Ryzen 3600X Tested Tight Ram Timings 2133/2400/2933/3200/3733/4000/4200

gerardfraser said:
dasa good charts and all ,I remember went you posted them back in Sept of 2016 on Intel machine.Since there is no real overclocking with Ryzen do you think these test apply to Ryzen in 2019.
Do you think Ryzen owners should not only downclock there CPU but also lower there settings and resolution also.Thanks for sharing ,I guess I could just go downclock Ryzen by 500Mhz to base clock ,lower settings /resolution and see how that works out but it will no be real gameplay scenario for me.
https://www.overclock.net/forum/180...0c11-2133c9-ddr4-2133c15-3000c12-4000c17.html
Click to expand...

Hell no. I wouldn't spend the money on a brand new Ryzen 7 or 9 system just to downclock it and lower my settings. The reason we lower details and resolution is to isolate the CPU to show differences between CPU's.

/dev/null said:
I pretty much aim for CL16 3200 or CL15 3000. CL14 @ 3200 is generally really expensive ram.
Click to expand...

You wouldn't get much for the money spent either. I've tested it on a few applications and that's the conclusion I've drawn. It shows up in some benchmarks, but that's about it.

Ready4Dis said:
Thanks for the great effort put forth. These Zen2 cpu's don't seem near as reliant on high speed ram as zen/zen+ cpu's, which is great to know we don't have to overspend on ram for the system to perform close to it's highest capacity. Would love to see some of these tested with other benchmarks besides just games as well, but I understand the amount of effort this takes and appreciate you just putting this together for everyone!
Click to expand...

Ryzen has never been reliant on memory clock speeds. Once your past DDR4 3000MHz or so, that's pretty much it. You'll probably see more impact from latency tuning than raw speed increases. Most people couldn't get Ryzen to go past DDR4 3400MHz anyway. Understand that when Ryzen 1000 series CPU's came out, DDR4 2133MHz and DDR4 2600MHz were far more common than they are now. Tons of articles came out professing how Ryzen benefits from more speed, but that wall was generally around 3200MHz give or take. I saw boards that could do DDR4 3600MHz and ones that could barely manage 2933MHz back then. Most were around 3000-3200MHz and that was it. AGESA code has been tweaked and greater speeds may be possible on some of those boards now, but not back then.

Essentially, the point of diminishing returns on RAM keeps going up a little each generation. 4 or 5 years ago, that was around DDR4 2666MHz. We had 3000MHz kits back then. They were really expensive and had horrendous timings, but they were there. Benchmarks at the time showed no reason to go for RAM that fast or faster. Now the line is around DDR4 3200MHz. This is because as the memory technology matures, so do the designs and applications that use it. Now, DDR4 3000MHz modules have good timings. Hell, we have faster modules with higher clocks than that and better timings. We've also seen Intel and AMD's architectures evolve and the newer CPU's scale better with RAM clocks performance wise. Haswell-E? 5 years ago? Not so much.
 
gerardfraser said:
I do not disagree with you ,I just shared my little part .Anyone can interpret it as they want to.If I can save someone a couple bucks by letting them know there new shinny Ryzen and new 2070Super /5700XT does not need a new X570 motherboard or update on new ram and what they have will supply the same gaming experience ,I say yep i did my part in sharing some information.
I have Intel gear and other GPU's but I do not care to test them or share anything about them as it has all been done before,but I can say that how I tested is not a common test.
Click to expand...
Fair enough it is a lot of work for which you would get nothing but some praise and more (constructive?) criticism :)

I just get a little frustrated when people who look at high res GPU limited tests conclude that RAM speed can't help when the benefits to the CPU are still there just as they are at 720p but there in reserve hidden by the GPU till you hit a section of gamplay where extra CPU performance is needed.
Generally nobody compares different speed CPU vs RAM speeds which is something that I just don't get

Here is a look at RAM speeds vs CPU speeds 2700X, 3700X, 9700K, 3000 v 3600 but they don't go into latency at all.
https://www.eurogamer.net/articles/digitalfoundry-2019-amd-ryzen-7-3700x-review?page=4
Edit: Snip, sorry was looking at the wrong data.

Farcry 5 shows the largest gains from RAM at around 7-12%
While Crysis 3 goes from seeing large gains with 2700X to much smaller gains from higher RAM speed on 3700X which is probably due to the amount of cache being mostly sufficient for this game.
 
Last edited:
dasa said:
Fair enough it is a lot of work for which you would get nothing but some praise and more (constructive?) criticism :)

I just get a little frustrated when people who look at high res GPU limited tests conclude that RAM speed can't help when the benefits to the CPU are still there just as they are at 720p but there in reserve hidden by the GPU till you hit a section of gamplay where extra CPU performance is needed.
Generally nobody compares different speed CPU vs RAM speeds which is something that I just don't get

Here is a look at RAM speeds vs CPU speeds 2700X, 3700X, 9700K, 3000 v 3600 but they don't go into latency at all.
https://www.eurogamer.net/articles/digitalfoundry-2019-amd-ryzen-7-3700x-review?page=4
Farcry 5 shows the largest gains from RAM so much so that 2700X with 3600 is faster than 3700X with 3000

i7 9700K | 3600MHz RAM 138FPS
i7 9700K | 3000MHz RAM 116FPS
R7 3700X | 3600MHz RAM 112FPS
R7 3700X | 3000MHz RAM 92FPS
R7 2700X | 3600MHz RAM 97FPS
R7 2700X | 3000MHz RAM 93FPS

While Crysis 3 goes from seeing large gains with 2700X to much smaller gains from higher RAM speed on 3700X which is probably due to the amount of cache being mostly sufficient for this game.
Click to expand...

I think the issue is that each game engine has to be evaluated on its own. Most people have neither the inclination, the hardware or the time to investigate performance all that deeply. Just the data I gathered for Destiny 2 alone took quite a bit of work. I only compared two processors but that was illuminating.
 
Dan_D said:
Hell no. I wouldn't spend the money on a brand new Ryzen 7 or 9 system just to downclock it and lower my settings. The reason we lower details and resolution is to isolate the CPU to show differences between CPU's.



You wouldn't get much for the money spent either. I've tested it on a few applications and that's the conclusion I've drawn. It shows up in some benchmarks, but that's about it.



Ryzen has never been reliant on memory clock speeds. Once your past DDR4 3000MHz or so, that's pretty much it. You'll probably see more impact from latency tuning than raw speed increases. Most people couldn't get Ryzen to go past DDR4 3400MHz anyway. Understand that when Ryzen 1000 series CPU's came out, DDR4 2133MHz and DDR4 2600MHz were far more common than they are now. Tons of articles came out professing how Ryzen benefits from more speed, but that wall was generally around 3200MHz give or take. I saw boards that could do DDR4 3600MHz and ones that could barely manage 2933MHz back then. Most were around 3000-3200MHz and that was it. AGESA code has been tweaked and greater speeds may be possible on some of those boards now, but not back then.

Essentially, the point of diminishing returns on RAM keeps going up a little each generation. 4 or 5 years ago, that was around DDR4 2666MHz. We had 3000MHz kits back then. They were really expensive and had horrendous timings, but they were there. Benchmarks at the time showed no reason to go for RAM that fast or faster. Now the line is around DDR4 3200MHz. This is because as the memory technology matures, so do the designs and applications that use it. Now, DDR4 3000MHz modules have good timings. Hell, we have faster modules with higher clocks than that and better timings. We've also seen Intel and AMD's architectures evolve and the newer CPU's scale better with RAM clocks performance wise. Haswell-E? 5 years ago? Not so much.
Click to expand...

I must be confusing it with the APU versions (those were due to the sharing of main memory). Or as you said, mostly thinking about 2133mhz vs 3200mhz before the point of diminishing returns takes effect.
 
Ready4Dis said:
I must be confusing it with the APU versions (those were due to the sharing of main memory). Or as you said, mostly thinking about 2133mhz vs 3200mhz before the point of diminishing returns takes effect.
Click to expand...

That reminds me. I have one of the APU's to review. I should test the effect memory speed has on that. That aside, yes, I believe that's the case. Both Ryzen 1000 and Skylake got a reputation for being sensitive to memory clocks, but that's back when DDR4 2666MHz was the sort of average memory speed in enthusiast spaces. A lot of people ran slower RAM than that. Gains were shown on both architectures going up from 2133MHz. I remember testing this on Haswell-E and seeing little benefit to increasing memory speeds at all. At least, not as far as gaming and common applications were concerned.
 
Sounds good, looking forward to seeing it as I am a fan on AMD's APU's as they're actually useful. Been wanting to do a really small mini itx build (I have one, but has a discrete GPU so not as small as it could be), but honestly can't find a real reason as I already have more computers than I need.
 
I debated whether or not to do it, but someone requested that we review it so I told AMD yes when they asked if we wanted to sample one. I also debated the merits of any game testing on it, and decided that ultimately we needed to test the actual APU part of it as its intrinsic to the processor. We are focused on PC gaming primarily, just as HardOCP was, but I did review an APU for the [H] way back in the day. Sometimes gamers and enthusiasts have secondary needs or they are extremely budget constrained, so I saw value in testing it.
 
Dan_D said:
I debated whether or not to do it, but someone requested that we review it so I told AMD yes when they asked if we wanted to sample one. I also debated the merits of any game testing on it, and decided that ultimately we needed to test the actual APU part of it as its intrinsic to the processor. We are focused on PC gaming primarily, just as HardOCP was, but I did review an APU for the [H] way back in the day. Sometimes gamers and enthusiasts have secondary needs or they are extremely budget constrained, so I saw value in testing it.
Click to expand...

Hey, I game on my 8550U-based ultrabook, and still have an ancient 17" Clevo with an i7-3150 or some such and GTX675M that I use for the bedroom TV, both of which an APU should handily beat.

It's a different set of games from the sig rig to be sure- MMOs, MOBAs, older releases. But even that ultrabook will put out 60FPS in League of Legends at 1080p.
 
IdiotInCharge said:
Hey, I game on my 8550U-based ultrabook, and still have an ancient 17" Clevo with an i7-3150 or some such and GTX675M that I use for the bedroom TV, both of which an APU should handily beat.

It's a different set of games from the sig rig to be sure- MMOs, MOBAs, older releases. But even that ultrabook will put out 60FPS in League of Legends at 1080p.
Click to expand...

See, I knew that APU performance had to matter to someone.
 
Dan_D said:
See, I knew that APU performance had to matter to someone.
Click to expand...

Dude I want a 1080p120 VRR panel on my next ultrabook- even if it still has Intel integrated. Would take an AMD APU in a heartbeat if they could get the power draw and heat output within range.
 
Dan_D said:
I debated whether or not to do it, but someone requested that we review it so I told AMD yes when they asked if we wanted to sample one. I also debated the merits of any game testing on it, and decided that ultimately we needed to test the actual APU part of it as its intrinsic to the processor. We are focused on PC gaming primarily, just as HardOCP was, but I did review an APU for the [H] way back in the day. Sometimes gamers and enthusiasts have secondary needs or they are extremely budget constrained, so I saw value in testing it.
Click to expand...
Having 3 kids I am often lugging hardware around to their friends houses or grandparents, and when they have friends over they also use a computer sometimes if they don't bring there own. I was looking at building a nice portable mini it's with a 3d printed case, low profile nactura fan, pico psu with an AMD GPU just for these instances. Much easier than lugging a full size and if it can run games decent it enough to get by it's good enough!

*Changed can to fan... Typing on my phone and it sometimes thinks it knows what I want to say more than I do..
 
What ITX boards are people using? Been looking to shrink my SMA8 into a Ncase/external radiator under the desk. Looking at a 3600X, 16GB, reuse 1080ti etc...

Edit: After looking into several reviews, the upgrade to 3600X is not worth it. The $50 increase does not warrant an increase of 1 FPS or 1-second reduction.
 
Last edited:
You must log in or register to reply here.
Back
Top