Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Jun 19, 2020.
Just adding in my 2c, using Hynix AFR. (2x16GB sticks)
Try 16-19-19-32 fabric of 1900 tied to 3800mhz RAM. Made a big difference in my system running 64gb of RAM. Think it was 58 or 54 other sub timings with 495 trfc I think.
Same thing only 3600mhz/1800mhz on my 128gb setup on my tRX40.
3800mhz or bust baby!! Same here with my 3900X. Wonder what numbers my tRX40 would throw at 3800/1900 setup on my 128Gb. She's not a fan with all eight DIMMs setup unfortunately. 3600mhz it is for now.
Thanks. Took a lot of trial and error to get it stable at that speed. But i think i've just got a golden set of AFR modules.
I should note. The Ryzen DRAM tool only goes up to 3533 for AFR modules.
Thanks for the article!
Thanks when I have time will try with those timings, I use to have 3800 with 1900 IF but while it was 100% stable in windows (game & P95), half the time in stuck in bootloop. In the end I settled for 3733.
Cool, so 3733 MHz with 1:1 Infinity divider, 1T, and at least CL16.
Right on. I'm going to assume I've been lucky as well. 128Gb RAM from three different kits. All the same Trident Z Neo from GSkill, but bought three separate times.
32Gb initially. Another 32 to round it off on my X570 Xtreme. And another 64Gb kit to complete it on my tRX40.
Now I need to treat the 570 with some more love and get her some more RAM to munch on as she's on a paltry 16Gb 3800mhz.
That is after I watercool it all this week. Just got the order in for father's day baby! Triple 140mm rad, dual 140 and a single 120mm going all inside my fractal 7.
I have been on Intel for ages, and was thinking to go AMD for my next build. But the more i read about Ryzen the more i want to go with Intel again.
I am not keen to use multiple tools and calculators to make my CPU and Memory work properly.
I dont want to buy all the fancy new hardware, just to find out, oh bugger this memory is not the right mix and loosing 30 percent of performance because of that.
If you even read the article you would have seen that if you choose a good 3600 kit with preferably 4 DIMMs, you'are set just with the XMP profile. You don't need to do anything else. A good 3200 kit will also behave the same. If you consider that "complicated", perhaps PC is not the proper platform for you. There are no "30%" performance loses anywhere, and if investing on a $150+ memory kit means that you can't read a single paragraph, then you'll get to issues and lost performance with Intel too.
I have been hoping for that kind of answer, thanks. Now my interest is back on AMD.
Well i didn't say it's complicated, nothing againg some BIOS settings. But seeing all those Apps to tweak the proper setting seemed to me as AMDs CPUs are still in development and we all are the beta testers.
Yes it is true, i haven't read all the pages and this particular article. I just summed up all the informations i have seen last couple of months, for example in the AMD Tweaking thread and some youtube videos.
The 30 percent loss was overexaggerated of course. Just wanted to make it more dramatic.
But you explained it already that it is me. I was just overthinking everything.
Great article! I did a test between 3800MHz CL16 1:1:1 32GB (4x8GB) and 3200MHz CL14 1:2:2 64GB (4x16GB) with Far Cry 5's Benchmark. This was done on a MSI - MEG X570 ACE. I ran it 4 times and restarted the game for the 4th run. The averages are 3800MHz CL16: MIN 104.25 FPS, AVE 127.5 FPS, and MAX 153 FPS. 3200MHz CL14: MIN 98 FPS, AVE 123 FPS, and MAX 147.5 FPS. The game was on Ultra settings and the GPU is a 1080 Ti.
You will also find that this amount of tweaking gives performance to Intel too. It's a matter of time vs gained performance, like anything. The setup itself is easy, thanks to articles like this. So a 3600 CL 14 kit, with 4 DIMMs, is really the sweet spot and it should work with just setting the XMP profile in the BIOS, like with any Intel system.
You should your report success (+failures) to Yuri (use Twitter, he reads it !)
That's how all those profiles appeared, people telling him what worked and what note.
There's also some google docs spreadsheets that anyone can fill (links in the tool itself)
Thanks Krzysztof. This is a good article. I have no need for it at the time, but it could be very useful in the future. I rarely bookmark specific articles but this one is.
I played with my RAM one evening and quickly decided it wasn't worth the time for the improvements possible. I did the easy thing from the start and bought 3200 CAS14 for my 2400G.
The replies to Loster were similar to what mine would be. Setting the XMP profile is easy and you'll lose little from it with good DDR4.
It's interesting to see that if I need more RAM in the future I could buy two more sticks and possibly get a slight performance improvement to go with the additional capacity.
Ive been running 32gig of Corsair Dominator Platinum 3600 CL18 19 19 39 since January, when i bought this i was running an i7 7700k but after recently changing to Amd Ryzen 7 3800x with an Asus prime x570pro board i wondered if changing to ram with tighter timings would be an improvement so i recently bought 16 gig of Gskill tridentZ 3600 cl 15 15 15 35 which i picked up from local post office today, i can`t for the life of me see any gain, ive run 3d mark time spy and skydiver but both give slightly lower scores of 100 points on each, admittedly the division 2 might be a tad smoother but that might be down to the room temps being a bit cooler with the crappy weather, also tried Arma3 and still see no benefit to having tighter timings, the only question now is....do i keep the tridentZ or the Dominator.
Don't compare scores done in summer and winter times with these new AMD CPUs.
Without even touching any config, they simply boost higher if they are cooler, resulting in (sometimes much) bigger scores.
I did an experiment in December last year on my 3700X, it was -8 degrees outside... got dressed like going into a polar expedition, stopped the heating in my room, opened all windows, waited until ambient INSIDE was -8...
Then ran benchmarks with all fans turned to 100% in that -8 degrees ambient.
Got the highest scores ever, sometimes with a massive 10% extra compared to normal 20-22 degrees ambient.
CPU was running 100% at only 15 degrees... with the freezing air being pushed through the cooler ;-)
In my case(but intel though) going from 4200c17#1.44v(tuned) to 3800 c14 firstname.lastname@example.org bring me 5-8 ms less in Aida latency.
Which, given same gfx2080s give me 5-10% more minimal FPS boost, some games even more. Pita process, but worth it?
Wonder what results I could achieve with this ram on future zen CPU....I really have high hopes they deliver something special in next release.
Ram is patriot viper steel 4400 c19. Highly recommended. 130$ area per 16gb.
Any chance you could repeat testing with Ryzen 5000 series and perhaps compare DR kits with 4000 low latency SR kits?