G.SKILL Flare X5 DDR5-6000 CL32 (AMD) 2x 16 GB

[ir_cow]

Well, considering that this G.Skill runs at 6000 MHz 32-38-38-96-130 with Hynix chips, and my Corsair is 6000 MHz 36-36-36-76-112 with Samsung, I have to say that Hynix chips don't seem to be as good as the hype around them.

If your not overclocking or tuning beyond the XMP/EXPO profile it doesn't matter what you have. Hynix M die scales up further and tighter subtimings compared to samsung current B-die (DDR5) that is just a little behind. Micron isn't even in the same world right now.

This will probably change as micron and Samsung release new die revisions in 2023.

I've noticed Corsair tends to have tighter profile timings.

 

[Mussels]

My point is that dual-rank RAM might not necessarily result in more performance, as some boards can only do lower speeds with dual rank modules

If you view it as "I've got the maximum overclock this CPU's memory controller can handle" then it seems like a loss - but you also prevent yourself ever adding more RAM, since you add ranks regardless


I'll speak of DDR4 since i'm more familiar with it, but why fight for DDR4 3800/4000 in two ranks, when you can get four ranks at 3200 and have it work out of the box? (or 3600 with minimal effort)
(In DDR5 there are these same points of diminishing returns, i just don't know them off the top of my head)

Would you rather:
32GB of 3600 for $178

Or 32GB of 3600 for $155, condensed into two sticks, with the option of 64GB in the future

2x32 kits start at $320 but have no direct comparison at the store i'm using
If all you wanted was 3600 for that compatibility and ease of getting it working, it's downright a bad idea to get 2x8 paying more for less performance and less upgrade potential


The fastest 2x8 SR kits they have cost as much or more than 32GB DR, and outside of very small latency gains you're not going to get any benefits from doing so - and four of those 4000 sticks sure wont run any faster than four 3600 sticks

 

[AsRock]

Exact same memory i have, they have been flawless.
I went for Flare X5 because the original Flare X worked so well with Previous Zen processors.

Yeah same here, been running mine 30cl\ other wise default. Worked well changing to 6200 too although i dropped the CL back to 32.

 

[AusWolf]

If you view it as "I've got the maximum overclock this CPU's memory controller can handle" then it seems like a loss - but you also prevent yourself ever adding more RAM, since you add ranks regardless


I'll speak of DDR4 since i'm more familiar with it, but why fight for DDR4 3800/4000 in two ranks, when you can get four ranks at 3200 and have it work out of the box? (or 3600 with minimal effort)
(In DDR5 there are these same points of diminishing returns, i just don't know them off the top of my head)

Would you rather:
32GB of 3600 for $178
View attachment 277138



Or 32GB of 3600 for $155, condensed into two sticks, with the option of 64GB in the future
View attachment 277139


2x32 kits start at $320 but have no direct comparison at the store i'm using
If all you wanted was 3600 for that compatibility and ease of getting it working, it's downright a bad idea to get 2x8 paying more for less performance and less upgrade potential


The fastest 2x8 SR kits they have cost as much or more than 32GB DR, and outside of very small latency gains you're not going to get any benefits from doing so - and four of those 4000 sticks sure wont run any faster than four 3600 sticks
View attachment 277140

I agree for DDR4. Dual rank, 3200 MHz out of the box is my tried and tested approach as well.

DDR5 is still too new, and I'm sure we'll see lots of improvement in the future. In about 2-3 CPU generations, DDR5-6400 dual rank will probably be the standard.

 

[bl4C3y3]

Thanks for the review.
When 'memory training' is happening, is there a standard procedure ? Something that is visible, like multiple POST cycles or is it motherboard/RAM dependent ?
In other words, is there a way to know that memory training is happening after settings have been changed ... so you know something is actually happening while waiting (vs failure to boot).

 

[AusWolf]

Thanks for the review.
When 'memory training' is happening, is there a standard procedure ? Something that is visible, like multiple POST cycles or is it motherboard/RAM dependent ?
In other words, is there a way to know that memory training is happening after settings have been changed ... so you know something is actually happening while waiting (vs failure to boot).

Unfortunately, not really. But it shouldn't take longer than 1-2 minutes.

 

[ir_cow]

Thanks for the review.
When 'memory training' is happening, is there a standard procedure ? Something that is visible, like multiple POST cycles or is it motherboard/RAM dependent ?

On AMD (AM5) the debug will say Code 15 during* the memory training. After that there is a bunch of other codes that are related but they go really fast. Code 55 is generally when your stuck on memory during post. This is why I don't like motherboards with just debug LEDs for B650/X670. The ram led will blink when its stuck and will blink when its training.

 

[Hankieroseman]

Anyone tried using 4 of the 16gb ones to get 64GB total?

Using F5-6000J3238G32X2-TZ5NR (2x32GB) modules in ASUS ROG Crosshair X670E EXTREME and Ryzen 9 7950X with no problems running MSFS2020.

 

[AusWolf]

On AMD (AM5) the debug will say Code 15 during* the memory training. After that there is a bunch of other codes that are related but they go really fast. Code 55 is generally when your stuck on memory during post. This is why I don't like motherboards with just debug LEDs for B650/X670. The ram led will blink when its stuck and will blink when its training.

On my board (in my specs), you usually see both the RAM and CPU debug leds lit up if it's stuck, and only the RAM when it's training... usually.

 

[Hankieroseman]

On my board (in my specs), you usually see both the RAM and CPU debug leds lit up if it's stuck, and only the RAM when it's training... usually.

It was odd the first time and it seemed to take 5 minutes or longer. I set the memory timings for 5200 and it failed. It was a while before I finally got it to boot. But this cute display on this EXTREME board told you what was happening while I looked up error codes in the asus bible. Thou shalt NOT change the memory timings in the BIOS. Use EXPO II, AI Optimized and the BIOS will take care of the rest. Perfecto! 7950X is Rockin' 5200 Megahertz. Stable. So far on Win11 Pro and MSFS2020.

 

[noel_fs]

Honestly - 1000% plus a run of 2.5b y-cruncher. After that the only thing you have to worry about is DRAM temps. Because passing when you first boot is easier because its cold. Being on for 8 hours / in-use can change the outcome. It isn't the 8 hours part so much, but the fact its at normal operating temperature for a extended period of time. You don't have to use karhu memtest. I just found it errors out a lot quicker vs MemTestPro or TM5. AIDA64 and Memtest86 is useless for DDR5. Those two will "pass" and instantly fail those other 3 programs.


Yes. Not with these specific ones. The outcome is the same. It depends on the motherboard, CPU and which platform your on (Intel or AMD).


Basically. It doesn't improve memory reviews because its 100% GPU bound. When the frame rate is the same between the GPU reviews and this, I don't see a reason to include them. Not until I get a more powerful video card and it shows a real impact on perf.

I totally understand, still would like to see them tbh. I get it wold be another time sink for testing so a good compromise i think would be to get rid of the medium preset 1080p benchmarks (i know the point of using it, just dont find it very relevant currently).

Personally im just curious overall and overtime it could prove useful.


On a side note, are you using resizable bar? I see no mention of it so just want to make sure since i feel like its something quite relevant.

 

[Mussels]

I agree for DDR4. Dual rank, 3200 MHz out of the box is my tried and tested approach as well.

DDR5 is still too new, and I'm sure we'll see lots of improvement in the future. In about 2-3 CPU generations, DDR5-6400 dual rank will probably be the standard.

The newness of DDR5 is exactly why I'm pushing the lower clocked, more ranks/modules approach: Fighting for MHz is harder on the newer stuff

 

[AusWolf]

The newness of DDR5 is exactly why I'm pushing the lower clocked, more ranks/modules approach: Fighting for MHz is harder on the newer stuff

That works as well, I guess. I'm happy with 6000 MHz single rank for now.

 

[bl4C3y3]

This is why I don't like motherboards with just debug LEDs for B650/X670.

yes, both my Crosshair VI/VIII Hero (X370/X570) have this code-LED ... but the X670E Hero being more twice the price of the X570 Hero was not something I could justify, so I settled for something more 'humble', but still stylish discrete black

you usually see both the RAM and CPU debug leds lit up if it's stuck, and only the RAM when it's training

interesting, I'll keep an eye on them (we have the same debug LED system). Couple of days ago, playing to get 6400 or 6600, I had CPU+RAM (red+orange) and needed to clear CMOS each time ...

 

[ir_cow]

On a side note, are you using resizable bar? I see no mention of it so just want to make sure since i feel like its something quite relevant.

By default it is enabled for AM5. I'll have to check again for Intel Z690/Z790. I think it is as well. Can't remember off the top of my head.

 

[alexcheveau]

Anyone knows how many layers this RAM PCB have?

 

[ir_cow]

Anyone knows how many layers this RAM PCB have?

Nope, but is it important?

 

[DrCR]

Nope, but is it important?

Significantly if it’s over 9000, but there’s no way it could be that high.

 

[ir_cow]

Significantly if it’s over 9000, but there’s no way it could be that high.

Not with this kit...ever. Maybe 1 DIMM with half disabled on LN2.

 

[Dr. Dro]

Not with this kit...ever. Maybe 1 DIMM with half disabled on LN2.

I think it flew right past you here man

Anyone knows how many layers this RAM PCB have?

I do not believe it's relevant, but if you are concerned, the Flare series are G.SKILL's midrange AMD-oriented memory series. If you want a higher-end product for any reason, you should purchase a Trident Z5 kit, or one of Corsair's Dominator Platinum kits.

 

[ir_cow]

Never watched DBZ, makes sense now

 

[JChris]

Sorry for reviving this topic, but your review was very useful and I have a question: if you were to pick a RAM today, which one whould you choose between the options below?

Their price are very similar to me, I'm in Europe so it's kind of limited the beands I can get.

- Kingston FURY Beast DDR5 6000MHz 32GB 2x16GB CL36 (KF560C36BBEK2-32)
- G.Skill Flare X5 2x16GB 6000 MHz CL32 AMD EXPO (F5-6000J3238F16GX2-FX5)
- G.Skill Flare X5 Series Perfil Bajo 32GB (2x 16GB) DDR5 6000 MHz CL30 (F5-6000J3038F16GX2-FX5)
- G.Skill Trident Z5 Neo 32GB (2x16) DDR5 6000 CL30 (F5-6000J3038F16GX2-TZ5N)

They will be used in a X670E with R7 7800x3d.

 

[ir_cow]

@JChris AMD is quite latency sensitive and if you aren't going to do any additional adjustments besides just enabling XMP/EXPO, lower primary timings are always the "best" for AMD.

However that X3D CPU is already doing some heavy lifting so ultra low latency memory isn't as beneficial.

I have 3 of those kits mention. Buy what you can afford and don't overspend

If you are going to tune it up, I think G.Skill Neo will give the best results. Flare X5 did decent too, in the review. So who knows...

 

[JChris]

@JChris AMD is quite latency sensitive and if you aren't going to do any additional adjustments besides just enabling XMP/EXPO, lower primary timings are always the "best" for AMD.

However that X3D CPU is already doing some heavy lifting so ultra low latency memory isn't as beneficial.

I have 3 of those kits mention. Buy what you can afford and don't overspend

If you are going to tune it up, I think G.Skill Neo will give the best results. Flare X5 did decent too, in the review. So who knows...

I will just overclock to the maximum from the EXPO, nothing manual. Money is not an issue as all of these RAMs are well within my budget, the problem is that sometimes not all modules are available where I live (Europe).

I will check the Neo as well, but any of those will work great and I won't notice any meaningful difference, is that right?

 

[Mussels]

Sorry for reviving this topic, but your review was very useful and I have a question: if you were to pick a RAM today, which one whould you choose between the options below?

Their price are very similar to me, I'm in Europe so it's kind of limited the beands I can get.

- Kingston FURY Beast DDR5 6000MHz 32GB 2x16GB CL36 (KF560C36BBEK2-32)
- G.Skill Flare X5 2x16GB 6000 MHz CL32 AMD EXPO (F5-6000J3238F16GX2-FX5)
- G.Skill Flare X5 Series Perfil Bajo 32GB (2x 16GB) DDR5 6000 MHz CL30 (F5-6000J3038F16GX2-FX5)
- G.Skill Trident Z5 Neo 32GB (2x16) DDR5 6000 CL30 (F5-6000J3038F16GX2-TZ5N)

They will be used in a X670E with R7 7800x3d.

EXPO is usually better since they can use extra values that dont exist on XMP.
In order of importance to me:
EXPO
Timings
voltage (If it went from 1.2v at CL32 to 1.5v at CL30, I'd not bother with the CL30)


The differences between those would be small, but the CL36 kit is noticeably slower than the rest