Adata sets new world record for DRAM overclocking

[LukeSavenije]

Sources: Tomshardware, Twitter (ADATA), HardwareBot

 

Adata's XPG Overclocking Lab (XOCL) has managed to overclocked their RGB-lit Spectrix D60G memory to a staggering 5,634 MHz, as recorded by HWBot. This beats HyperX Predator as the fastest DDR4 memory on earth. The overclock has been done on a Z390i gaming edge ac and a 9900k

 

Quote

Quote

As expected, the overclocking team resorted to exotic cooling, in this case liquid nitrogen, to cool both the CPU and memory module. Of course, when you're aiming to break world records, you have to push your hardware to its limit, so your typical cooling methods have no place in the competitive overclocking world.When it comes to pushing for raw speed, memory efficiency is basically thrown out the window. In order to hit 5,634 MHz, the XPG Overclocking Lab had to use very loose timings that you simply won't see in normal day usage. The overclockers set the Spectrix D60G's CL timings to CL31-31-31-46 3T. Although Adata didn't provide any insight on the operating voltage, we suspect it was about 2V.

 

[Uptivuptiz]

Imagine in 10 years we might get RAM with speeds of that caliber right out of the box...

[LukeSavenije]

13 minutes ago, Uptivuptiz said:

Imagine in 10 years we might get RAM with speeds of that caliber right out of the box...

i mean... about what happened with ddr3

[Master Disaster]

50 minutes ago, Uptivuptiz said:

Imagine in 10 years we might get RAM with speeds of that caliber right out of the box...

At those timings it's pretty pointless tbh, it's more a "look what we did" thing than anything useful.

[LukeSavenije]

1 minute ago, Master Disaster said:

At those timings it's pretty pointless tbh, it's more a "look what we did" thing than anything useful.

it's like cl30 or something stupid like that

[XR6]

4 minutes ago, Master Disaster said:

At those timings it's pretty pointless tbh, it's more a "look what we did" thing than anything useful.

Yeah, it's nothing more than a publicity stunt. I think it's still interesting to see RAM at 5000MHz+ though, even if the timings are at CL31.

[Master Disaster]

3 minutes ago, LukeSavenije said:

it's like cl30 or something stupid like that

Quote

CL31-31-31-46 3T

3T, ooof.

[Master Disaster]

3 minutes ago, XR6 said:

Yeah, it's nothing more than a publicity stunt. I think it's still interesting to see RAM at 5000MHz+ though, even if the timings are at CL31.

Oh without a doubt, I very often find myself looking at world record overclocks. Then I look at my 4.2Ghz Ryzen and sigh LOL

[LukeSavenije]

3 minutes ago, Master Disaster said:

Oh without a doubt, I very often find myself looking at world record overclocks. Then I look at my 4.2Ghz Ryzen and sigh LOL

then i look at my memory clocked at 3600 cl17

 

even harder sigh

[porina]

58 minutes ago, Uptivuptiz said:

Imagine in 10 years we might get RAM with speeds of that caliber right out of the box...

Possibly closer than you think. DDR5 is aiming to double bandwidth over DDR4, so officially defined 6400 could be expected at some point. Not necessarily day 1. As an indication standards compliant 5200 has been developed https://www.anandtech.com/show/13605/sk-hynix-develops-first-16-gb-ddr5-chip-demos-ddr5-rdimm Don't know if anyone has claimed faster since then.

7 minutes ago, Master Disaster said:

At those timings it's pretty pointless tbh, it's more a "look what we did" thing than anything useful.

5634C31 is equivalent to 3200C18 so, still lower latency than standards compliant 3200. A bit higher latency than most enthusiast kits typically C16, or faster if you throw more money at it.

 

Of course, with the requirement of exotic cooling it isn't for normal usage, but I could really use that bandwidth. Note not all use cases care about latency. Bandwidth can really help too. Given a choice I'd take more bandwidth over lower latency any day.

[jetforce]

Response times be faster than human response times wouldn't they

[Uptivuptiz]

29 minutes ago, Master Disaster said:

At those timings it's pretty pointless tbh, it's more a "look what we did" thing than anything useful.

I didn't mention the timings. At that point I hope the timings have been optimized.

[LukeSavenije]

1 minute ago, Uptivuptiz said:

I didn't mention the timings. At that point I hope the timings have been optimized.

ddr4 had some problems with that already... but that aside

[Atmos]

1 hour ago, Uptivuptiz said:

Imagine in 10 years we might get RAM with speeds of that caliber right out of the box...

doubt it lol.

Our current technology has slowed in advancement pretty heavily since about 2017 when we hit the transistor gate, and we can only get so much smaller without developing to an entirely different cpu architecture using ridiculous resources and materials. Unless of course we start developing massive hardware modules liek the ryzen cpus... We've almost hit the brick wall of our current technology in terms of this size factor.

 

Here's hoping though that we find some kind of super-material able to process the kinds of data we stream into it so we can continue to blow past previous expectations without increasing size massively... like amd is starting to realize we need to do to get real performance upgrades in the same package.

[Dabombinable]

28 minutes ago, LukeSavenije said:

ddr4 had some problems with that already... but that aside

Bandwidth increases actually negate higher timings. To a point. Go too far and of course there are diminishing returns. I found that out with DDR2 667, 800 and 1066.

[LukeSavenije]

3 minutes ago, Dabombinable said:

Bandwidth increases actually negate higher timings. To a point. Go too far and of course there are diminishing returns. I found that out with DDR2 667, 800 and 1066.

yeah... true

 

and at the start of ddr4 the same case

[Humbug]

But can it run crysis....... without crashing?

[LukeSavenije]

1 minute ago, Humbug said:

But can it run crysis....... without crashing?

[Doobeedoo]

Would be great to see these memory clocks on new Ryzen chips. 

[LukeSavenije]

Just now, Doobeedoo said:

Would be great to see these memory clocks on new Ryzen chips. 

well... good luck running ln2 all day then

 

and the fact that it's cl31

[cj09beira]

1 hour ago, Master Disaster said:

3T, ooof.

didn't even know that was possible 

[Doobeedoo]

2 minutes ago, LukeSavenije said:

well... good luck running ln2 all day then

 

and the fact that it's cl31

Heh well we'll at least have much faster memory support with new series. 

[DildorTheDecent]

https://hwbot.org/submission/4156022_ogs_memory_frequency_ddr4_sdram_2863_mhz

 

http://valid.x86.fr/cc42lc

 

ADATA record was short lived. But that's how HWBOT works. They probably have several high clocked results on hand. All ready to resubmit when their previous score is beat.

 

[Gegger]

Definitely cool, but usually most people don't have a giant tank of LN2 lying around. However, I would love to see these speeds in the future.

[Ground]

6 hours ago, DildorTheDecent said:

https://hwbot.org/submission/4156022_ogs_memory_frequency_ddr4_sdram_2863_mhz

 

http://valid.x86.fr/cc42lc

 

ADATA record was short lived. But that's how HWBOT works. They probably have several high clocked results on hand. All ready to resubmit when their previous score is beat.

 

unlikely for this one - bdie is at its limit (and apparently a lot harder to run). The new run is using micron