[SOLVED] Should I sell my RAM to buy Micron Rev E for Ryzen 5000?

Status
Not open for further replies.

JaSoN_cRuZe

Honorable
Mar 5, 2017
457
41
10,890
I have bought G.Skill 3600Cl19 2*8GB Kit for nearly 100 dollars when RAM prices started to peak in the hope of setting up my build for Ryzen 5000 series.

Unfortunately for me, the RAM prices have further fallen and I'm thinking of selling this RAM to get Crucial Ballistix 3000 cl15 2*8GB for a price of 65 dollars as these sticks were not available during the time of my initial purchase or was heavily overpriced.

Seeing that Micron Rev E are confirmed to be in the Crucial sticks compared to the luck of the draw in the G.Skill one,
should I sell my G.Skill RAM and get the Micron Rev E sticks for the Ryzen 5000 series build, what do you guys think.
 
Solution
Keep what you have, why would you want to swap for slower RAM ?
3000mhz CL15 has a true latency of 10nS. 3600mhz CL19 has a true latency of 10.5nS. The 3000mhz CL15 kit would be faster, however, the timings on a CL15 kit are not going to be conducive to good compatibility with Ryzen platforms. You would be a lot better off to find a set of high end sticks from G.Skill, or whoever you prefer if you don't like G.Skill which it seems maybe you don't, that are either CL14 3200mhz or CL16 3600mhz, because the 3200mhz CL14 sticks would assuredly be compatible with your Ryzen platform for 99% of 3200mhz CL14 kits, since those are going to be Samsung B-die chips and would have an 8.75nS true latency, or a 3600mhz CL16 that would be only...
It's a lot of time and activity for very little difference, if any, depending on what your pair it with. Especially since we haven't even seen any reviews yet and have zero idea what it is or is not going to like or what it's limits are going to be. I would recommend not doing ANYTHING at all until reviews are out and you have a chance to read them, so you can make an INFORMED decision.
 
  • Like
Reactions: DMAN999

JaSoN_cRuZe

Honorable
Mar 5, 2017
457
41
10,890
It's a lot of time and activity for very little difference, if any, depending on what your pair it with. Especially since we haven't even seen any reviews yet and have zero idea what it is or is not going to like or what it's limits are going to be. I would recommend not doing ANYTHING at all until reviews are out and you have a chance to read them, so you can make an INFORMED decision.

Thanks for the input @Darkbreeze .

Just to understand, if I get a Ryzen 7 3700X as my processor then which would be the better pick?
 
Keep what you have, why would you want to swap for slower RAM ?
3000mhz CL15 has a true latency of 10nS. 3600mhz CL19 has a true latency of 10.5nS. The 3000mhz CL15 kit would be faster, however, the timings on a CL15 kit are not going to be conducive to good compatibility with Ryzen platforms. You would be a lot better off to find a set of high end sticks from G.Skill, or whoever you prefer if you don't like G.Skill which it seems maybe you don't, that are either CL14 3200mhz or CL16 3600mhz, because the 3200mhz CL14 sticks would assuredly be compatible with your Ryzen platform for 99% of 3200mhz CL14 kits, since those are going to be Samsung B-die chips and would have an 8.75nS true latency, or a 3600mhz CL16 that would be only marginally slower at 8.8nS true latency.

"Speed/frequency", contrary to what a lot of people will tell you, does not "trump" true latency. True latency IS the measuring stick for how fast memory operations will generally be based on a COMBINATION of speed and CS latency. So sticks with a lower true latency are going to be faster, period. 4000mhz sticks with a CL22 latency are not going to be "faster" than 3200mhz CL14 sticks because the 4000mhz sticks are going to have a true latency of about 11nS.

ALSO, on Ryzen platforms, so far at least, once you go beyond 3600mhz you are almost certainly (To the point of being guaranteed for most configurations, although some systems with a very good CPU sample, high end motherboard and superior memory modules might be able to go as high as 3733 and in very rare cases possibly even 3800mhz although I haven't seen anybody actually accomplish this and show proof of it. It's been entirely anecdotal from what I've seen.) going to lose the benefit of a 1:1 ratio state for the infinity fabric as for MOST configurations, going beyond 3600mhz and certainly for the vast majority of users, anything over 3733mhz is going to incur a 1:2 ratio penalty. That is going to SIGNIFICANTLY increase the latency along the lines of 5-9nS in most cases depending on what the actual frequency is set to for the memory.

If you buy a very high end 2 x8GB 3733mhz CL16 kit, you'd end up with an 8.57nS true latency which, short of trying to further tighten those timings down, which might be possible, you'd end up with about the fastest memory configuration you could legitimately have without incurring the ratio penalty from the infinity fabric, IF you can get them to run at 3733mhz without decoupling.
 
  • Like
Reactions: JaSoN_cRuZe
Solution

JaSoN_cRuZe

Honorable
Mar 5, 2017
457
41
10,890
3000mhz CL15 has a true latency of 10nS. 3600mhz CL19 has a true latency of 10.5nS. The 3000mhz CL15 kit would be faster, however, the timings on a CL15 kit are not going to be conducive to good compatibility with Ryzen platforms. You would be a lot better off to find a set of high end sticks from G.Skill, or whoever you prefer if you don't like G.Skill which it seems maybe you don't, that are either CL14 3200mhz or CL16 3600mhz, because the 3200mhz CL14 sticks would assuredly be compatible with your Ryzen platform for 99% of 3200mhz CL14 kits, since those are going to be Samsung B-die chips and would have an 8.75nS true latency, or a 3600mhz CL16 that would be only marginally slower at 8.8nS true latency.

"Speed/frequency", contrary to what a lot of people will tell you, does not "trump" true latency. True latency IS the measuring stick for how fast memory operations will generally be based on a COMBINATION of speed and CS latency. So sticks with a lower true latency are going to be faster, period. 4000mhz sticks with a CL22 latency are not going to be "faster" than 3200mhz CL14 sticks because the 4000mhz sticks are going to have a true latency of about 11nS.

ALSO, on Ryzen platforms, so far at least, once you go beyond 3600mhz you are almost certainly (To the point of being guaranteed for most configurations, although some systems with a very good CPU sample, high end motherboard and superior memory modules might be able to go as high as 3733 and in very rare cases possibly even 3800mhz although I haven't seen anybody actually accomplish this and show proof of it. It's been entirely anecdotal from what I've seen.) going to lose the benefit of a 1:1 ratio state for the infinity fabric as for MOST configurations, going beyond 3600mhz and certainly for the vast majority of users, anything over 3733mhz is going to incur a 1:2 ratio penalty. That is going to SIGNIFICANTLY increase the latency along the lines of 5-9nS in most cases depending on what the actual frequency is set to for the memory.

If you buy a very high end 2 x8GB 3733mhz CL16 kit, you'd end up with an 8.57nS true latency which, short of trying to further tighten those timings down, which might be possible, you'd end up with about the fastest memory configuration you could legitimately have without incurring the ratio penalty from the infinity fabric, IF you can get them to run at 3733mhz without decoupling.
I totally agree with the true latency and Ryzen would detach 1:1 ratio of IF clock is set above 3600.

I was said that the overclocking potential is rather high for Rev E compared to Hynix. Using the Ryzen dram calculator you can find a fast profile of rev e running @cl15 @3733 whereas Hynix on best would run @cl16 for the same speed.

If the rumor that the IF clock for the upcoming Ryzen 5000 can maintain 2000 then the Rev e can do it with looser timings whereas Hynix may not post at that speed.

I may be wrong but the true latency can be reduced by tweaking the sub timings and it is said that Rev E is superior in this aspect hence I'm confused. I'm definitely gonna tweak the subtimings to reduce latency and won't stick with the default xmp profiles.

Also these crucial kits are on offer with the price of 65 dollars usually priced 100 and if we are lucky 75. Hardwares are expensive in India. Samsung b die are heavily over priced 200+ dollars for any of those kit and they are out of stock for the majority of the time.
 
In this regard, if you have to BUY a set of sticks, I'd have to agree with Count Mike that buying into a 3000mhz kit for Ryzen doesn't make a lot of sense. Not when the prices are not especially higher for much faster kits these days. But obviously that is a choice you have to make for yourself.
 

JaSoN_cRuZe

Honorable
Mar 5, 2017
457
41
10,890
In this regard, if you have to BUY a set of sticks, I'd have to agree with Count Mike that buying into a 3000mhz kit for Ryzen doesn't make a lot of sense. Not when the prices are not especially higher for much faster kits these days. But obviously that is a choice you have to make for yourself.

I have decided to buy those crucial sticks in discount check with the combination of 5600X with B550 Aorus Pro. Based on the overclocking capabilities and latency results from both will either sell the G.Skill or the Crucial. If it's G.Skill, I would be at a loss but if I wish to sell crucial, I would probably make the spent amount in re-selling it. Yes RAM prices are low but the availability of RAM in my region is very bad. It is either very high or very low tier, The budget B-die chips like Flare X were last seen 2017. We have to make it work with what we have here.
 

JaSoN_cRuZe

Honorable
Mar 5, 2017
457
41
10,890
UPDATE
I have decided to return the Micron Rev E Kit. As buildzoid stated that Ryzen 5000 won't go past 1900 FCLK and still has the same limitations of the 3000 series processors in terms of memory overclocking, should have guessed it as the IO die is the same on both.
Anyways at 3800 Hynix CJR Kit or the Rev E are more or less the same.
 
Status
Not open for further replies.