Getting up to 2933MHz on an AMD-2200G - Does dual-channel matter?

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
I just put together my first AMD based system in years with the following parts:

1 AMD Ryzen 2200G cpu
1 ASRock AB350M Pro 4 motherboard
1 Patriot Viper 4 DDR4 1x8 3000MHz memory stick (from a 2x8GB kit - PV416G300C6K)

I turned on the XMP Profile in the BIOS and set the memory speed @ 2933MHz but the system seems unstable at this selection.

So, I went back into the BIOS and set the memory speed @ 2666MHz.

I'm getting the system to boot up fine @ 2666MHz but not @ 2933MHz. The reason I have a single 1x8GB stick in this system is because I pulled it from it's twin brother in an older system due to the raping and pillaging of the DDR4 memory manufacturers. I'm trying to weather the price gouging storm by splitting my 16GB kit between 2 systems for now but it's tough.

I have an paper insert that I got (see linked image) with my ASRock AB350M Pro4 motherboard and I'm trying to interpret what it means. I'm leaning towards the possibility that your level of success depends on whether you are using 2 sticks in dual-channel mode or just 1 stick in single-channel mode. It actually looks like you are more likely to get the highest rated speeds of your memory if you stick with a single stick of RAM versus running 2 sticks in dual-channel mode. Am I reading the paper right?

v0Zc1yo.png


I also seen on the insert (which I couldn't believe) that the actual physical DIMM slot you use on the motherboard seems to matter. Good grief!!! So, just to have all my t's crossed I moved the single stick of ram from A1 to A2 per the paper insert after I read it. Of course, the paper insert says "Summit Ridge" and not "Raven Ridge" but I'm assuming it's pretty similar for the latter.

I always perceived 4 DIMM slots over 2 DIMM slots on a motherboard as a good thing. After seeing this paper insert I think it may actually be a waste of money on the motherboard itself and the money spent on the faster rated sticks of memory. The more you plug in, the slower they go. WTF?

Is this revelation applicable to the Intel 8th-Gen overclocking motherboards as well? Or is this just some "free gift" (tube of vasoline anyone) with AMDs Ryzen platform?
 
Last edited:
  • Like
Reactions: Drazick

alexruiz

Platinum Member
Sep 21, 2001
2,836
556
126
What BIOS version do you have?
Put the single stick on the second dimm slot from the CPU
 

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
What BIOS version do you have?
Put the single stick on the second dimm slot from the CPU

No, I did that already. I said that in my original post but since I've just finished editing it you may not have seen it then. The single 1x8GB stick is in DIMM slot A2 per the paper insert.
I don't know what BIOS I have but it's the original that came with my motherboard. Or to say it another way, I didn't have to flash it first before I installed my AMD 200G (Thank God!, *wiping the sweat off my brow*).

**EDIT**: I guess I should have titled my question "Getting up to 2933MHz on and AMD 200G - Does dual-channel make it worse???".
 
Last edited:

alexruiz

Platinum Member
Sep 21, 2001
2,836
556
126
You need dual channel for best performance with the APU, and yes, the BIOS version DOES matter.
AsRock AM4 boards got peak RAM compatibility on summit ridge with bios ver 3.30
Versions 4.xx with PinnaclePi AGESA broke it.
Unfortunately, you need BIOS 4.xx for Raven Ridge.
 

whm1974

Diamond Member
Jul 24, 2016
9,436
1,568
126
I always perceived 4 DIMM slots over 2 DIMM slots on a motherboard as a good thing. After seeing this paper insert I think it may actually be a waste of money on the motherboard itself and the money spent on the faster rated sticks of memory. The more you plug in, the slower they go. WTF?

Is this revelation applicable to the Intel 8th-Gen overclocking motherboards as well? Or is this just some "free gift" (tube of vasoline anyone) with AMDs Ryzen platform?
Well yes populating all 4 memory slots will reduce the speed if the modules, but you will not notice the loss anyway. But you gain the ability to upgrade your memory but just adding two more modules instead of replacing two.
 

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
Well yes populating all 4 memory slots will reduce the speed if the modules, but you will not notice the loss anyway. But you gain the ability to upgrade your memory but just adding two more modules instead of replacing two.

That is absolutely unacceptable when memory chips are sold at premiums per their speed ratings. Now that I know this, I will probably make different purchasing decisions in the future. I'll never install memory in more than 2 sticks at a time from now on.

You need dual channel for best performance with the APU, and yes, the BIOS version DOES matter.
AsRock AM4 boards got peak RAM compatibility on summit ridge with bios ver 3.30
Versions 4.xx with PinnaclePi AGESA broke it.
Unfortunately, you need BIOS 4.xx for Raven Ridge.

I'm in the BIOS and it doesn't say "BIOS x.xx" anywhere that I can see. I see "Microcode Update: 810F10/8101004". I'll keep looking though.

EDIT: I guess it's "UEFI Version P4.50".
EDIT #2: You say I need to run the memory in dual-channel mode for best performance and for maximum theoretical bandwidth performance I agree. I never realized this came at the expense of reduced memory speeds, however. I'm going back to read some reviews and figure out why this is never mentioned.
 

whm1974

Diamond Member
Jul 24, 2016
9,436
1,568
126
That is absolutely unacceptable when memory chips are sold at premiums per their speed ratings. Now that I know this, I will probably make different purchasing decisions in the future. I'll never install memory in more than 2 sticks at a time from now on.
The reduction is rather small and I highly doubt that you would even notice at all. And why waste money by throwing away sticks if you need to double your memory?
 

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
The reduction is rather small and I highly doubt that you would even notice at all. And why waste money by throwing away sticks if you need to double your memory?

Obviously, I'm not going to throw away memory sticks especially given the current market prices. But what I will do is use the knowledge to make future purchasing decisions. So that, for instance, if I see a 4x8GB kit versus a 2x16GB kit that I want to purchase. I'm going for the kit with the smallest amount of sticks. Why not, if it's faster? So what, if the loss is 5% 4% or 3%. A loss is a loss and if I make one decision over the other to avoid that loss I'm gonna go that way. All other things being mostly equal.

EDIT: Btw, what you call small is a reduction in memory speed by 266MHz. Now, according to my math on a 2666MHz rating memory kit, that is a 10% reduction in speed. That's not what I would call small or insignificant myself. Who do you work for? Which memory manufacturer? Patriot or G.Skill?

Underlined portion done by moderator.
User call-outs (or accusations) are not allowed in
the tech forums.

AT Mod Usandthem
 
Last edited by a moderator:
  • Like
Reactions: VirtualLarry

whm1974

Diamond Member
Jul 24, 2016
9,436
1,568
126
Obviously, I'm not going to throw away memory sticks especially given the current market prices. But what I will do is use the knowledge to make future purchasing decisions. So that, for instance, if I see a 4x8GB kit versus a 2x16GB kit that I want to purchase. I'm going for the kit with the smallest amount of sticks. Why not, if it's faster? So what, if the loss is 5% 4% or 3%. A loss is a loss and if I make one decision over the other to avoid that loss I'm gonna go that way. All other things being mostly equal.
I would go with 2x16GB stick as well, but mainly in case I would need to double the ram in the future. You do know that modern consumer boards can handle up to 64GB up memory, right?
 

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
I would go with 2x16GB stick as well, but mainly in case I would need to double the ram in the future. You do know that modern consumer boards can handle up to 64GB up memory, right?

You probably missed my edit in the last message but your talking away serious issues like they mean nothing. Your comment total dismisses my entire point.
 
Last edited by a moderator:

whm1974

Diamond Member
Jul 24, 2016
9,436
1,568
126
You probably missed my edit in the last message but your talking away serious issues like they mean nothing. Your comment total dismisses my entire point.

This issue isn't all that serious to begin with. In fact, you are the first person I meet in years who is really making a mountain out of a molehill about this. And no, I don't work for any memory manufacturer or own stock in any.

How come we are not seeing other posters make such a ruckus over this as you are?
 
Last edited by a moderator:
  • Like
Reactions: KompuKare

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
This issue isn't all that serious to begin with. In fact, you are the first person I meet in years who is really making a mountain out of a molehill about this. And no, I don't work for any memory manufacturer or own stock in any.

How come we are not seeing other posters make such a ruckus over this as you are?

So, if we translate your question, you want me to get into everybody else's mind and tell you what they're thinking? How would I know? I speak for me. And I'm making a big deal about it because I care personally. If I spend good money on 4 memory sticks rated and promised to run at a certain speed that won't run at that speed because of the very fact that they are sold in a pack of 4, that responsibility falls on the memory manufacturers back and I don't see ANY memory kits explaining this information. Somebody jump in here and correct me if I'm wrong. But I just learned it today from a paper insert I got from ASRock. So, maybe a lot of people don't know about this issue. Making a ruckus brings awareness and awareness brings change. So maybe it's worth the ruckus.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,570
10,200
126
and promised to run at a certain speed that won't run at that speed
"Overclocking not guaranteed".

Any memory that requires using XMP or AMP, is overclocking the RAM. And as we all know, overclocking is not guaranteed.

So, stick to JEDEC standard RAM ONLY, and run your AMD AM4 CPU or APU at ONLY supported RAM speeds, which is 2667 for Raven Ridge, with single-sided one-DIMM-per-channel.
 
  • Like
Reactions: Drazick and whm1974

whm1974

Diamond Member
Jul 24, 2016
9,436
1,568
126
Actually it is not the memory manufacturers fault but the CPU's memory controllers.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,570
10,200
126
I also seen on the insert (which I couldn't believe) that the actual physical DIMM slot you use on the motherboard seems to matter. Good grief!!!
This has been true now for several years, with all modern DDR4-using platforms. The timings and speeds are so tight, that trace routing and impedance / termination matters.

These aren't your father's Super Socket 7 boards.
 

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
You need dual channel for best performance with the APU, and yes, the BIOS version DOES matter.
AsRock AM4 boards got peak RAM compatibility on summit ridge with bios ver 3.30
Versions 4.xx with PinnaclePi AGESA broke it.
Unfortunately, you need BIOS 4.xx for Raven Ridge.

Heh. I updated the BIOS to UEFI P4.70 and guess what? That did it. The 3000MHz RAM I bought almost 3 years ago is finally running at most of it's maximum potential. "Run Forest, RUN!" The Linux command prompt reports a speed of 2934MHz. So, thank you. I'm a little bit more of a happy camper than I was a few moments ago. I'm learning a lot with this experience.

This has been true now for several years, with all modern DDR4-using platforms. The timings and speeds are so tight, that trace routing and impedance / termination matters.

These aren't your father's Super Socket 7 boards.

I haven't built a new system for awhile. Thanks for the input.

EDIT: Hmmm, now I'll be greedy and stingy and press my luck and try both sticks in dual-channel mode and see if I still get 2934MHz. Hehehehehe, maybe tomorrow.
 
  • Like
Reactions: lightmanek

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
"Overclocking not guaranteed".

Any memory that requires using XMP or AMP, is overclocking the RAM. And as we all know, overclocking is not guaranteed.

So, stick to JEDEC standard RAM ONLY, and run your AMD AM4 CPU or APU at ONLY supported RAM speeds, which is 2667 for Raven Ridge, with single-sided one-DIMM-per-channel.

I completely missed the point you made on my first read through about Raven Ridge only officially supporting 2666MHz memory speeds. I don't think that's correct. I believe they officially support 2933MHz memory speeds. Except your point is being made with a specific chip so I guess I need to look more closely at my chips. Do you have a link to this information similar to the paper insert image I posted above but 100% relevant to Raven Ridge?
 

whm1974

Diamond Member
Jul 24, 2016
9,436
1,568
126
To be clear this "issue" the OP is making a fuss about, is caused by both the memory controller in the CPU and the motherboard's design. Yes it is possible to design a board that would allow for full speeds with all slots populated, but how much are you willing to pay for a consumer motherboard? This is a bit simple, but is mostly a cost reduction decision to allowing for decreased speeds. And very users will notice anyway.
 

Topweasel

Diamond Member
Oct 19, 2000
5,437
1,659
136
EDIT: Btw, what you call small is a reduction in memory speed by 266MHz. Now, according to my math on a 2666MHz rating memory kit, that is a 10% reduction in speed. That's not what I would call small or insignificant myself.
No one really responded to expect for the call out bit. But people place waaaay to much on the numbers of a particular hardware piece, or even the system as whole, 10% this 50% that. Rather than the actual performance of a system. The key is the actual affect on system performance in the tools, programs, and games that you will use. So a 10% drop in memory performance might come off as a large drop it isn't. Look at it like MPG, is 90 MPG that much different than 100 MPG? If you look at Fuel expense over the course of like 20000 miles at $3 a gallon, you get a difference of $66 a year. Even if we were looking 22.5 versus 25 MPG its $240 for a year or an extra $20 a month in gas. Cept we aren't looking at one stat to gather the performance. You have dozens of stats all with varying levels of importance depending on the task that will affect your overall system performance in a given application. That 10% drop might only equal a 2% performance drop in even the most memory intensive tasks. Just look at Meltdown patching, the potential for performance hit to a tool could be as high as 30% but on average the hit is only about 1-3%.
 
  • Like
Reactions: whm1974

whm1974

Diamond Member
Jul 24, 2016
9,436
1,568
126
No one really responded to expect for the call out bit. But people place waaaay to much on the numbers of a particular hardware piece, or even the system as whole, 10% this 50% that. Rather than the actual performance of a system. The key is the actual affect on system performance in the tools, programs, and games that you will use. So a 10% drop in memory performance might come off as a large drop it isn't. Look at it like MPG, is 90 MPG that much different than 100 MPG? If you look at Fuel expense over the course of like 20000 miles at $3 a gallon, you get a difference of $66 a year. Even if we were looking 22.5 versus 25 MPG its $240 for a year or an extra $20 a month in gas. Cept we aren't looking at one stat to gather the performance. You have dozens of stats all with varying levels of importance depending on the task that will affect your overall system performance in a given application. That 10% drop might only equal a 2% performance drop in even the most memory intensive tasks. Just look at Meltdown patching, the potential for performance hit to a tool could be as high as 30% but on average the hit is only about 1-3%.
Which is why I have been explaining to the OP that he is making mountains out of small molehills about the tiny loss in performance with all slots filled. This is something very users, even high end ones, will even notice.
 

TheDarkKnight

Senior member
Jan 20, 2011
321
4
81
Which is why I have been explaining to the OP that he is making mountains out of small molehills about the tiny loss in performance with all slots filled. This is something very users, even high end ones, will even notice.

Did you miss seeing the image of the paper insert I posted above? You have the very real "potential" to go from running memory @2666MHz using 1 stick of memory down to a low 1866MHz running 4 sticks of memory. Benchmarks on gaming website prove there is a very real difference, all the more relevant, on AMDs latest Ryzen CPUs with integrated Vega graphics. Maybe you need to read some reviews on these new CPUs and the real impact the memory speeds can have on performance. It's not negligible at all. I can post links if you like but that might invalidate your point. Id be happy to though.

@Topweasel, obviously I'm talking about games here. Not sending out an email. I just assumed we all knew that but I apologize. So, there is no depending on the application in this thread. It's all about the games. A little math for you here: 1866/2666 = ~.70. So were talking about a 30% drop in performance if you aren't aware of the "subtleties" involved in these scenarios, i.e., "Did you know if you use a single slot it has to go into A2?" "Did you know if you use up all 4 slots on your motherboard, there is a potential 30% drop in performance?", "Did you know that single-rank versus dual-rank matters?" I just learned all these small but important distinctions most recently. Yes, I'm sorry, it matters and it's noticeable.
 
Last edited:

whm1974

Diamond Member
Jul 24, 2016
9,436
1,568
126
Did you miss seeing the image of the paper insert I posted above? You have the very real "potential" to go from running memory @2666MHz using 1 stick of memory down to a low 1866MHz running 4 sticks of memory. Benchmarks on gaming website prove there is a very real difference, all the more relevant, on AMDs latest Ryzen CPUs with integrated Vega graphics. Maybe you need to read some reviews on these new CPUs and the real impact the memory speeds can have on performance. It's not negligible at all. I can post links if you like but that might invalidate your point. Id be happy to though.

@Topweasel, obviously I'm talking about games here. Not sending out an email. I just assumed we all knew that but I apologize. So, there is no depending on the application in this thread. It's all about the games. A little math for you here: 1866/2666 = ~.70. So were talking about a 30% drop in performance if you aren't aware of the "subtleties" involved in these scenarios, i.e., "Did you know if you use a single slot it has to go into A2?" "Did you know if you use up all 4 slots on your motherboard, there is a potential 30% drop in performance?", "Did you know that single-rank versus dual-rank matters?" I just learned all these small but important distinctions most recently. Yes, I'm sorry, it matters and it's noticeable.
You are still making a bigger deal out of this then warranted. Just use two 8GB sticks of memory in dual channel and be done with it. Although to honest, with 16GB of memory you would been better getting the 2400G instead of the 2400G, but by the time you will need 32GB of RAM chances are you need more then 4 cores anyway.

I don't think you are very likely to to fill four slots up anyway. So why the tempest in a teapot?
 

alexruiz

Platinum Member
Sep 21, 2001
2,836
556
126
Did you miss seeing the image of the paper insert I posted above? You have the very real "potential" to go from running memory @2666MHz using 1 stick of memory down to a low 1866MHz running 4 sticks of memory. Benchmarks on gaming website prove there is a very real difference, all the more relevant, on AMDs latest Ryzen CPUs with integrated Vega graphics. Maybe you need to read some reviews on these new CPUs and the real impact the memory speeds can have on performance. It's not negligible at all. I can post links if you like but that might invalidate your point. Id be happy to though.

@Topweasel, obviously I'm talking about games here. Not sending out an email. I just assumed we all knew that but I apologize. So, there is no depending on the application in this thread. It's all about the games. A little math for you here: 1866/2666 = ~.70. So were talking about a 30% drop in performance if you aren't aware of the "subtleties" involved in these scenarios, i.e., "Did you know if you use a single slot it has to go into A2?" "Did you know if you use up all 4 slots on your motherboard, there is a potential 30% drop in performance?", "Did you know that single-rank versus dual-rank matters?" I just learned all these small but important distinctions most recently. Yes, I'm sorry, it matters and it's noticeable.

We have been trying to help you, but you need to chill down and understand what the situation entails.
All AM4 boards have that sheet showing the "memory speeds", but that sheet was created over a year ago, BEFORE the launch of Ryzen. The sheet was intended to show what was guaranteed as far a memory speed, and that anything else beyond that, you were lucky.

At Ryzen launch, one year ago, clocking memory higher than 2133 had varied degrees of success. The sheet was intended to protect the board makers as far as guaranteed ram speed. It served its purpose, but what it meant was minimum speed, not maximum. Keep that in mind.

Finally, as others have said, 2 dimms will always be better than 4.
In any modern CPU, 4 dimms drive higher load and strain on the CPU IMC. For the same ram capacity, 2 dimms win.
 
  • Like
Reactions: whm1974