• Welcome to Autism Forums, a friendly forum to discuss Aspergers Syndrome, Autism, High Functioning Autism and related conditions.

    Your voice is missing! You will need to register to get access to the following site features:
    • Reply to discussions and create your own threads.
    • Our modern chat room. No add-ons or extensions required, just login and start chatting!
    • Private Member only forums for more serious discussions that you may wish to not have guests or search engines access to.
    • Your very own blog. Write about anything you like on your own individual blog.

    We hope to see you as a part of our community soon! Please also check us out @ https://www.twitter.com/aspiescentral

DDR4 Versus DDR5

Judge

Well-Known Member
V.I.P Member
This is driving me nuts. Trying to find an online source in plain English that cites that MT/s equates to Mhz. And if not, is there a site that converts these two measurements in either direction? (Like equating Sones to Decibels in terms of noise.) I'm just looking for mathematical conversions, not complex technical explanations.

Seems a number of people are having real issues with overclocking, in not doing some homework over whether their CPU and motherboard can accept the right kind of memory and latency. Frustrating when sources like Intel cite only MT/s pertaining to both DDR4 and DDR5. Arrrrgh! I just want to verify that I have the right memory for the right CPU and motherboard.

That said, DDR5 continues to stump me. And yes, I've gone to "PartPicker" hoping to use it as a tool to make such determinations for me. And indeed it provides succinct answers. Conversely I've also read in Amazon comments about various motherboards supposed to work with various DDR5 memory where PartPicker claimed it would work, but that it didn't. Not sure if this involved operator error as few people seem to check the memory limitations of a CPU as opposed to the motherboard.

It just all seems terribly confusing compared to looking up latency requirements for DDR4 RAM.

Can anyone shine some light on this ? Or should I rely on sources like PartPicker as long as I correctly identify the right memory sticks for the right motherboard? (I've used it as a reliable source for years.) I have no issues in putting together a DDR4 system at 3200Mhz without any overclocking. But DDR5 requirements still seem confusing.

In essence, is Corsair Vengeance 32GB (2x16) DDR5 - 4800 MT/s CL40 memory optimal to run with an Intel i5 12400 CPU and an Asus TUF Gaming B760M-PLUS WiFi ?
 
Last edited:
It can get tricky, MB, CPU, and ram need to be matched or nothing works as it should. I normally check the specs of the processor first, then choose ram to suit that and search for a board compatible with both afterwards. It always pays to check back and forth both ways.

The CPU you've chosen is compatible with both DDR4 and DDR5 but the motherboard is not, it's DDR5 only.

You are better off with DDR5, the extra speed and extra bandwidth are worth it. I also noticed in the MB specs that 4800 MT/s is the slowest ram that board supports, you may or may not want to choose something a bit quicker, depending on budget constraints.

I'm also not familiar with this new MT/s term, I still think of it in Mhz and I think they're both the same.

In essence, is Corsair Vengeance 32GB (2x16) DDR5 - 4800 MT/s CL40 memory optimal to run with an Intel i5 12400 CPU and an Asus TUF Gaming B760M-PLUS WiFi ?
This appears to be a workable combination.

[Edit] You can't choose a faster set of ram because 4800 MT/s is the fastest that that CPU supports.
 
I'm also not familiar with this new MT/s term, I still think of it in Mhz and I think they're both the same.

This appears to be a workable combination.
I would tend to agree, but it bothers me that I can't seem to find a source to plainly state that MT/s mathematically equates to Mhz. And I'm basically relying on PartPicker to validate my choice of RAM. Choosing Corsair is a "no-brainer" IMO, but the latency issues still haunt me.

Apparently it's all quite confusing for a number of persons involved with motherboards that exclusively use DDR5 memory. Leaving me to seriously consider skipping on the ASUS DDR5 motherboard and going with an MSI DDR4 motherboard at 3200Mhz, sans any overclocking.

I want the increased speed of DDR5 without overclocking, but only if it truly gets along with the motherboard. If MT/s truly equates to Mhz, then the problem so many PC builders are having is that they are paying attention only to motherboard requirements rather than considering CPU requirements as well. But to date I haven't discovered any sources that clearly say MT/s=Mhz.
 
Last edited:
Leaving me to seriously consider skipping on the ASUS DDR5 motherboard and going with an MSI DDR4 motherboard at 3200Mhz, sans any overclocking.
This might not be a clever idea, thinking of future upgrades. We've been using DDR5 for a while now which usually means DDR4 won't be produced for much longer.

I also never overclock.

And I've been doing a bit of a search and checking specs, it looks like you won't find an intel processor that uses ram any quicker than 4800 MT/s unless you're prepared to spend a considerably larger sum of money than I would.

That bodes well for your choice of MB though, future upgrades compatible.
 
This might not be a clever idea, thinking of future upgrades. We've been using DDR5 for a while now which usually means DDR4 won't be produced for much longer.

I also never overclock.

And I've been doing a bit of a search and checking specs, it looks like you won't find an intel processor that uses ram any quicker than 4800 MT/s unless you're prepared to spend a considerably larger sum of money than I would.

That bodes well for your choice of MB though, future upgrades compatible.
It's proving to be a very skinny tightrope for me to walk. Choosing a midrange CPU that still operates at a lower wattage for most uses, especially considering only light gaming, if any. With no interest in XMP technology, let alone overclocking. Thermodynamics are always on my mind, along with noise from cooling fans. All major concerns if I choose to use my existing (and tiny) mATX computer case, the one I showed in early threads.

But above all, I want to make sure that whatever parts I purchase to build a new system, won't be the wrong ones. I'm fine with nominal speed and latency RAM, as long as they run properly in a particular motherboard with a particular CPU.

And some of these combinations of parts are starting to become scarce with sources like Amazon and Newegg. And I never purchase used computer parts. That's a deal-breaker for me.
 
Last edited:
According to everything I've read the combination you've picked out should be fine. I didn't use part picker, I went to the manufacturer's websites and check the specs from there.
 
According to everything I've read the combination you've picked out should be fine. I didn't use part picker, I went to the manufacturer's websites and check the specs from there.
Thanks for the feedback. I've considered both forms and the metrics they use...but I suppose it still spooks me to read a few people's comments that they took a similar route and yet failed. But then maybe they chose some other more obscure and cheaper RAM. From what I've seen in the past, it's hard to go wrong with Corsair memory any more than their renown power supplies.

But trying to understand latency with DDR5 continues to vex me. I want the most stable memory- not the fastest. But I also want to avoid the slowest memory, if and when possible. Above all, I gotta live up to my autism and keep overthinking it all. :rolleyes:

My sense of logic tells me that MT/s and Mhz are two different metrics. Yet it would then seem illogical that they can be quantitatively expressed as identical integers. Or am I missing something obvious here? :oops:
 
Last edited:
Just out of curiosity I did a quick search for computer shops in the US and noticed a profound difference between here and there.

Most of my search results came up with computer repair shops or places selling outdated or second hand parts, I'd never go near them. In Australia most of the shops are more aimed at the components for custom builds in a similar style to NewEgg. I wonder if this is a cultural difference or more related to our proximity to Asia.
 
On the MT/s vs. Mhz topic - I believe the comparison is like IPS (Instructions Per Second) vs. clock cycle on CPUs.

Modern CPUs can do all sorts of fancy techniques like instruction-level parallelism, pipelining, out-of-order execution, and superscalar execution that older CPUs can't. Which is why they are able to do many more operations today given the same clock speed - IPS is a way of measuring that.

However, IPS might not be accurate in real-world scenarios because of poorly written code and memory stalls. I have read that it is estimated that a program is "lucky" to use 20% of a CPU's capabilities. So this is where CPU speed is still useful, it gives another view of a CPU's capabilities.

From what I've read, RAM MT/s is like CPU IPS and RAM MHz is like CPU GHz, MT/s is measuring the rate of data transfer. Some RAM architectures can do more data transfer per clock cycle these days, hence the need for MT/s. So it is indeed getting more complicated to understand RAM performance, but Outdated has a good handle on recommendations.
 
And I've been doing a bit of a search and checking specs, it looks like you won't find an intel processor that uses ram any quicker than 4800 MT/s unless you're prepared to spend a considerably larger sum of money than I would.

That bodes well for your choice of MB though, future upgrades compatible.

Money isn't an issue for me. It's picking a CPU at a threshold where temperatures are still likely to be lower compared to any and all Intel CPUs beyond this model or generation. With much higher processor base power, translating into much more heat the minute the computer is turned on.

Processor Base Power - 65 Watts
Maximum Turbo Power - 117 Watts

Now if I were to consider abandoning my obsession with thermodynamics, I'd probably opt for a much more power CPU, though I'd be dealing with even more thermodynamic concerns broadening beyond a cpu and RAM, into a realm of GPU, case fans, creating positive airflow and having to find a very specific case that would likely sacrifice noise for airflow.

Mhz is looking better and better right now. Much easier to understand, though DDR4 motherboards are not in plentiful supply given DDR5.
 
Last edited:
@Judge

Mhz isn't a good measurement for "memory speed". It looks like it may have been chosen quite long ago because the industry had "trained" consumers" in what Mhz means for processors.

Note though - Mhz/Ghz isn't an ideal measurement for processor capability either. But as is often the case in IT, there's "good enough" and "takes too long to educate even the smartest consumer" (as opposed to serious hobbyists - they'll spend the time if they must).

What actually matters with memory is the effective data transfer rate. But memory these days is far from simple. The cycle time of the electronics seems to be getting less and less useful as a measure of transfer rate.

This has happened in the past a few times with CPU effectiveness (notably with IBM's POWER chips - there are probably articles on that on the web).

So it seems the industry is trying to switch to transfer rate instead of cycle time.

These "metric switches" typically happen at times when the old measurements have become misleading, and the new one is designed to capture the real situation more accurately and more consistently.
Which means it's probably the worst possible time to make a conversion scale :)

BTW - MT/s needs to be adjusted by the "channel width". If the memory H/W is moving multiple bits per transfer, you get the data rate by multiplying MT/s by channel width in bits.
Which suggests (but I didn't check) that memory tech is getting good at doing multi-bit transfers, and CPU I/O buses are able to handle that.

I remember this kind of thing happening with CPU pipelining (on-chip parallelization for early execution of instructions), and again with on-processor memory caching - both messed up the meaning of CPU metrics, but in those cases they kept using processor cycle time anyway.
That makes sense for CPUs (cores) BTW - the problem with cycle time for memory is that it's always been an indirect metric: what you want has always been the data rate.
 
Last edited:
All I am really looking for is whether or not the following combination of CPU, DDR5 memory and motherboards will work properly in building a new computer, without regards to overclocking the memory.

In building systems up to now I've never experienced such confusion about a fundamental component like memory and latency. Keeping in mind at all times I only want top grade, but also low profile memory so a beefy cpu heatsink can clear the memory on the motherboard, whether all four banks are filled or not.

Hence I felt forced to rely on "PartPicker" to match a specific CPU that emphasizes memory types up to DDR5 4800 MTS. But some folks apparently claimed PartPicker did them wrong. Whether this reflects reality or operator error I cannot say. Though it seems that both myself and @Outdated came to the same conclusion of my choice of memory, PartPicker withstanding.

Motherboards:

* MAG B760 TOMAHAWK WIFI

* TUF GAMING B760M-PLUS WIFI - Tech Specs|Motherboards|ASUS USA

While the CPU objectively offers capability only "up to DDR5 4800 MT/s", both motherboards offer capability of non-overclocked memory at 5600 MT/s. So I presume that in one's BIOS settings, 4800 is as high as it may show or go given the CPU limitation. Or is this an incorrect assumption?

I've seen a number of comments reflecting confusion over such concerns, not only with the limits of the CPU and motherboards, but how they impact overclocking as well. Leading people to purchase higher MT/s memory which may or may not work.

____________________________________________________________________________________________

PartPicker list of available DDR5 memory compatibility for the MSI MAG-B760 Tomahawk WiFi DDR5:


Logical choice of DDR5 memory based on the specified CPU: Corsair Vengeance 32GB (2X16) DDR5 4800 ?

____________________________________________________________________________________________

PartPicker list of available DDR5 memory capability for the Asus TUF Gaming B760M-Plus WiFi DDR5:


Logical choice of DDR5 memory based on the specified CPU: Corsair Vengeance 32GB (2X16) DDR5 4800 ?

____________________________________________________________________________________________

THE BOTTOM LINE: Do y'all concur with PartPicker's selection of compatible memory based on these two motherboards, and the 4800MT/s limitation of Intel's i5 12400 CPU? Does that CPU limitation in fact, truly dictate the proper amount of stable memory to be installed on a motherboard?

And that a motherboard's bios is only capable of reading the limit of memory based on the CPU, and not the motherboard itself? (CPU being 4800 while both motherboards start at 5600 MT/s).

I don't want to debate the change in metrics. Only to be able to accurately choose the right CPU (4800)for the right motherboard (5600) and right memory without considerations to overclocking. ;)
 
Last edited:
I don't want to debate the change in metrics. Only to be able to accurately choose the right CPU (4800)for the right motherboard (5600) and right memory without considerations to overclocking. ;)
What ram you get is dictated by the CPU. If the CPU can't run it then it doesn't matter what motherboard you use.
 
What ram you get is dictated by the CPU. If the CPU can't run it then it doesn't matter what motherboard you use.
That's precisely the impression I got from disgruntled posters in Amazon purchase comments. But in their case I don't think they put "two-and-two" together. That they only considered the memory requirements based on the motherboard alone. And that virtually all of them sounded like they expected to overclock their memory. With them all citing exasperation over the same thing- that their BIOS recognized lower memory than they expected.

So my choice of using Corsair Vengeance DDR5 4800 is the one to go with. Resulting in stable and compatible memory regarding both the motherboard as well as the CPU. Not so?
 
So my choice of using Corsair Vengeance DDR5 4800 is the one to go with. Resulting in stable and compatible memory regarding both the motherboard as well as the CPU. Not so?
Absolutely. And what ever overclocking you may or may not do has no bearing whatsoever in this decision. The processor specs dictate everything, if the CPU can't run it then how are you going to overclock it? How can you overclock the CPU if it can't identify the ram? If the CPU can't correctly run the ram it shuts down again immediately, it's a critical error.
 
One last question just for the hell of it.

With those DDR5 4800 MT/s sticks, could I then successfully overclock them to something a bit higher?

Though we both agree, overclocking RAM in general is just not something I've ever wanted to do. Another added heat consideration even if it were to work. I just want to make sure I understand at least the basics correctly!

I've pushed my CPU with overclocking a few times...but I just didn't see the point of it. Particularly since I stopped gaming decades ago.
 
Last edited:
Other than the confusion revolving around DDR5, my main concern is just to double my memory to 32GB, in consideration of future technologies like AI. I just don't want to be caught again without adequate hardware, for at least a few more years.
 
With those DDR5 4800 MT/s sticks, could I then successfully overclock them to something a bit higher?
I'm really not sure there, not something I've ever looked at. In fact I haven't played around much with ram in many years it may have all changed since then. My experience is that if you start playing with cas, latency and page rates you start getting all sorts of weird behaviour, and that's if you're playing within limits that the CPU can handle, step outside of that range and the thing won't boot at all until you hard reset the bios.
 
If my mental model is correct, RAM overclock would help with your average CPU-bound program as the majority of CPU clock cycles are wasted in memory stalls. But if you are getting to that point, you are better off compiling source code from scratch and tweaking the makefiles. And a lot of programs that might seem CPU-bound aren't truly CPU-bound because they're wasting time on blocking system calls.
 
I'm really not sure there, not something I've ever looked at. In fact I haven't played around much with ram in many years it may have all changed since then. My experience is that if you start playing with cas, latency and page rates you start getting all sorts of weird behaviour, and that's if you're playing within limits that the CPU can handle, step outside of that range and the thing won't boot at all until you hard reset the bios.
Yep. I could consider going to a much more powerful CPU that would give me a higher memory ceiling of performance, but the tradeoff in thermodynamics doesn't appeal to me. And I never had any intention or real need to overclock the CPU, GPU or memory anyways, apart from all the potentials of unstable performance.

I just want enough computing to use an ancient version of Photoshop, without burning the house down.

I may ultimately have unfounded concerns, but my impression of the technology market at present is that they are so hellbent on improving performance, that they are ignoring the damage potential of all the heat put out by CPUs and GPUs. Gamers may eat it all up hook, line and sinker...but me? Not so much.
 
Last edited:

New Threads

Top Bottom