Ryzen - Tweaks?

Bankie

2[H]4U
Joined
Jul 27, 2004
Messages
2,462
Hey everyone, so I just built a 3700x system and while the synthetic benchmarks are amazing the gaming performance isn't what I expected (with a 2080ti @ 1440p it's anywhere from a bit slower to equal to my old [email protected]). So I'm looking for things that I may need to do to help it stretch it's legs.

Some of the things I've tried:

1) Set RAM to 3600Mhz and IF to 1800Mhz. Used the Ryzen Mem Calc and tweaked the RAM timings. This helped quite a bit (AC:Odyssey gained like 3fps on the minimum framerate and 14fps on the maximum).
2) Installed Ryzen Master but haven't messed with it that much yet as it seems to reduce the desktop refresh rate to like 20hz? As soon as it's open the mouse movement on the windows desktop just dies.
3) Installed the AMD Ryzen drivers and switched to the Ryzen power plan.

I haven't messed with PBO/overclocking yet as the Asus ROG Strix B450-I board's BIOS is convoluted as all hell and I haven't looked up any guides on it. I mean most motherboards would have have an XMP setting to enable but on this thing the memory timings are enabled through D.O.C.P...

What I've been used to with Intel systems is basically just, "Set FSB, set Multiplier, install chipset drivers, done" but it seems like there's quite a bit more to getting Ryzen working optimally. Is there anything that I'm missing?
 
I think it will really be slower than a 7700K at 4.9ghz unless you're running something that will heavily load up more than 4 threads.

Approximately Ryzen 3rd gen has about 10-15% more IPC than Sky/Kaby/Coffee but approximate operating speed will be around 4.2-4.3 max.

quick and dirty computation:
4.3 ghz ryzen * 1.1 ipc advantage = 4.7ghz sky/kaby/coffee lake core.
 
With a 2080Ti @ 1440p, you're not going to see much difference between the processors unless you use applications that can utilize the extra threads, and that will hold true for even something like a 9900K compared to the 7700K.
 
BrotherMichigan agree 100%
Bankie
Here is a tweak for a few more FPS for your Ryzen CPU. With your mother board,this will give more CPU clock,adjust EDC and test and always at your own risk.I game at 4k but I like the free extra FPS.Here is an example below BF5 at 1080 for example,did at 1080 for nut bars who only think 720/1080 is the only way to test CPU.

Also at 2560x1440 adjust /tune ram timings and your good to go for best results,does not matter if you run DDR4 4200Mhz ram or DDR4 2133Mhz ram.I will also add video for 2933Mhz vs 3866Mhz Ram
so your 3600Mhz is fine,just tune timings.

PBO limits to manual
PPT limit =0
TDC limit =0
EDC limit =1
precision boost overdrive scalar - manual =10x
max cpu boost clock override =200mhz

BF5 Default vs tweak,free FPS


BF5 and GTAV at 1920x1080 with Ram 2933Mhz vs Ram 3866Mhz


RDR2 and Sleeping dogs 1920x1080 and 2560x1440 with Ram 2933Mhz vs Ram 3866Mhz
 
Which games are you comparing exactly? What clock speeds is your 3700X hitting while in games? I've seen some gains on my 3800X @ 4.5Ghz vs my old 3970X @ 4.4Ghz even at 4K.
 
Which games are you comparing exactly? What clock speeds is your 3700X hitting while in games? I've seen some gains on my 3800X @ 4.5Ghz vs my old 3970X @ 4.4Ghz even at 4K.

The most cpu-intensive games I have tested so far are RDR2 and AC:Odyssey. AC:O gained quite a bit of performance from memory timings but it basically just put the 3700x on par with the 7700k versus a few percent behind. The framerates in most other games are so high that it doesn't really matter; LoL for instance is a good bit faster on the 7700k but does it really matter if you're still getting 250+ fps?
 
Last edited:
7700K is still a beast in games which don't gain much from having more than 4 cores. And in those cases, the extra clockspeed is going to win. However, if you look at CPU reviews, there are plenty of games nowadays, which do scale pretty well with 6 cores. And even some which do have appreciable gains with 8 cores.
 
I'll have to disagree with the 7700k being a beast at least in some games. My 5.0GHz chip did not play well with BFV. In-game CPU usage pegged 100% most of the time while my 2080Ti sat at 50-60% usage. Clearly there is a bottleneck with that game. I upgraded to 3900X and the problem was solved now the opposite happens GPU 80-100% and CPU never goes over 50% usage on Ultra settings. I'm not familiar with AC:Odyssey but I suspect it would not be much different.
This video may be of some help tuning you setting for better performance.
 
Yeah, those are the thread-limited situations I was mentioning. Shadow of the Tomb Raider is another one that seems to like access to more threads.
 
I'll have to disagree with the 7700k being a beast at least in some games. My 5.0GHz chip did not play well with BFV. In-game CPU usage pegged 100% most of the time while my 2080Ti sat at 50-60% usage. Clearly there is a bottleneck with that game. I upgraded to 3900X and the problem was solved now the opposite happens GPU 80-100% and CPU never goes over 50% usage on Ultra settings. I'm not familiar with AC:Odyssey but I suspect it would not be much different.
This video may be of some help tuning you setting for better performance.

I gave up on BF being a worthwhile game long ago so I don't have BFV to test. I don't really need more performance in AC:O; both CPUs just run it and pretty much everything else identically. I don't shut down background apps and frequently have Youtube playing while I game and even though the 7700k can sometimes get to 100% usage it will still usually have a little bit higher FPS than the 3700x which will only be at like 36%. It's just kind of annoying that I spent $250 on a 7700k 3 years ago and 3 years later I can spend the same amount on a CPU that's basically the same in games while also running 10-15C warmer making the system louder. It's pretty awesome on the synthetic benchmark front (my friend with a 9900k was flabbergasted that I benchmarked higher in a few things) but for gaming it's been pretty meh.
 
Last edited:
You haven't specified what the clock speed is when gaming. It may be lower than expected. You're also comparing an extremely highly clocked CPU to a stock one in very limited case examples. Even when the results are in favor of the 7700K in some cases by a few percent, it's not shocking in the least.
 
I've seen it go up to 4.475 on 1-2 cores a handful of times in games. In AC:O it seems to go to 4.375-4.40. This is on a 280mm AIO and the PBO settings listed above.

My friend with the 9900k has the exact same 2080ti as me but his isn't under water like mine. I might have him log in to my Uplay account and see if he gets any better on AC:O or if it is just the limit of the GPU.
 
put EDC to 11 - 15. dont do "1" like post above said, your single thread performance WILL TANK
 
put EDC to 11 - 15. dont do "1" like post above said, your single thread performance WILL TANK
I never understood why a lower value gives better performance that the Motherboard value which is 280A on my board. Can you enlighten me or point me to an article?
 
put EDC to 11 - 15. dont do "1" like post above said, your single thread performance WILL TANK
This is not true if you know what you are doing,can some EDC values with Windows power plans not work 100% at times with single thread or multi thread,why yes that is true.OP can test any value he want's like I suggested
adjust EDC and test and always at your own risk

I never understood why a lower value gives better performance that the Motherboard value which is 280A on my board. Can you enlighten me or point me to an article?
IT is a Bug in BIOS ,which has been around since at least July/August from what I remember and used,here is a thread with some information and different motherboards and CPU 's suggestions.
https://www.overclock.net/forum/13-amd-general/1741052-edc-1-pbo-turbo-boost.html
 
This is not true if you know what you are doing,can some EDC values with Windows power plans not work 100% at times with single thread or multi thread,why yes that is true.OP can test any value he want's like I suggested


IT is a Bug in BIOS ,which has been around since at least July/August from what I remember and used,here is a thread with some information and different motherboards and CPU 's suggestions.
https://www.overclock.net/forum/13-amd-general/1741052-edc-1-pbo-turbo-boost.html

Thanks. Will read. Mine seems to do well when I set the BIOS to set the power to Motherboard instead of Auto. I get boosts of 4.55Ghz according to HWInfo64 and an all core of 4.125Ghz in R20. Although I think that was higher on my other X570 board than this new one.
 
Thanks. Will read. Mine seems to do well when I set the BIOS to set the power to Motherboard instead of Auto. I get boosts of 4.55Ghz according to HWInfo64 and an all core of 4.125Ghz in R20. Although I think that was higher on my other X570 board than this new one.
Well with tweak I can get 4400+Mhz all core Cinebench20 no problems,tweak works fine for me.In single thread and multi.It is worth a try if someone wants an extra 100Mhz + on CPU clocks.
49462447431_e5fb86e65e_k.jpg4400-Cinbench20
 
If all want is a pure gaming machine you wasted your money and should have stayed 7700k.

Basically man... if you CBr15 single thread test a 7700k at 4.9 it might get a 201 but a ryzen 3700 will pull an easy 208 or 210 depending. Sounds to me like your expectations demand more than the top tier hardware in market can deliver.

I dont think people understand just how significantly better a 7nm 3700x is than a 7700k.

Try 30+mb of cache on for size as a starter.

Sorry for typos using phone
 
Last edited:
Well with tweak I can get 4400+Mhz all core Cinebench20 no problems,tweak works fine for me.In single thread and multi.It is worth a try if someone wants an extra 100Mhz + on CPU clocks.
View attachment 2235054400-Cinbench20

I tweaked my 3900X. I get 150Mhz higher all core for 300pts increase in R20 MT, got 11pts higher on Single for a max core of 4.65Ghz on at least 3 cores according to HWInfo64. Highest I ever got before was 4.525Ghz. All core I am running 4.275Ghz compared to 4.125Ghz before. Ended with EDC=16 and PPT=500 and TDC=210, scaler at 6x and 100Hz auto OC(it did best here and not with it at 200). If I do 0 on those others I get stock power locked so it would not go past the stock Wattage. I did the EDC as suggested on the link above for the 3900X processors. Running AIDA64 I got at least 200Mhz above previous ratings which put me right behind the 3950X in a few of the benches, there used to be an Intel in between in those benches, the CPU Queen bench came in at 4.625Ghz.

Temps went up about 4C on the tests over previous. Crept to 72C where it was at 68-69C during R20.

Oh and power plan set to Ultimate Performance
 
Last edited:
Sounds great,that 3900X is a great CPU as are all the Ryzen new CPU's.I set mine as follows
49431899601_595a301865_b.jpg
 
Sounds great,that 3900X is a great CPU as are all the Ryzen new CPU's.I set mine as followsView attachment 223592

Yeah I did not try a 1 on EDC because that thread you pointed out said the 3900 and 3950 were different. So I tried that suggestion first. I did set the PPT and TDC to 0 but that default to the AMD ratings and would not go past those limits power wise. Set the way I have it they go up to the appropriate levels. Set at 0 limited my PPT to like 105W and it would not go past that. 10X on the scalar actually lowered my max clock, 6 kept it where it could hit 4.6 or higher on some cores. My shitlet will not push any core past 4450 where my good one can push several to 4650. I think it ended up good. R20 MT is 7300 and Single is 511. Before I could barely hit 7000 except once or twice to 7002. I could hit as hit as 501 single thread but normally it was 498. As for max boost I did 150, I can only go to 200 and it seemed to get lower scores by a few points. So I left it there, I started at 100. So far 5 of my 6 cores on the CCD1 hit 4600 while the remaining has hit 4575.
 
I am not sure why people like the Cinebench20 scores but your single thread scores are still off with 3900X,should be 530+ with the tweak because of the extra CPU Mhz.
I only play games so I am happy with the AMD Ryzen's I owned but looking for second hand 3900X to play with

Video:Highest normal Boost is still 4650Mhz. With 4600Mhz average boost clock




3800X light gaming
49430283162_81e7b480a2_k.jpg


AMD 3800X Bclk 4700Mhz+
49391171163_62f3eed2f4_h.jpg
 
I am not sure why people like the Cinebench20 scores but your single thread scores are still off with 3900X,should be 530+ with the tweak because of the extra CPU Mhz.
I only play games so I am happy with the AMD Ryzen's I owned but looking for second hand 3900X to play with

Video:Highest normal Boost is still 4650Mhz. With 4600Mhz average boost clock




3800X light gaming
View attachment 223601

AMD 3800X Bclk 4700Mhz+
View attachment 223602


I have yet to see a review of the 3900X hitting 530+ single thread. And the Tweak, if you read the overclock.net, does not really help single thread. Now mine actually performs single thread to what the reviewers have got. But multi is 200 pts over most reviews. So I am ok with that. Mine was not even hitting what most reviewers got multi threaded, and I have tried 2 X570 boards so I doubt it is that. I have six cores that don't seem to go over 4.45Ghz and most stay at 4.4ghz on that CCD. Hell just looking at a 3950x review it said R20 single was 511 on his test.

At the end of the day though I should not have to do any of these tweaks in the BIOS, I should be able to just turn PBO on and maybe tell it to use the motherboard power settings. Which enabling PBO is supposed to do, allow more power. I should not have to trick it to get it to operate withing spec.
 
OK Cinebench is like a drug for some reason,you scores are low and review scores are low and I agree with ,should have to not do any tweaks but for some reason you either do not know or never been told.AMD before release of Ryzen 3xxx or Zen2 implemented the 200Mhz


Second you literally did not even look at the thread ,I am actually the second poster in the thread,with other post and I have known about the tweak/bug sine July release.
I see you are saying single thread scores are low with the tweak in Cinbench20 at EDC1. Did you actually try a different Windows 10 powerplan.
I suggest Ultimate Power plan vs using (which is a great power plan)1usmus power plan or power saver plans.
My 3800x just setting EDC 1 in BIOS and scaler 10x and nothing else Cinebench20 still 530 score with Ultimate power plan but power saver plans I have seen 478.Try it out if you want or anyone.I suggest back to back runs.

This screen shot is not the highest or even average for my 3800x,so I guess I would have to agree with single scores are low.
Desktop-Screenshot-2020-01-24-09-27-58-82.png


Here is a screen grab of a youtube video I did in cinebench20 in 2019.
Desktop Screenshot 2020.02.15 - 20.46.16.82.png
 
So trying to help people,get negative people responses ,so I whipped out my dick,now beat it bitch.

I am going to put this here,Cinebench20/15 is not a mesure of Ryzen CPU people.

Anyway I just did 4 Cinebench20 runs one after another, with Ryzen tweak/bug 3800x.
Jamie Marsala is correct,I went from 541 single to 531 single while recording runs,I cut the video ,so it is only last part of last run.


I should say I am j?k and do not read this like post like other post and get triggered because you do not comprehend how stuff works
 
Last edited:
So trying to help people,get negative people responses ,so I whipped out my dick,now beat it bitch.

I am going to put this here,Cinebench20/15 is not a mesure of Ryzen CPU people.

Anyway I just did 4 Cinebench20 runs one after another, with Ryzen tweak/bug 3800x.
Jamie Marsala is correct,I went from 541 single to 531 single while recording runs,I cut the video ,so it is only last part of last run.


I should say I am j?k and do not read this like post like other post and get triggered because you do not comprehend how stuff works


Oh I was just saying this stuff helped. I only use R20 as my own bench really, sure it is nice to compare with others as well but every system is different. I use it to see what has changed and if it is for the good or bad. Same reason I use the same stretch of road to test the tune revisions on my car, so I have a baseline and then somewhere to hopefully get to. It is great that your 3800x hits those numbers as they seem to get higher clocks than the 3900x does. I may just have a not so good chip since there really seems to be a heck of a difference between CCD1 and CCD2. I am just trying to get the most for what I paid. Maybe someday the BIOS will catch up to the tweaks but I assume AMD does it to keep the power levels, Watts, in check. But it would be nice if out of the box it would at least come close to the 4.6Ghz, mine never did on stock settings with no tweaks or changes to BIOS.

I will add that it did help my AIDA64 numbers increase nicely as well. And my Time Spy numbers increased as well. I really was not that bored today to run anything else.
 
If you read my whole post ,I was fucking with yeah and of course it is always prudent to compare with others on your own system with reviews.
Just to let you know again,My 3600x x 2 scored higher than your 3900x in single thread cinebench .It is not an insult on you,it me saying to you ,you can improve with such a great CPU.Also review guys are playing on a different playing field,they are stuck to out of the box performance on which ever gear they have at the time.I am always blown away that people can not realize that stuff.
 
If you read my whole post ,I was fucking with yeah and of course it is always prudent to compare with others on your own system with reviews.
Just to let you know again,My 3600x x 2 scored higher than your 3900x in single thread cinebench .It is not an insult on you,it me saying to you ,you can improve with such a great CPU.Also review guys are playing on a different playing field,they are stuck to out of the box performance on which ever gear they have at the time.I am always blown away that people can not realize that stuff.

I keep feeling it should be higher as well, but like I said comparing to others and online reviews it seems inline with them. I saw a reviewer that got a 511 on the 3950x sample as well and one that got 524. Maybe the big ones just can not single core clock as high as the single CCD models. Who knows. That same reviewer that got a 524 on R20 on the 3950 got a 502 on the 3900 and a 506 on the 3800x. So it would seem the 3800x shines.

And I figured you were fucking around. But you never can tell. lol. I am just happy that I am getting better than what the reviewers got on the multi thread. Which it is now. And these reviewers are using the MSI Godlike board where I did not want to spend quite that much and got a MSI X570 Ace instead. I am sure if I played around with that EDC value that maybe I could somehow find out if this board has a magic number but I am a little saddened that you need to set it low to get it to boost to where it should be. I mean my board goes to 280 on its own when set to Motherboard instead of AMD power settings. And when benching it my EDC is 700% over that 16. I mean just by changing that EDC value I get 150Mhz more on my all core and close to what I can get stable OC manually. Best I got was 4.325Ghz with just adjusting the voltage. The downside is now running R20 I am at 72C instead of 68-69, which pushing it past that magic 70C mark may be limiting me more. Next step may be to bump up from a 280 AIO to a 360. Anyway this thread helped me at least get here. I will keep playing.
 
You have a great CPU as is,it is fun to mess around with the CPU's but I would not judge the performance on a Cinebench run ,I would play games and some real world things like encoding or something.Remember Ryzen 3900X @4300Mhz is as fast as Intel @5000Mhz.
I have the cheapest MSI X470 Gaming Plus motherboard and I have all the performance I want from the CPU/Motherboard.My time to sell X570 and go X470 LOL J/K
 
LOL! I am hoping the 4000 series will work with the X570 and will see what those look like. Although I really do not need the 3900x, I just wanted it. After I am doing playing I will set it back to just normal PBO. Seems for some reason I am getting re-boots when I am nowhere near the system. like asleep or out of the house. Temps are not an issue so who knows, could be a power hit and I just am not here to see it. I do need a UPS just to cheap to get one. I did set the scalar back to auto instead of 6x and the temps are down a hair at all core 100% now.
 
I am buying a 3900X ,just waiting on used ones.I owned 2x3600X/3800X and I enjoy just messing around also. I think the Ryzen 4000 will work depending if AMD releases it in 2020.
 
Put CPU Voltage on Normal and both PBO settings to enable. Im getting boost clocks up to 4650ghz on a number of cores.
 
Put CPU Voltage on Normal and both PBO settings to enable. Im getting boost clocks up to 4650ghz on a number of cores.

Oh I tried that. Never got a a core to go over 4550 according to HWInfo64, most would not even hit 4500 ever. Now all cores on CCD1 hit at least 4600. None on CCD2 hit over 4450. I was happy before and just wanted to see what it really could do. Now I know it will at least hit where it is supposed to. It hits those speeds when not testing now.

If I leave it all stock, stock power levels, and PBO set to enabled then the max all core speed I get is 4.0Ghz, If I kick the power setting to Motherboard then I get 4.125Ghz. With the EDC set to 16 I get 4.275Ghz. And now normally it is trying to idle at 4.4Ghz instead of 4.25Ghz.
 
Oh I tried that. Never got a a core to go over 4550 according to HWInfo64, most would not even hit 4500 ever. Now all cores on CCD1 hit at least 4600. None on CCD2 hit over 4450. I was happy before and just wanted to see what it really could do. Now I know it will at least hit where it is supposed to. It hits those speeds when not testing now.
Not that it probably matters I meant to say the bios setting CPU VDDP and setting it to Normal, also I use MSI Afterburner along with HWinfo to monitor CPU clocks, voltage and temps.
 
Not that it probably matters I meant to say the bios setting CPU VDDP and setting it to Normal, also I use MSI Afterburner along with HWinfo to monitor CPU clocks, voltage and temps.

Yeah my voltages are set to Auto. I tried the AMD Overclocking setting with Auto still on and that did nothing different. So I set it back to auto. I had the Scalar set to 6x which gave me a few point on R20 single thread but more heat on Multi so I set it back to Auto. I just ordered some quieter but better fans for my 280 so I can run the fans at a higher RPM to keep the AIO cooler. I can only take them at about 50% during normal use, under 55C proc temp. Anything higher annoys me. But that only keeps my AIO Coolant temp just below 31C, ambient at my floor is 17C, where the fans would suck in from. I am thinking with some more air flow through the Rad I could get the temp down a hair to thus keep the processor temp down below 70 at full load. Well plus I wanted some RGB front fans for this all glass paneled case. Lol. If I can find 8 screws for the old fans to screw into the back side of the rad I may try a push/pull. But then I would add the noisy fans back in the mix so we shall see.
 
One thing that makes a huge difference is ambient temps, if the room where your system is at 70F that will be your starting point for everything, it doesn't matter if your using air or a AIO. I keep a cool house year round, Winter is typically 64F on the thermostat, so its probably a couple degrees colder than that in the computer room on the other side of the house, during the summer we set the central AC to 68F. Electric bill sucks but Im cool and so is the computer LOL! Im able to use the AMD Wraith and keep my temps the same or even better than a AIO, I typically never break 65-66C at full load, a big part of that is the low ambient temps.
 
Last edited:
One thing that makes a huge difference is ambient temps, if the room where your system is at 70F that will be your starting point for everything, it doesn't matter if your using air or a AIO. I keep a cool house year round, Winter is typically 64F on the thermostat, so its probably a couple degrees colder than that in the computer room on the other side of the house, during the summer we set the central AC to 68F. Electric bill sucks but Im cool and so is the computer LOL! Im able to use the AMD Wraith and keep my temps the same or even better than a AIO, I typically never break 65-66F at full load, a big part of that is the low ambient temps.

I'm assuming you mean C?
 
One thing that makes a huge difference is ambient temps, if the room where your system is at 70F that will be your starting point for everything, it doesn't matter if your using air or a AIO. I keep a cool house year round, Winter is typically 64F on the thermostat, so its probably a couple degrees colder than that in the computer room on the other side of the house, during the summer we set the central AC to 68F. Electric bill sucks but Im cool and so is the computer LOL! Im able to use the AMD Wraith and keep my temps the same or even better than a AIO, I typically never break 65-66F at full load, a big part of that is the low ambient temps.

My floor, which the computer sits on, is hardwood and the floor temp is 63F or 17C. The coolant temp is at ~31C so it is a bit warmer than Ambient. I am hoping better fans will help this situation. I do not expect the coolant to be the same as ambient but a little closer would be nice, or the 3900X is just a mini cooker that keeps the liquid that warm just idling. The CPU is in the low 40C range when just doing normal stuff, not gaming. So if I can drop the coolant temp a few degrees than that would be great. Not a big deal since the processor does not sit pegged at 100% load pretty much ever but I like a cooler starting point.
 
Back
Top