G.Skill unveils blazing 10,934 MT/s DDR5 and ultra-high-capacity 512GB R-DIMM at Computex 2025

[DragonSlayer101]

What just happened? G.Skill used Computex 2025 to debut several high-performance DDR5 memory solutions, including a 512GB DDR5-6600 kit and a single 64GB CAMM2 module running at DDR5-10000 speeds. The company also demonstrated a 2×24GB kit overclocked to 10,933 MT/s on an Asus ROG Maximus Z890 APEX motherboard with an Intel Core Ultra 9 285K CPU.

A notable highlight at G.Skill's booth was a 512GB (8×64GB) DDR5-6600 CL42 R-DIMM kit built with SK Hynix ICs. The kit ran on an AMD-based system featuring a Ryzen Threadripper PRO 7985WX processor and an Asus Pro WS WRX90E-SAGE SE motherboard.

G.Skill also had a 256GB (4×64GB) DDR5-7000 CL38 kit running in two overclocked AMD-powered systems. One featured a Ryzen 7 9800X3D paired with an Asus ROG Crosshair X870E Hero motherboard, while the other used a Ryzen 9 9900X with an MSI MEG X870E Godlike board.

Other showcased products included a 128GB (2×64GB) DDR5 CL48 kit running at 8,400 MT/s on an MSI B850 M-ATX motherboard with a Ryzen 9 9600X processor and a 64GB (2×32GB) DDR5-9000 CL48 kit running on an Intel-based system with an MSI MEG Z890 UNIFY-X motherboard and a Core Ultra 9 285K CPU.

The company also demoed several ultra-low-latency setups, led by a 192GB (4×48GB) DDR5-6000 CL26 kit running on an MSI MAG B850M Mortar WIFI motherboard. Another low-latency build featured a 192GB (4×48GB) DDR5-6400 CL28 kit installed on a Gigabyte X870E AORUS MASTER motherboard. Both examples used a Ryzen 9 9900X processor.

Also on display was an ultra-high-speed 64GB DDR5-10000 CAMM2 module with CL56-70-70 timings, installed on an Intel system powered by a Core Ultra 9 285K processor. Another demo featured a white Trident Z5 NeoX RGB DDR5 gaming memory kit running at 10,747 MT/s with CL68-128-128 timings, validated on an Intel PC using the same processor.

While some kits were purely for demonstration, most will eventually hit the market. However, G.Skill has yet to announce official pricing or availability timeframes. Given the advanced specs and performance on display, many enthusiasts will be eager to see when these products become available.

Permalink to story:

G.Skill unveils blazing 10,934 MT/s DDR5 and ultra-high-capacity 512GB R-DIMM at Computex 2025

 

[Bunnings]

Perfect for a tradie themed build.

 

[moon982]

What is DDR5-10000? What is 10000 after the DDR5 the speed?

Is it the speed in GHz or the name.

I see DDR5 9000 there also 6600.

 

[TooQik]

What is DDR5-10000? What is 10000 after the DDR5 the speed?

Is it the speed in GHz or the name.

I see DDR5 9000 there also 6600.

Yes, the number refers to the speed of the RAM, specifically the amount of data the RAM can transfer in one second, measured in MT/s.

Some people/manufacturers also state this number as the frequency of the RAM in MHz, but this is technically not correct for DDR RAM, as the actual frequency the RAM runs at is half of the stated value. For example, DDR RAM advertised at 10000 MHz runs at a frequency of 5000 MHz.

 

[Manoa]

This demo is nice, but you will not, in reality have this, this is highly selected combinations of hardwares (aka: like intel used chiller under the table), in reality you should not expect to get anywhere near this level of performance, unless you are ready to buy 10 mobo and another 10 processors to find out which of them combination is able to get this performance, as a demo this corporations do this all the time becuase they can: they got all the money and it is worth for them to buy 10 of each to find out which can - becuase the demo will make people belive they can get this performance, to make them buy it: they may spending 100,000$ on buying and rebuying 10 mobo's and 10 processors, but if the demo makes people go buy 1,000,000 units of they hardwares becuase they blieve they can get this performance - it's worth for them.

the only real most-advanced memory support right now is the new Zen 5 thredripers, because AMD says they improved the IO die. the memory iteself is very far from being the bottleneck right now - both on desktop and high desktop plataforms