G.SKILL Reveals World's First Large Capacity 256 GB (64 GB x4) DDR5 U-DIMM Memory at DDR5-6000 CL32 Overclocked Speed

[GFreeman]

G.SKILL International Enterprise Co., Ltd., the world's leading brand of performance overclock memory and PC components, is thrilled to unveil DDR5 memory with an ultra-high kit capacity of 256 GB (64 GB x4), overclocked to DDR5-6000 CL32. Built with high-performance SK hynix DDR5 ICs, this overclock memory speed and timing pushes the boundaries of memory overclocking by fully loading 4 modules of high-capacity DDR5 64 GB U-DIMM modules on a single system; marking a new milestone for high-performance computing, content creation, AI applications, and advanced workstation workloads.

Overclocked, Ultra-Capacity 256 GB (64 GB x4) at DDR5-6000 CL32 In content creation and recent advancements in AI, the need for higher DRAM capacity is essential. Whether calculating large language models or working with complex or high-resolution video content, high-speed memory of high-density 64 GB modules would help to provide a smoother and more efficient workflow. Loaded with AMD EXPO memory overclock profiles, see the DDR5-6000 CL32 256 GB (64 GB x4) kit being stress-tested in the screenshots below on the ASUS ROG CROSSHAIR X870E HERO motherboard with the AMD Ryzen 7 9800X3D desktop processor, as well as on the MSI MPG X870E CARBON WIFI motherboard with the AMD Ryzen 9 9900X desktop processor.



Showcasing Extreme Overclock at up to DDR5-7000 CL38-50-50 with 256 GB (64 GB x4)
Through extensive research and development, G.SKILL memory also demonstrates the overclock potential and limits of ultra-high capacity 64 GB memory modules. Combining 4 of these high-density 64 GB modules for a total of 256 GB, the following screenshot shows this massive capacity kit operating at DDR5-7000 CL38-50-50 on the MSI MEG X870E GODLIKE motherboard and AMD Ryzen 7 9800X3D desktop processor.

Additionally, the massive kit capacity of 256 GB is showcased in low latency DDR5-6400 CL32 in the screenshots below, on the ASUS ROG CROSSHAIR X870E HERO motherboard with the AMD Ryzen 9 9900X3D desktop processor and the MSI MAG B850M MORTAR WIFI motherboard with the AMD Ryzen 7 9800X3D desktop processor



View at TechPowerUp Main Site | Source

 

[Chaitanya]

Another vaporware.

 

[Onasi]

I fail to see the need for 256 gigs of RAM on a consumer desktop platform, but I suppose it’s more of a halo product for GSkill than anything else. Then again, I suppose there ARE weird niche uses for this in SOME workloads paired with a 9950X.

 

[TumbleGeorge]

I fail to see the need for 256 gigs of RAM on a consumer desktop platform, but I suppose it’s more of a halo product for GSkill than anything else. Then again, I suppose there ARE weird niche uses for this in SOME workloads paired with a 9950X.

Consumer platforms also is used as lite and more price accessible workstations.

 

[Dirt Chip]

Those kit's are wonderful, will get them when price will be reasonable.
256GB is good to have primer pro with 4k+ projects together with after effects simultaneously.
128GB can also do it (from personal experience), but it on the edge.

 

[bonehead123]

I fail to see the need for 256 gigs of RAM on a consumer desktop platform, but I suppose it’s more of a halo product for GSkill than anything else. Then again, I suppose there ARE weird niche uses for this in SOME workloads paired with a 9950X.

Well, as they saying goes "it's better to have it & not need it (all the time), as opposed to needing it & not having it"

OR, as I always say: "Ram, bam, thank ya Ma'am" hehehe

 

[Onasi]

Consumer platforms also is used as lite and more price accessible workstations.

True, though I feel that with those capacities (and higher) quad-channel memory becomes more important. Issue is that, yeah, that’s no longer even a remotely affordable feature since TR platforms are now eye-wateringly expensive. The price gulf between consumer desktop and workstation platforms became clear insane in the last few generations.

 

[TumbleGeorge]

True, though I feel that with those capacities (and higher) quad-channel memory becomes more important. Issue is that, yeah, that’s no longer even a remotely affordable feature since TR platforms are now eye-wateringly expensive. The price gulf between consumer desktop and workstation platforms became clear insane in the last few generations.

I also meant many pre-assembled computers, from different brands, not just DIY. They often use motherboards with professional chipsets from C series, and the supported processors are both consumer and their professional variants, for example, low-end Xeon or Ryzen Pro. TR and WS/Server Xeon, yes it's are too expensive for entry level workstations in this days.

 

[Dr_b_]

I fail to see the need for 256 gigs of RAM on a consumer desktop platform, but I suppose it’s more of a halo product for GSkill than anything else. Then again, I suppose there ARE weird niche uses for this in SOME workloads paired with a 9950X.

would like to see more PCIe lanes in the consumer platforms

 

[CosmicWanderer]

I thought these motherboards would reduce the memory frequency when 4 RAM sticks are used. Or has that changed with X870?

 

[A Computer Guy]

True, though I feel that with those capacities (and higher) quad-channel memory becomes more important. Issue is that, yeah, that’s no longer even a remotely affordable feature since TR platforms are now eye-wateringly expensive. The price gulf between consumer desktop and workstation platforms became clear insane in the last few generations.

Weren't they always eye-wateringly expensive? If you are willing to take a hit in top performance Zen2 based thread-rippers can be quite affordable for around $800 to $1300 (CPU+MB+RAM) on the used market. I nearly snagged a 5000 series combo for $1000 (just CPU and MB) but that scenario appears to be rare and there is some difficulty in assurance of getting non-vendor locked CPU's. I had to ask several times on ebay listings if the CPU they were selling was vendor locked and they didn't know so now I just ask if it came out of Dell or Lenovo.

Looking on ebay what's really frustrating is vendor locked TR CPU's are often much cheaper but once you try to get a refurbished DELL then the expense goes right back up on "recertified/renewed" brand name alone. At one point I was just looking at getting some shucked DELL TR board for $125 + PSU + Dell locked CPU for under $800 but then I came to my senses when I realized I'll have to build a case for that kind of Frankenstein motherboard as well as finding a proper wiring harness just for the power switch - ugh no thanks, not that desperate yet.

 

[TumbleGeorge]

I thought these motherboards would reduce the memory frequency when 4 RAM sticks are used. Or has that changed with X870?

Maybe is not accidental that in both experiments are used 9800X3D and not CPU's with two computational chiplets.

 

[A Computer Guy]

Maybe is not accidental that in both experiments are used 9800X3D and not CPU's with two computational chiplets.

I don't think it should matter as there is only one I/O die and it needs to go through that anyway.

 

[bonehead123]

Also, since there is no mention of it, and all the screenshots show AMD-powered rigs, does this mean that these sticks have NOT been tested in or won't work that well in an Intel rig ?

Just curious

 

[R-T-B]

I fail to see the need for 256 gigs of RAM on a consumer desktop platform, but I suppose it’s more of a halo product for GSkill than anything else. Then again, I suppose there ARE weird niche uses for this in SOME workloads paired with a 9950X.

AI. AI sees memory and goes nom nom.

I use 128GBs on my current workstation and it helps a bit with my work, but probably less than it would with an AI workload by far. I generally don't use more than 90GBs even with VMs, in truth.

 

[Dr. Dro]

I fail to see the need for 256 gigs of RAM on a consumer desktop platform, but I suppose it’s more of a halo product for GSkill than anything else. Then again, I suppose there ARE weird niche uses for this in SOME workloads paired with a 9950X.

64 GB sticks make 128 possible on a 1 DPC board, and that's enough reason for them to exist. I'll probably have to upgrade my RAM in the future, 32 GB main memory is not enough to make the best out of the RTX 5090.

 

[igormp]

AI. AI sees memory and goes nom nom.

I use 128GBs on my current workstation and it helps a bit with my work, but probably less than it would with an AI workload by far. I generally don't use more than 90GBs even with VMs, in truth.

I've been maxing out my 128GB rig quite often, to the point of facing some OOM errors. Shitty pic before the OOM hit since the system became mostly unresponsive:

Managed to workaround it for now by really lowering my batch sizes, but now my GPUs are becoming really data starved and training runs are taking 10~20 hours

I should really finish testing up my new 256GB rig and move things over to make my life easier.

 

[R-T-B]

I've been maxing out my 128GB rig quite often, to the point of facing some OOM errors. Shitty pic before the OOM hit since the system became mostly unresponsive:
View attachment 396171

Managed to workaround it for now by really lowering my batch sizes, but now my GPUs are becoming really data starved and training runs are taking 10~20 hours

I should really finish testing up my new 256GB rig and move things over to make my life easier.

Raise swap size maybe? I mean short of adding memory or lowering requirements, obviously. My swap is usually the same as my ram size too, if only for proper hibernation support. That works for me but ymmv.

 

[igormp]

Raise swap size maybe? I mean short of adding memory or lowering requirements, obviously. My swap is usually the same as my ram size too, if only for proper hibernation support. That works for me but ymmv.

My swap size is already the same size of the RAM amount, issue is mostly that the working set is bigger than the total RAM size, so it ends up paging wayyy too much and moving things to a crawl (and I guess my OOM settings may be a bit too aggressive).

I lowered the requirements (as in lower batch sizes) for the time being as a short-term (and slower) solution, but the proper one would be more memory within the new setup.

 

[_roman_]

Really big room for improvement. I miss the windows 11 pro DRAM software. Was it called zen-timings? I think one of those timing values looked already worse.

I saw they are using 4 dimm modules. I barely use my 2x32Gib DRAM. The timings for that product in this news piece could be much better compared with my beginner overclocking daily stable result.

 

[onoffertrips]

I fail to see the need for 256 gigs of RAM on a consumer desktop platform, but I suppose it’s more of a halo product for GSkill than anything else. Then again, I suppose there ARE weird niche uses for this in SOME workloads paired with a 9950X.

You need only 8GB ram then. enough for your use.

 

[Raiden85]

Finally, fast single 64GB sticks. 128GB with two sticks here I come.

 

[Wirko]

Consumer platforms also is used as lite and more price accessible workstations.

Yes. As a consequence, now that quarter-terabyte RAM is within reach, many more PC builders will come to these forums, complaining that most AM5 mobos have far too few PCIe lanes to connect a capture card, a 2x25GbE card, a third 4090 and two storage controllers.

 

[A&P211]

I fail to see the need for 256 gigs of RAM on a consumer desktop platform, but I suppose it’s more of a halo product for GSkill than anything else. Then again, I suppose there ARE weird niche uses for this in SOME workloads paired with a 9950X.

You might not need it but someone else would need it.

 

[lexluthermiester]

Another vaporware.

How so? It's GSkill. They're not known for making empty product announcements.

I fail to see the need for 256 gigs of RAM on a consumer desktop platform

I can think of a few, but those are more Prosumers type things.