Corsair Dominator Platinum RGB DDR5-6600 CL32 2x 16 GB

[P4-630]

Well board, Cpu, memory, Samsung 980pro was 626$ total. 575$ without the SSD.
Replaced my 8700K with 4267mhz B-Die on Maximus X Hero, and it clobbers the living piss out of B-Die and all at like frequencies or less. 4.7ghz vs 4.7ghz I saw roughly an average of 1000pts increase on most of the 3DMarks. Comparing both systems using a GTX 980, not the fastest card in the bunch, but I'm still happy at 1080p myself. Though I do have RTX 2060's, I haven't slapped one in the 12th gen system yet for comparisons.

Here's budgeting performance based on reading TechPowerUp reviews and forum threads.
This is the review page that got me looking into the particular kit I'm using right now.

Patriot Reveals Signature DDR5 Memory Series

PATRIOT, a global leader in performance memory, SSDs, gaming peripherals, and flash storage solutions, today is proud to announce their first DDR5 high-performance memory will be available in the market in November 2021. The Signature DDR5 memory is approximately 75% faster performance over...

www.techpowerup.com

This screen shot below.
12400F + C36 series overclocked from 5200mhz using XMP timing set and additional Vdimm 1.250v
Motherboard is an Asus B660-G. I chose this specific board for BCLK overclocking specifically.

Note: Cpu at 4.7ghz Cache at 4.4ghz. (hardly breathing.... let the screeny speak for it's self)




View attachment 258492

Mine 69,7 ns at the same speed.
What is not right, yours compared to mine

 

[INSTG8R]

Crap. I knew I forgot something. I am aware of the hardware lighting and funny thing is with a DDR4 review, I turned the LEDs pink to test it and a year later put it in another computer. Yeah they are still pink because I just didn't want to install icue again lol.

I had a set of Dominator DDR4 and absolutely dreaded "needing" iCUE, turns out you just had to set it once and is just "remembered" as you experienced. Of course iCue was promptly uninstalled and never needed again...

 

[ThrashZone]

Mine 69,7 ns at the same speed.
What is not right, yours compared to mine
View attachment 258508

Hi,
That's what I was talking about the high latency of 6000 c36/... yours being nearly 70

 

[Asni]

Mine 69,7 ns at the same speed.
What is not right, yours compared to mine

That's what happens if you don't tweak secondary timings (tFAW,tRRD and tRFC). People keep talking about first word latency (2000/freq*cas) but primary timings don't really matter.

 

[Deleted member 185158]

Mine 69,7 ns at the same speed.
What is not right, yours compared to mine
View attachment 258508

Of course it aint right....

We are trying to compare Memory bandwidth with one of the most unreliable and inaccurate benchmarks available. We might as well compare systems on UserBenchmark. Which they say my rig is in the 100th percentile, meaning way badass. But is it believable??? 12400F vs 12700K? Is there a difference? Does it matter?

I think @buildzoid would agree that we should be using Intel MLC which is vastly more accurate and will test latency beyond 128mb as well as being consistent.

But hey, thought I'd just roll with the sheep like the usual and see what's what. Your latency needs improvement.

 

[ir_cow]

That's what happens if you don't tweak secondary timings (tFAW,tRRD and tRFC). People keep talking about first word latency (2000/freq*cas) but primary timings don't really matter.

The first read of AID64 latency is usually always high. Double click the square you want to retest and be amazed how it drops 5-10 . Do that a few times and it should bounce around 1-3 in value. Take that reading 3 times and you have the average value.

 

[Deleted member 185158]

The first read of AID64 latency is usually always high. Double click the square you want to retest and be amazed how it drops 5-10 . Do that a few times and it should bounce around 1-3 in value. Take that reading 3 times and you have the average value.

Or just use MLC. There's even a GUI and quick test for it too. Later I can shoot some examples and see if the forum will let me upload the GUI for it so it's easier for everyone.

Only if you all are interested of course.

 

[justiceleague]

This only proves DDR5 RAM is still a waste of money. Still too slow, still too expensive, still no new DRAM fabs till 2024. I know there are always growing pains for a new DDR standard, but DDR5 seems at least a year too early. Who knows how well these boards will even work with the good stuff when it finally comes out.

Is twice the frequency as ddr4 that is slow? Some games like overwatch benefits greatly on high frequency than latency. Using some of the best ddr4 Samsung b die rams vs a 6000mhz ddr5 kit you gain upwards of 30-50fps on 1080p and with huge improvement on the 1% low. You can get a 6000mhz 32gb for around $200 that is actually cheaper than a good Samsung b die ddr4 16gb kit

 

[dicobalt]

Is twice the frequency as ddr4 that is slow? Some games like overwatch benefits greatly on high frequency than latency. Using some of the best ddr4 Samsung b die rams vs a 6000mhz ddr5 kit you gain upwards of 30-50fps on 1080p and with huge improvement on the 1% low. You can get a 6000mhz 32gb for around $200 that is actually cheaper than a good Samsung b die ddr4 16gb kit

That comes with (more often than not) over twice the latency. Have a look at PcPartPicker, it shows longer first word latency.

 

[ir_cow]

Hey I noticed in your screen shots in the review of CPU-Z displays 2200mhz Cache frequency. Just wondering if performance would be a lot greater running a fixed frequency? Another thing I've noticed with my platforms that turning off HT seems to increase memory bandwidth with a decent percentage. Have you witnessed that by any chance?

@ShrimpBrime -retired I went back and checked why it was 2200Mhz. On ASUS Hero (probably others) by default it is set to bin down when idle. If I put a load on the CPU, it jumps to 3600MHz. I have yet to look deeply into the performance gains from higher cache frequency. I'm not sure if a BIOS update fixed this or it is just how Alder Lake works, but the cache had to be 0-2x below the E-cores or the system wouldn't boot. Only way around this was to disable the E-cores and you could go up to like 4600 easily. Since E-cores on the 12th gen are pretty much limited to 4.3Ghz, that is the limit for cache as well. I do the benchmarks with a locked CPU at 5Ghz P / 3.8 E (and 3.6 Cache *auto*) for a clean comparison between memory kits.

 

[Deleted member 185158]

@ShrimpBrime -retired I went back and checked why it was 2200Mhz. On ASUS Hero (probably others) by default it is set to bin down when idle. If I put a load on the CPU, it jumps to 3600MHz. I have yet to look deeply into the performance gains from higher cache frequency. I'm not sure if a BIOS update fixed this or it is just how Alder Lake works, but the cache had to be 0-2x below the E-cores or the system wouldn't boot. Only way around this was to disable the E-cores and you could go up to like 4600 easily. Since E-cores on the 12th gen are pretty much limited to 4.3Ghz, that is the limit for cache as well. I do the benchmarks with a locked CPU at 5Ghz P / 3.8 E (and 3.6 Cache *auto*) for a clean comparison between memory kits.

That really helps explain a few things. Like the gross waste of die space, could have just added 2 or 4 more performance cores and called it a day. That's my opinion only.

I'm rocking a 12400F, so getting a grip on memory performance with E cores is quite baffling to me, why the big little design on desktop.

If you ever have time, try e cores and ht disabled and play with the performance numbers.

I don't do a lot of gaming, I'd imagine though from past experience to turn these things off to gain per core and memory performance.

Just food with thoughts.