Change AMD 5900 for intel? Advice please

[markturner1960]

Hi, I just ordered a new PC and its arrived, initial impressions are that despite the step up from a 2080ti to a 3090 card, and an extra 32 GB of RAM, my old PC ( 9900K) is running DCS faster than the new one ( 5900) .....

 

This was admittedly a quick side by side, using free flight mission. Pretty sure I had all settings the same as well. The old PC was approx 80 FPS and the new one around 65 FPS. On using 3dmark, the new PC was approx 20% higher scores across the board. But 3Dmark is not DCS....

 

Its got me thinking perhaps I made a mistake going for the AMD processor? I want to use it for VR in DCS a lot, so the whole idea was getting the fastest chip. I do have the option to send back and get the parts changed over. What do you think? And what processor would you recommend to run DCS the fastest? 

 

thanks

[Rudel_chw]

22 minutes ago, markturner1960 said:

Its got me thinking perhaps I made a mistake going for the AMD processor?

the AMD chip should be 20% faster on DCS, judging from its single-core performance ... when you did your test, both PC used the same DCS version?

[Lt_Jaeger]

Look for ctr 2.1. My 5900 with 3080ti is a beast. 

[Hoirtel]

Yeah single core performance is what you need to  check. You should run HWINFO and see if there are any other bottlenecks that might be reducing perf outside of CPU and GPU. First thing I think of is RAM and its speed. I'm not very experienced with this on AMD but the effective speed this runs at is all controlled by the CPU. 

[markturner1960]

Hi, thanks, yes pretty sure it was the same version and same settings. I have not had a lot of time to look into this deeply, so will investigate the CTR software and do some further checks.  I am glad that in theory at least, my choice should be better for running DCS....even better than say an 11900K ? 

[Hoirtel]

6 minutes ago, markturner1960 said:

Hi, thanks, yes pretty sure it was the same version and same settings. I have not had a lot of time to look into this deeply, so will investigate the CTR software and do some further checks.  I am glad that in theory at least, my choice should be better for running DCS....even better than say an 11900K ? 

Based on single core performance it should be better, however its the sum of all the parts and setting them up that has the final outcome. Any PC can run DCS badly....

Edited August 10, 2021 by Hoirtel

[Lt_Jaeger]

Ctr will check the different cores for max clock speed. Since I'm a little on the loser side of the silicon lottery, I'm able to run only 4850 / 4900mhz on the first 2 cores. With the 3900x I had to make sure to keep the fast cores restricted to dcs, but currently I don't deem it necessary any more. 

 

For me, one of the benchmarks for dcs performance, is the Anton bomber intercept mission. With all these planes in the air, my 3900x / 2080ti went to stutterville, the current combo 5900x / 3080ti heats the world climate but is running kinda decently (all in vr). 

 

 

[Thinder]

2 hours ago, markturner1960 said:

Hi, I just ordered a new PC and its arrived, initial impressions are that despite the step up from a 2080ti to a 3090 card, and an extra 32 GB of RAM, my old PC ( 9900K) is running DCS faster than the new one ( 5900) .....

 

Right. You obviously haven't get the memo about the Ryzen CPUs and I'm not sure you researched this PC specs before purchase, in any case, check the BIOS setting, some manufacturers just don't bother optimizing their systems before shiping, same for the build looking at those specs. 2666MHz/Cl16? Seriously?

 

The Ryzen architecture is optimized for low latency, best results are obtained with Cl14/3200MHz, so what is your RAM latency? Also, CPU controllers are limited to a number of ranks not exceeding 4 (FOUR) in the case you only use RAM at the recommanded frequency which is 3200MHz, anything more than 4 ranks and your CPU will throttle back under load.

 

As it is, your CPU looks slower than my 5600X which can run all cores at 4506.97MHz in CPU-Z tests and would give me a 3.47% GPU bottleneck headroom if I was to swap my GTX 1080Ti for a 3080.

 

Since you use 3DMark (I use pro), here are the results of some test I ran back to back at 4 K after getting the RAM kit I use now.

 

(Crucial RAM first. G.GKILL second)

3D Mark Pro Fire Strike Ultra 4K gaming with DirectX 11

3840 x 2160 (16:9) (4K) MSAA x 2

MSI Afterburner ON. AMD Ryzen Master ON; Game mode.

>>>>>>

Graphics score: 6 496 vs 6 583. +1.33%
 
Physics score: 23 894 vs 25 339. + 6.04%

Combined score: 3 605 vs 3 654. + 1.35%.

>>>>>> 

 

This suggest to me that RAM is much more important when running Ryzen CPUs than people usually think, if you try to O.C yours with a liquid cooler, you won't get anywhere close to half the performance gain I got from using this RAM simply because the CPU doesn't throttle back and this affects all buses including CPU-to-GPU.

 

So check your RAM compatibility with your motherboard, bounding with your CPU (Cl 16? Number of Ranks?) and make sure they are exactly of the same chip batch first if the previous two are a pass.

 

If not, you just find the source of your problem, then consider using a Cl14/3200MHz/1 Rank kit for a 4X stick configuration, you don't need 64GB of RAM, 4 X 8GB would do and lower the price of the upgrade.

Edited August 10, 2021 by Thinder

[Hoirtel]

18 minutes ago, Thinder said:

 

Right. You obviously haven't get the memo about the Ryzen CPUs and I'm not sure you researched this PC specs before purchase, in any case, check the BIOS setting, some manufacturers just don't bother optimizing their systems before shiping, same for the build looking at those specs. 2666MHz/Cl16? Seriously?

 

The Ryzen architecture is optimized for low latency, best results are obtained with Cl14/3200MHz, so what is your RAM latency? Also, CPU controllers are limited to a number of ranks not exceeding 4 (FOUR) in the case you only use RAM at the recommanded frequency which is 3200MHz, anything more than 4 ranks and your CPU will throttle back under load.

 

As it is, your CPU looks slower than my 5600X which can run all cores at 4506.97MHz in CPU-Z tests and would give me a 3.47% GPU bottleneck headroom if I was to swap my GTX 1080Ti for a 3080.

 

Since you use 3DMark (I use pro), here are the results of some test I ran back to back at 4 K after getting the RAM kit I use now.

 

(Crucial RAM first. G.GKILL second)

3D Mark Pro Fire Strike Ultra 4K gaming with DirectX 11

3840 x 2160 (16:9) (4K) MSAA x 2

MSI Afterburner ON. AMD Ryzen Master ON; Game mode.

>>>>>>

Graphics score: 6 496 vs 6 583. +1.33%
 
Physics score: 23 894 vs 25 339. + 6.04%

Combined score: 3 605 vs 3 654. + 1.35%.

>>>>>> 

 

This suggest to me that RAM is much more important when running Ryzen CPUs than people usually think, if you try to O.C yours with a liquid cooler, you won't get anywhere close to half the performance gain I got from using this RAM simply because the CPU doesn't throttle back and this affects all buses including CPU-to-GPU.

 

So check your RAM compatibility with your motherboard, bounding with your CPU (Cl 16? Number of Ranks?) and make sure they are exactly of the same chip batch first if the previous two are a pass.

 

If not, you just find the source of your problem, then consider using a Cl14/3200MHz/1 Rank kit for a 4X stick configuration, you don't need 64GB of RAM, 4 X 8GB would do and lower the price of the upgrade.

 

I knew there was something a bit more involved when it came to AMD and RAM speeds, but not had any experience with it. Seems you have researched it very well!

[markturner1960]

Sorry, I did not update my signature completely. Ram is this: https://www.scan.co.uk/products/64gb-2x32gb-corsair-ddr4-vengeance-rgb-pro-black-pc4-28800-3600-non-ecc-unbuffered-cas-18-22-22-42-1

 

 

[Hiob]

Have you set the XMP-Profile in your Bios?

This step has to be done basically with every new build in order to boost performance.

The 5900 is plenty power to not be the bottleneck, even with a 3090. (and though CTR is beneficial - I use it myself - it is in now way needed to get the performance for DCS)

XMP is the only major "one-flip" bottleneck that comes to my mind.

 

EDIT: Next step would be checking all important drivers for latest updates (nvidia, bios, chipset)

Edited August 10, 2021 by Hiob

[markturner1960]

Thanks, some great advice here.......AassssssAlways...!

[Hiob]

aaaaaaand...

check for background processes/autostart apps.

Some software can be pretty intrusive from time to time.

[reece146]

Whatever the issue is that you are having it is not the CPU causing it.

 

This sounds like a system optimization issue (i.e. software and/or OS level optimization) and nothing to do with the hardware set itself.

 

Full disclosure: I have Intel 10700K, 11700k and 5600X CPUs running DCS and they are all within pissing match of one another - not really a big difference between them - at the seat of the pants  without measuring. Long story short - you should be happy with your config once it is optimized - no need to change anything.

Edited August 11, 2021 by reece146

[Thinder]

15 hours ago, Hoirtel said:

I knew there was something a bit more involved when it came to AMD and RAM speeds, but not had any experience with it. Seems you have researched it very well!

 

To be honnest, it took me month of request emails to AMD, MSI and later on G.Skill to figure this one out, I was first put on track by an advanced AMD user on the MSI forum, yes it is tedious and perhaps the reason why people don't bother looking further than increasing the number of sticks and/or frequency.

 

But this approach with these Ryzen is logical, they were conceived to take advantage of low latency RAM.

 

Personally I took a different approach than O.C, because I do not have the budget to replace RAM,  CPU or GPU and believe me, I did BBQed some in the past, I also saw how DELL shiped me the last PC I purchased in 2011, all BIOS settings were set to default, at least I know my PC upside down because I put it together myself and set it up.

 

So this solution looks the best in view of my budget and research of performances at 4K (VR), as it offers more than twice the gain I could expect with fitting a liquid cooler to my 5600X with little difference in cost when it comes to the Premium you pay for those kits, I'll go through the same process for a Nex-Gen future PC, in a few years time.

 

All I do now is to pass on the informations gathered and obtained from manufacturers but it wasn't obvious, hope it help anyway.

 

Samsung B-die memory

 

B-Die Finder  Find Samsung B-Die DDR 4 memory kits on Amazon, Newegg and many more.

 

I'm not sure how much you can rely on this but this is one of my CPU-Z tests results.

 

Edited August 11, 2021 by Thinder

[markturner1960]

I have heard back from Scan, they say that XMP was used, it was required to run the memory at the 3600 frequency. I am not with the PC until the weekend, so can’t check, however, they are a well respected and capable performance PC builder and I have no reason to doubt them or their abilities. Pretty sure all the drivers etc will be correct and up to date. 
 

Any other suggestions as to where to look for a cause of the discrepancy? I have only installed DCS, track ir, voice attack, steam and chrome, so there should be no issues with other apps etc

 

thanks

[reece146]

Ensure the OS and NVIDIA drivers are set to performance mode over power saving, game mode, etc., etc.

 

Revisit the DCS settings to ensure that everything matches to the old system. At a high level:

 

The following things severely impact the CPU:

• Visibility Range
• Shadows
• Terrain Object Shadows
• Trees Visibility

The following impact the GPU:

• Obviously resolution
• MSAA, SSAA, AAs in general.

The following impacts system/GPU ram

• Textures
• Terrain textures
• Preload radius

Your new machine should be able to match the old machine without breaking a sweat. But, to get your old machine to where it is now may have taken a while to get where it to where it is now if you are a tinkerer. Any DCS performance/visual mods ever installed for example? Taz's trees help FPS a touch for example.

 

FWIW, 3600MHz RAM should be a sweet spot for AMD (cf: Infinity fabric) but the CLs could be better. I'm not convinced that it is enough of a deficiency to chase that hardware though. The relative difference in FPS from expected to generated is larger than what I'd expect the CL rating to affect if you follow what I mean.

 

When you are sitting at the machine again, ensure that all the background crapware like virus scanners (anything more than Windows Defender is a waste of CPU/money/cycles), mobo vendor tool sets, sound card BS, joystick BS, social media BS, etc., etc. is not running. Check what's running and consuming CPU & memory via the task manager.

 

Is your DCS install via Steam? Can you disable Steam as a background process if that is the case? Dunno - I'm standalone.

 

Random thought - the CPU and/or GPU aren't thermal throttling, right?

 

HTH

 

Edited August 11, 2021 by reece146

[markturner1960]

Thanks, was thinking more OS and hardware rather than DCS...I ported over my saved games folder to the new installation, so my DCS in game settings should all be the same ( and indeed, on checking they are) 

[Hoirtel]

8 hours ago, markturner1960 said:

I have heard back from Scan, they say that XMP was used, it was required to run the memory at the 3600 frequency. I am not with the PC until the weekend, so can’t check, however, they are a well respected and capable performance PC builder and I have no reason to doubt them or their abilities. Pretty sure all the drivers etc will be correct and up to date. 
 

Any other suggestions as to where to look for a cause of the discrepancy? I have only installed DCS, track ir, voice attack, steam and chrome, so there should be no issues with other apps etc

 

thanks

There is a twitch Flight Sim streamer that has had a  really bad recent experience with scan. They had destroyed a mobo, graphics card and left the cooling loop loose so the thermals went crazy. Sounded a bit crazy, don't know any more details....

[markturner1960]

Well, this is my 4th 3XS system from them and I buy all my business PC's from them as well and they have have always been top notch. Of course, I imagine, like in any business, mistakes do get made occasionally, but I have never had any issues. Dont get me wrong, in 3D mark, the PC is everything you would expect compared to the old PC, like I said, approx15 -20% better across the board, so I am thinking its more some kind of DCS problem or a setting somewhere that affect DCS particularly......

[Hoirtel]

7 minutes ago, markturner1960 said:

Well, this is my 4th 3XS system from them and I buy all my business PC's from them as well and they have have always been top notch. Of course, I imagine, like in any business, mistakes do get made occasionally, but I have never had any issues. Dont get me wrong, in 3D mark, the PC is everything you would expect compared to the old PC, like I said, approx15 -20% better across the board, so I am thinking its more some kind of DCS problem or a setting somewhere that affect DCS particularly......

I've always thought they were good although I've only ever bought parts from them. Good to know you have had good experience with thier build side. Like I said I don't know what happened lots of sides that story probably. 

[Thinder]

On 8/11/2021 at 7:40 AM, markturner1960 said:

it was required to run the memory at the 3600 frequency. I am not with the PC until the weekend, so can’t check, however, they are a well respected and capable performance PC builder...
 

 

So is DELL, yet you end up with a PC running with a £359.99 Cl18 RAM kit above recommanded manufacturer (AMD) frequency (3200MHz) and which was shiped to you with the wrong BIOS settings, the only positive for your CPU with this RAM kit is that it can't be more than 4 ranks IF it is the two stick, but at 3600MHz, it's already throttled down by the controller...

 

Chances are they cram those Cl 18 kits into their bundles because they can't sell them otherwise, Corsair advertises a Cl 16 kit, not Cl 18 but I was right to recommand that you check your BIOS in the first place, people who build those bundles don't give a damned about performances, they do not optimize those PCs, they sell them, it's not a simple mistake.

 

 

That's precisely why I don't trust anyone but myself when it comes to mine, and you guys keep posting frequencies which basically mean nothing when running a Ryzen designed for low latency, chances are your RAM kit is slower than a 32GB Cl14 3200MHz kit even at a higher frequency and double the amount of RAM, that's not how those CPU works.

 

Now, run 3D Mark Pro Fire Strike Ultra 4K gaming with DirectX 11 3840 x 2160 (16:9) (4K) MSAA x 2 and see if you obtain the performances gain you can expect from your Ryzen 5900X, that would be 1% lower latency, 3% faster on single core... 

 

CPU-Z is free btw, it will tell you exactly what is going on during the test.

 

https://www.cpuid.com/softwares/cpu-z.html

 

Here are AMD specs and recommanded frequency for your CPU and it is UP TO 3200MHz, perhaps those trustworthy people can explain to you why they think they know better and fit a Cl18 kit above this frequency into your PC?

 

https://www.amd.com/en/products/cpu/amd-ryzen-9-5900x

 

 

 

Now, in the future, after this experience, you will remember that there is a difference between a consumer and a PC user, this industry sells to consumer stuff they don't need, like frequencies above the CPU manufacturers best tested speeds and there is only one reason for this: It makes them make money, those who had it with being consumers and decided to get the best of their PCs look at the specs of the parts first and bound their RAM to their CPUs.

 

At least, G.Skill are half-honnest when they say that they manufacture their 3600MHz + kits "for those who want them"...

[BitMaster]

I honestly don't think that any kit faster than 3200 is consequently slowing down the CPU. Your promoted solution is indeed great, no doubt, but it is not the only

config supported and giving great results.

 

As far as I can remember AMD mentioned 3600MHz as the sweetspot and some even accept 3733 to 3800MHz before the Infinity Fabric switches modes.

I do agree, this is then above the specs and outside warranty, just to not get nailed to the floor if I dont mention this. 

 

If it wasn't for the money, I would prefer 3600 over 3200 any day if the CL is not off. A 3600 CL16 has 8,88ns latency, just 0,13ns slower than 3200CL14 @ 8,75ns.

Balancing Bandwith gain vs. Latency increase vs. cost.  As both kits are about in the same ballpark price wise ( with true CL16 or CL14 settings, if you go i.e. CL16-18-18-42 the price drops significantly ), I'd prefer the higher bandwidth and take the little latency hit.

 

My personal rule of thumb, memory ./. bandwidth ~ < 1   64GB ./. ~55-60GB/sec ~1.1      If you have 128GB and DDR4-2133 in Dual Channel you can go and get a coffee if you shuffle lots of GB around as it takes roughly 5 sec to read that amount of RAM in Best Case Scenario, just to give an idea. With a 4, 6 or 8 channel config that goes waaay quicker, or boost up the MHz  .

 

Above all, I think in the future, with Win11 and new GPUs etc. the amount of VRAM is getting more important than before once we can load from storage directly into VRAM. Then it may also pay back if you can read fast, very fast or very very fast. 

 

The market is forced to change and we can call ourselves lucky if gamers with big fat x86 CPU's dont get left behind. It all moves towards SoC, ARM etc for the main market. How much 

care can we expect from Intel and AMD if their main focus shifts away towards SoC etc..

 

Apple has had such a strong impact with their M1 chip that MS is back into developing it's own ARM chip and I dont expect it to fail. Things are moving, Dell and HP have sold really a lot less Notebooks since this M1 came out, they are all forced to do something and Apple is about to launch the next iteration in 2 months already, taking the crown in HE Notebooks, likely.

 

If the money stops to come in, things get changed that were once before written in stone, and this M1 was the beginning of the end of x86 chips as we know them and it will come fast.

 

Once AMD and Intel have branched to issue ARM designs as well, how much R&D will go into our aging class of x86 CPU's ?  Most software these days is already ARM compatible due to the mobile market, there are gazillions of apps to get started with, this is not an empty AppStore scenario as in the early days.

...and that's why I will likely prefer a new Apple notebook over a CPU-Mobo-RAM upgrade but I have not decided yet which route I will go, maybe I will get my last x86 CPU, a 5900X..LoL

 

 

Sorry for hijacking this thread

[reece146]

RE: RAM - yes, agree, there is not much in it with respect to 3200CL14-14-14-3X and 3600CL-16-16-16-3X. At least for Intel. IIRC for AMD the 3600 has it's advantages since the stock setting Infinity Fabric runs at 1800... RAM should be in multiples of 1800MHz and if you overclock then the RAM should be in multiples of whatever the Inf Fabric is set to... 1900 Inf -> 3800 RAM for example. It's not something I've messed with - just regurgitation of what I've read online in the usual places.

 

I suppose there might be something to trying 3x1800 = 5400 MHz RAM. IIRC it exists but probably negligible difference when measured (if any) otherwise it would be more of a "thing". There's probably stuff out there discussing it...

 

FWIW, the AMD machine I have is 3600CL15-15-15-36 (CPU-z - prob really 3600CL16). I got lucky when I bought the machine as a pre-built at the height of the GPU-apocalypse - vendor put together good configuration in spite of zero details of what I was buying from their web site (Details were 5600X, 3070, 16GB RAM and a picture of the completed system). This is my first AMD since the '90s. Peppy box - but I do miss the iGPU for use as a capture device. Enough so that I may swap it with my 11700K Linux box just for that reason.

 

While we are hijacking...

 

I don't have a problem with the new ARM based stuff at all but we need a more general, high performance platform than what Apple is offering with the M1. Getting locked into the Apple MacOS ecosystem is not something I am interested in. I see that Windows 10/11 can run on the M1 now and there is work to get Linux working on it but without a more open BIOS/EEPROM layer... it's problematic at best. I haven't looked into the specifics of these OS installs but expect that it is still gatekeep'ed by Apple in some form. Having run Linux on older Macs - it sucks working around the proprietary-ness of Apple.

 

I have a few ARM SBCs running various flavours of Linux and they work well enough. But, for a power user they are not sufficient for every day use. I basically use them as over built embedded devices (because I'm too lazy to mess with Arduino)..

 

$0.02

[Thinder]

2 hours ago, BitMaster said:

I honestly don't think that any kit faster than 3200 is consequently slowing down the CPU. Your promoted solution is indeed great, no doubt, but it is not the only

config supported and giving great results.

 

As I keep saying, it all depends on how much your CPU controller can handle, how many stick you have, how many ranks per sticks, CPU load etc, the more of it, the deeper the issue and playing DCS comes with its range of issues, not heavy RAM use which makes faster frequencies and anything more than 32GB unnecessary, but heavier than average, heavy CPU load under some scenarios.

 

Under those conditions your CPU controller will struggle to cope with the amount of data it has to process especially at higher frequencies. Those are the limits.

 

The Ryzen RAM-to-CPU bound not only determine how fast your CPU can run, but also how much it will restrict the buses speed, reason why I also gained 1.33% in GPU performances with this simple RAM swap, none of the buses were hold back after that and the only remaining limitations I get are my GPU bottleneck and thermal limits.

 

MSI and AMD specified to me that the controller will throttle back above its limits and those are 3200MHz and 4 ranks, you might fit higher frequencies RAM and notice a small gain at moderate load but under heavy load (gaming) it will throttle back when a better bound will not.

 

So that leads me to think that if his Ryzen 9 5900X can give some good results at medium settings during bench tests, when it comes to DCS it's a completely different story, reason why I never test-bech a low-to-mid settings but 4K 3840 x 2160 (16:9) (4K) MSAA x 2, that's the closest one can get to DCS requierements with a benchmark.

 

What people want to do out of those specs is not the manufacturers problem but that of those who chose to push their CPU controller beyhond its limits because it is fashionable, more to it, I speak with experience, my PC had a RAM-to-CPU bottleneck resulting in a loss of 6.04% in CPU speed with a 2 X 2 ranks 3200MHz Cl16 kit.

 

Now I don't have this bottleneck, simply by swaping RAM from Cl16 to 14 with the same frequency and the same number of ranks. only the number of sticks changes and the quality of the chips (B.Die), and the fact that this kit can make better use of interleaving, It's not rocket science and everyone should comprehend this simple math.

 

So how much do you think his CPU is losing with Cl 18 RAM if my Cl16 was losing 6.04% to Cl14?

 

Quote

As far as I can remember AMD mentioned 3600MHz as the sweetspot and some even accept 3733 to 3800MHz before the Infinity Fabric switches modes.

I do agree, this is then above the specs and outside warranty, just to not get nailed to the floor if I dont mention this. 

 

That's the commercial thing, not the AMD enthusiast choice. Did you notice that RAM manufacturers rarely mention ther latency and numberr of ranks in their advertising? One of the reason is that people buy RAM like cookies and have no clue if it fits their CPU in the first place, it got nothing to do with performances.

 

Quote

if you go i.e. CL16-18-18-42 the price drops significantly ), I'd prefer the higher bandwidth and take the little latency hit.

 

 

Same here, good luck increasing your CPU performances by 6.04% this way: You need to limit the number of ranks per stick to one in a 4 X sticks configuration and 2 in a 2 X sticks configuration, in any case, at high load, the little frequency gain you made is nulified when the CPU throttle back, after that people will spash dosh in a liquid cooler to gain less than 50% of what gain one can get from this sort of RAM kits.

 

In his case, it's not Cl16, it's Cl18, if you look at 6.04% gain in performances for - Minus 2 Cl per stick you end up with a hit of above 12% under load, and again, I reiterate, it's UNDER LOAD not moderate load what you're doing spreasheets, it is what is relevant to gamers in general, not mainstream average PC user.

 

Quote

  If you have 128GB and DDR4-2133 in Dual Channel you can go and get a coffee if you shuffle lots of GB around as it takes roughly 5 sec to read that amount of RAM in Best Case Scenario, just to give an idea. With a 4, 6 or 8 channel config that goes waaay quicker, or boost up the MHz

 

That's providing he has a motherboard supporting more than two channels, not mentioning the RAM sticks themselves, the standard is 2, the real issue is the number of ranks per stick and a maximum of 4 ranks.

 

What they don't tell you if you don't ask: "and also officially supported memory speed by AMD decrease accodingly"

Edited August 14, 2021 by Thinder