDR5 Hynix for half the price I would like to know.
G.Skill TridentZ RGB Series - 32GB: 2x16GB - DDR4 - 4400MHz CL19 (19-26-26-46) 210€
In my country Dell DDR5 are double the price compared to the G.skill listed and I am guessing they need extra cooling to get to 7000+, if the CPU is even able to hit that high.
Horus-Anhur
We also have to consider that you have one massive memory overclock. So it's to be expected that your gains would be big.
-Tj-
And overvloted to the max, not suitable for daily user.
nizzen
1.55v isn't near max. It's doable for 24/7 aircooling 🙂
Now you learned something new today also 😉
I'm betting that more than me is using 1.5-1.55v for ddr4 on Zen 3 too, to run 3800c14 tight 😉 Even my daughter is doing that, and she is doing it on air.
nizzen
I wrote 4000c14 2x16 kit......
4400c19 is Hynix ddr4, and noone want's it for a reason 😛
TLD LARS
Why compare a unicorn DDR4 kit to bare PCB DDR5 kits?
The 4400c19 kit still cost half of what the Dell DDR5 sticks cost.
My 3600 B-die sticks are still 40% less then the bare Dell DDR5 sticks, Half price when the extra cooling on the naked DDR5 sticks are factored in.
Naked DDR5 sticks can reach 80 degrees at 1.35-1.4V.
