• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

does ram speed even matter? LTT vid

I "downgraded" from 16GB of 3600 DDR4 RAM to 32GB of 3200 DDR RAM with my Ryzen 5800x and the only performance difference I noticed was being able to run twice as much stuff in RAM as before with no slowdowns.

On a more serious note as that isn't exactly apples to oranges, RAM speed or throughput is somewhat of a useless metric for the vast majority of computer users including power users. There is very little software outside of certain professional or scientific workloads which can make use of the throughput we currently have. It's the reason practically no one sees any difference outside of benchmarks and typically only synthetic benchmarks. If the software is incapable of making use of the throughput there won't be any real world differences. It's not a difficult idea to grasp but most people don't understand how few pieces of software make use of high throughput.
 
I turned off XMP on both my rigs to drop the voltage from 1.4v to 1.1v for less heat and less wear n tear on my kits since the price of my 64GB kits have shot up to over a grand now.

Literally absolutely no difference at 4k. Exactly the same fps exactly the same gaming experience. No way any human being can tell in a random test with XMP on or off. The difference is indiscernible lol.
 
Last edited:
El_Capitan said:
This is like 4 and a half year-old news at this point, lol. But some people just need constant reminders.
Click to expand...
About time they made this video, back when I was buying RAM the videos out there were a few years old. I wasn't sure how much the meta had changed, I'd see online that 6000 MT/s is best with the new AMD chips and for X3D it doesn't really matter what speed you get. Other videos were over DDR4 and DDR3.

I knew it was basically pointless to buy the fastest RAM out there, but I wasn't sure if there was a decent benefit to hitting 6000MT/s vs not. I considered getting a couple more FPS to hit a constant 60 is well worth the price back then

I did notice a huge difference when I was training an autoencoder in MATLAB, when I enabled XMP on my DDR4 build going from 2133 to 3200 (50% increase in speed) I saw a 33% decrease in training time. Made a bigger impact than my 5070 GPU did, before it was barely running as fast as my laptop.
 
I went from 6400 to 6000mhz when I upgrade to 64gb because it couldn't hit 6400 with 4 sticks. Didn't make a difference I could see.
 
I went from 16gb cl18 ddr4 to 32gb b-die, didn't notice anything. But then again I don't OC and no gaming. I'm about to sell this b-die kit and put in generic OEM.🤔
 
Latency often matters more than speed in games. 6400 CL32 and 6000 CL30 is the same latency. The 32 and 30 are measured in clock cycles. Then a small difference doesn't matter much. You can see a noticable difference between XMP/EXPO and JEDEC, but even then it's usually just a few percent in terms of FPS. I buy XMP/EXPO ram that doesn't cost much more than the basic JEDEC speed stuff and doesn't run at some crazy high voltage. Yeah you want max speed if you want to win at benchmarks. It's not worth paying for if you're playing single player AAA games with the graphics cranked all the way up like I do. Then if you are going PvP I'd get an X3D before even thinking about fancy high end ram.
 
El_Capitan said:
This is like 4 and a half year-old news at this point, lol. But some people just need constant reminders.
Click to expand...
Or that single channel is in majority of cases barely slower in practice than dual channel. Or that you, in fact, can mix different sticks without issues more often than not.
 
El_Capitan said:
This is like 4 and a half year-old news at this point, lol. But some people just need constant reminders.
Click to expand...

Channels like Linus and Jays (especially Jays) have faced a real issue for the past 6 years or so...there are only so many things they can do with domestic PC tech. I bet Jay has a list of 30 topics that he just goes round and round on. "I'm switching to AMD!" a year later "I'm switching to Intel!" "You dont need these apps!" "You need these apps!" How big a PSU do you need?" "What speed ram?" and so on.
 
Last edited:
daglesj said:
Channels like Linus and Jays (especially Jays) have faced a real issue for the past 6 years or so...there are only so many things they can do with domestic PC tech. I bet Jay has a list of 30 topics that he just goes round and round on. "I'm switching to AMD!" a year later "I'm switching to Intel!" "You dont need these apps!" "You need these apps!" How big a PSU do you need?" "What speed ram?" and so on.
Click to expand...
This...

I mean there are new people who get into the hobby / field, but for those of us who have been around, most of those tuber's is just rinse, repeat with the newer toys and the end is often the same every time.
 
pendragon1 said:
apparently not that much. and unless youre playing benchmarks you probably wont notice the difference. or if youre making money then maybe go up in speed. for most people the difference isnt gonna matter.


View: https://www.youtube.com/watch?v=05HRLuU--jI
Click to expand...

In theory, he’s wrong; in practice, he’s right—in fact, there are applications that take advantage of, let say, faster* RAM, while others do not.
So he can prove that it doesn’t matter, and then in the next video prove that it matters a huge amount.
Linus was just putting on a show for his audience.
The truth is, RAM speed does matter.

*"Faster" depends on certain speeds/timings/latency/combinations from all them, etc.
 
AVATARAT said:
The truth is, RAM speed does matter.
Click to expand...
It does, but for gaming the "main" rule still applies most of the time - get a better vid card before you go nuts on ram. The caveat there is that "mainstream" gaming ram doesn't cost much more than no-heatsink JEDEC stuff, so you might as well get it. Or at least that's how it was before prices went nuts. 6400 CL32, 6000 CL30 - not a lot more $. Going from JEDEC to basic/standard/mainstream gaming ram with an XMP or EXPO profile is the biggest jump. So IMHO it's not worth blowing a big pile of cash on the fastest gaming ram unless you have a top end rig or maybe you're really hardcore into some e-sports game that doesn't need a lot of GPU but it is worth getting "standard" gaming ram.

One thing I noticed from that video is the 6000 CL26 won a lot of the time. So latency, not throughput. That doesn't surprise me at all with games. The main CPU threads in games are branchy. They're running a bunch of often complicated logic, not streaming assets from main memory and doing the same little bit over and over to them. That's the vid card's job. It's why the big cache on the X3D chips helps so much, and why lower latency helps on Intel procs too. CL is measured in clocks so it's relative to the speed of the ram. "Standard" DDR5 gaming ram is 200:1, so 6000 CL30, 6400 CL32, etc. JEDEC stuff is way worse, or you can pay up for a little bit better. It's the "little bit better" that's generally not worth it unless you're already balls to the wall with a high end proc and high end vid card, with perhaps a few exceptions for e-sports.
 
Last edited:
Gaming general priorities:
  1. Graphics card - Get the fastest you can afford. If gaming at 4K, then more than 12GB VRAM. Overclock and/or keep it cool.
  2. CPU - Speed over threads. If gaming at 1080P, go for 3D V-cache. If at 1440P or 4K, whatever's cheapest and has the fastest boost clock. The 9600X (5.4GHz) will beat the 13600K at stock (5.1GHz), but if you're getting a 360 AIO, there's plenty of meat on that bone. Overclock it, and baby, you've got a stew going (up to 5.6GHz up to 5.7GHz easily at 1.35V).
  3. Monitor - All that FPS is worthless if your monitor can't match it.
  4. RAM - Doesn't really matter as much. Overclock your graphics card first, then your CPU, then your RAM. For RAM, tighter timings > speed, which is why you don't see much of a gap between DDR5 7000MHz vs DDR4 3200MHz. If DDR5, SK Hynix A-die or M-die, and if DDR4, Samsung B-die for overclocking. Focus on CL vs Speed. For example, 2x16GB DDR5 6000MT/s CL36 vs 6400MT/s CL 36. Most of the time, the 6400MT/s kit will cost more. However, you can easily overclock the 6000MT/s CL36 to 6400MT/s with CL36 timings most of the time. Command Rate at 1T is better than CL (to a degree), and CL is better than everything else.
All that being said, overclocking is free, but not without potential consequences. Once you understand and gain experience, there's little to no pitfalls. For example, you should know by now not to push 3D V-cache CPU's, or 13th and 14th gen CPU's too hard with Vcore. Same thing with DRAM voltage and graphics card voltage.

The whole point of this thread, hopefully, is to spend the $1K on a RTX 5070 Ti, rather than on a 2x32GB DDR5 6400MT/s RAM kit, if you're upgrading from a RTX 2060 with an AMD Ryzen 5 9600X with 2x8GB DDR5 5200MT/s.
 
Yes, there are many variables. XMP/Expo profiles sometimes have no effect because, when switching to higher MT/s and higher CL, this simply increases the data transfer rate, which rarely helps with gaming, since reducing latency is more important for games. That’s why manual tuning is necessary, or the motherboard needs to have custom profiles for that specific RAM kit.

And if we look beyond the frame rate, RAM speed improves the performance of the entire system, not just the FPS.
 
AVATARAT said:
Yes, there are many variables. XMP/Expo profiles sometimes have no effect because, when switching to higher MT/s and higher CL, this simply increases the data transfer rate, which rarely helps with gaming, since reducing latency is more important for games. That’s why manual tuning is necessary, or the motherboard needs to have custom profiles for that specific RAM kit.

And if we look beyond the frame rate, RAM speed improves the performance of the entire system, not just the FPS.
Click to expand...
Turning on XMP or EXPO pretty much always improves latency. Yeah it's possible to get no benefit. Some apps just won't, and shitty XMP/EXPO ram exists. There are multiple speed bins for JEDEC ram, but the fastest DDR5 ones are 5600 CL40, 6400 CL52 (CUDIMM) and 8800 CL62. Lower end DDR5 gaming ram is something like 6000 CL36, and 6000 CL30 and 6400 CL32 really aren't premium. Then looking at high end stuff like CUDIMMs for Arrow Lake builds and the same 200:1 ratio is common. So stuff like 8800 CL44. CL is measured in clocks, so all the 200:1 modules are about the same latency regardless of transfer rate. It's the same for DDR4.
 
You must log in or register to reply here.
Back
Top