Memory maker aims to develop DDR5 modules clocked at over 10,000 MHz

jsilva

Posts: 47   +0
Staff
Forward-looking: We still have to wait a bit more for the adoption of DDR5 as the new standard, but manufacturers are already hard at work on it. Micron, Samsung, Team Group, and SK Hynix are some of the companies known for developing DDR5 memory modules, but Netac is joining them with something new. The Chinese memory manufacturer plans to develop DDR5 memory modules clocked at over 10,000 MHz.

The DDR5 memory standard released by JEDEC set clock speeds between 4,800 MHz and 6,400 MHz, but SK Hynix and Samsung have already promised more than that. Samsung stated that its DDR5 memory would go up to 7,200 MHz, and SK Hynix went a step further claiming it would reach speeds as high as 8,400 MHz.

Netac, which has just announced that it received the first batch of DDR5 DRAM memory chips from Micron, plans to push DDR5 memory speeds even higher, going beyond the 10,000 MHz barrier.

The Micron chips Netac received have the IFA45 Z9ZSB FBGA code. According to Micron, these are engineering samples with the MT60B2G8HB-48B ES:A part number, featuring a 16 GB (2Gx8) capacity and 40-40-40 timings.

Considering Micron does not rate their chips above 6,400 MHz, Netac has quite the task in its hands. The Chinese manufacturer will have to increase voltages (VDD/VDDQ/VPP) considerably and loosen up timings to clock the memory at above 10,000 MHz.

Compared to DDR4, however, the achievement seems quite possible. DDR4's JEDEC standard speed is set at 2,133 MHz, but manufacturers released memory kits at more than double that, while professional overclockers even went above the 7,000 MHz mark.

Image credit: IT Home

Permalink to story.

 

QuantumPhysics

Posts: 4,721   +5,131
I'm definitely waiting for DDR5, an Intel Motherboard, and an appropriate Core i9Extreme processor for my next build.

Hope it's worth the wait!.
 

Danny101

Posts: 1,608   +691
Okay. Finally figured it out. Time new builds at the end of a memory generation. By that time all of the other components for that generation will be about as good and as stable as they're going to get. I jumped on board too early with AMD with the 3xxx chipset. However, I still don't see a reason to go for the 5xxx chipset unless the motherboard breaks. We all have our strategies and is often fluid. There's probably no best one I suppose.

After doing this for 25 years, I'm still learning. I'm also remiss to
just toss old hardware and still looking for ways to re-use them. Electronics are caustic and I'm trying to more efficient in the life cycles. By waiting and investing until the end of a generation, it should still be power efficient with enough performance to carry me through until the end of the next. I was 2 generations behind this time around and was chomping at the bit to upgrade and jumped early into the AMD Ryzen platform. Granted it was about as power efficient as an upgrade I've ever had. Unfortunately, I'm going to end up buying 2 CPUs instead of one and ultimately that is the main reason that proper to prefer to push late into generational platforms.
 
Last edited:

Danny101

Posts: 1,608   +691
Yeah, 10GHz DDR5... Probably with a CAS latency of like, 50.
Kind of plays into my point. It's much faster than the rest of the components currently available and has to be artificially hampered to work with the rest of the system. It's fast, but the response time will be slow as compared to DDR 4.
 

neeyik

Posts: 1,839   +2,151
Staff member
Yeah, 10GHz DDR5... Probably with a CAS latency of like, 50.
A CAS latency of 50 just means 50 clock cycles between addressing a column and when the data becomes available. DDR5-10000 has an I/O bus clock of 5 GHz and a DRAM clock of 1.25 GHz, so 50 cycles equates to 40 nanoseconds. For something like DDR4-2400, this time period would be equivalent to a CAS latency of 12. In other words, if DDR5-10000 did have a tCAS of 50, then that would be perfectly okay.
 

kiwigraeme

Posts: 293   +241
I'm definitely waiting for DDR5, an Intel Motherboard, and an appropriate Core i9Extreme processor for my next build.

Hope it's worth the wait!.
Why ? - yes most of us will be building a DDR5 system in the future - but I won't tie myself to Intel or AMD - I 'd wait like I always do for it to settle down .

Fast memory does not do much for most general users - for your gaming 3200 DDR4 or 4200 DDR4 won't be much different .
I know stuff all about it - as it's mostly a hobby - just to get fast speeds - I assume it only makes a difference on server/threadripper like boards folding proteins or something - You will need a lot of it and you will need superfast M2 drives .
For gaming at 4K you can put most of your money in the GPU and a nice fat fast M2 drive, cooling, keyboard/mouse and the display - most other things don't matter too much

whoops remember you like flight sim - so maybe enough ram
 
I think you will hardly notice any fps difference in games compared to DDR4 coz games are hardly memory bandwidth limited. I'm baffled that even at single channel compared to dual channel the loss in fps is minimal. The biggest bottleneck is mostly the GPU. So putting more money there is (for gaming) almost always the best choice, if you can find one at a reasonable price that is.
 

godrilla

Posts: 194   +116
The elephant in the room is the 2.4 voltage needed to achieve such an exotic clock but then again overclocking memory usually does require higher power and thus needing ln2 cooling. I could be wrong if they recreated the wheel I mean ram cooling. The question is can you take advantage of the higher clocks by lowering the clocks and improving a highest clock to lowest timings ratio? Obviously at a practical use case scenario*.
 

Rayneofpayne

Posts: 318   +297
Meh wait till they have kits that have high mhz clocks and lower timings, jumping on early has some drawbacks like early process nodes with high failure rates and bad timings to keep it stable. Just like AM5 platform, I would wait a cpu gen before dropping cash down so you aren't a beta tester, besides it will take a few years for technologies to get adopted on the software side anyways.
 

Watzupken

Posts: 131   +102
I would wait and see what benefits will DR5 bring to the table before jumping onto the bandwagon. In any case, highly overclocked ones will not be cheap anyway, so no point rushing into it. The frequency will anyway improve over time and even the 10000Mhz will sound obsolete in a few years.
 

Avro Arrow

Posts: 1,130   +1,259
TechSpot Elite
A CAS latency of 50 just means 50 clock cycles between addressing a column and when the data becomes available. DDR5-10000 has an I/O bus clock of 5 GHz and a DRAM clock of 1.25 GHz, so 50 cycles equates to 40 nanoseconds. For something like DDR4-2400, this time period would be equivalent to a CAS latency of 12. In other words, if DDR5-10000 did have a tCAS of 50, then that would be perfectly okay.
Yeah, so my guess was accurate. It happens from time to time. (I do know what CAS latency means BTW.) :laughing:

I think that the first DRAM modules that I used in a build were SIP. That's how long I've been working with RAM. I knew that 50 would sound funny but I also reasoned that, based on previous CAS latency numbers, 50 wouldn't be an unreasonable prediction. You confirmed it, so thank you. :D
 

neeyik

Posts: 1,839   +2,151
Staff member
My previous calculation was off anyway! I forgot that tCAS is based on the I/O clock, not the DRAM one. JEDEC has ratified 3 categories for DDR5 (A, B, and C-spec) and something like DDR5-4800A has a tCAS of 34 cycles, which is a latency of 14.2 nanoseconds (for DDR4-2400, this would equate to a tCAS of 17, so not bad, but not great either - I'm using DDR4-3000, with a tCAS of 15).

For the main JEDEC DDR5 A-specs, they fall around the 14 to 14.5 ns region, but JEDEC only goes up to DDR5-6800 at the moment (46 cycles). So realistically, DDR5-10000 is likely to have a tCAS of around 75 to 80.