G.Skill has created the world's first air-cooled 5000MHz DDR4 memory kit

[WYP]

This was only possible on LN2 a few years ago.



Read more about G.Skill creating the world's first air-cooled 5000MHz DDR4 memory kit.

 

[AlienALX]

OK I don't get this out very often, but..

https://www.youtube.com/watch?v=DcJFdCmN98s

Ram has gone ham.

 

[Dicehunter]

Would this actually have any real world benefits for applications or games over say 3000MHz ?

 

[NickOmega]

Ryzen gonna love that :lol:

 

[TheF34RChannel]

Would this actually have any real world benefits for applications or games over say 3000MHz ?

Nope. None whatsoever.

 

[ByteMyAscii]

The real benefit of this is very questionable.
I don't see the benefit of this extending to anything outside of artificial benchmarks.
Are there any real tasks shown to be benefitting from this, with the increased latency that comes with that higher speed.
To me it seems that it would require a pretty specialist workload to offer anything to a buyer, even designing around the hardware.

 

[NeverBackDown]

Ryzen gonna love that :lol:

Probably would see the most benefit but not sure if you'd notice after 4000mhz

 

[Avet]

This one is for pro overclockers. Real world benefit... None.

 

[AlienALX]

This one is for pro overclockers. Real world benefit... None.

Real world gaming usage, probably none yes. However, throughput and boring server like stuff ? it will make a big difference.

You will probably see AIDA benchmarks for memory on this eventually, judge it then

I also think that once Ryzen gets faster you will find that faster ram will also help it. Once they start hitting 5ghz and or beyond? the IF will need to function a lot faster than it does now. 3200 is about where Ryzen does well atm, but remember Ryzen is only running 3.8-4ghz.

 

[Avet]

Real world gaming usage, probably none yes. However, throughput and boring server like stuff ? it will make a big difference.

You will probably see AIDA benchmarks for memory on this eventually, judge it then

I also think that once Ryzen gets faster you will find that faster ram will also help it. Once they start hitting 5ghz and or beyond? the IF will need to function a lot faster than it does now. 3200 is about where Ryzen does well atm, but remember Ryzen is only running 3.8-4ghz.

Servers and boring stuff still run on ECC and that doesn't go to 5GHz. Neither can HEDT CPUs use that speed. And also only few cream of pre-binned CPUs can run it. Not every memory controller can handle it. It is at the edge of present tech.

 

[AlienALX]

Servers and boring stuff still run on ECC and that doesn't go to 5GHz. Neither can HEDT CPUs use that speed. And also only few cream of pre-binned CPUs can run it. Not every memory controller can handle it. It is at the edge of present tech.

I wasn't talking about today fella. Give it time I mean, it's pretty obvious that at some point in the future we will be all running 5ghz memory.

 

[NickOmega]

Probably would see the most benefit but not sure if you'd notice after 4000mhz

Oh I don't doubt that, was just an off the cuff remark based on Ryzen performing better with fast RAM
Maybe in a couple of years as the Infinity Fabric and Zen matures...

 

[Avet]

I wasn't talking about today fella. Give it time I mean, it's pretty obvious that at some point in the future we will be all running 5ghz memory.

Yes but it won't be DDR4 5GHz memory. Maybe DDR5 or 6. That doesn't increase usefulness of this kit.

 

[grec]

Ryzen APUs are very bandwidth constrained on the GPU portion too. With the prices of standalone GPUs nowadays and with increasing abilities to share graphics loads between both the dGPU and iGPU (Which I assume will be a bigger thing once AMD brings out some lower end Vega 20-28 based cards(Using the Vega M die) to pair with those Vega 8-11 iGPUs), some 4.7Ghz+ memory could really help close the bandwidth gap with a dGPUs GDDR/HBM.

 

[Avet]

Ryzen APUs are very bandwidth constrained on the GPU portion too. With the prices of standalone GPUs nowadays and with increasing abilities to share graphics loads between both the dGPU and iGPU (Which I assume will be a bigger thing once AMD brings out some lower end Vega 20-28 based cards(Using the Vega M die) to pair with those Vega 8-11 iGPUs), some 4.7Ghz+ memory could really help close the bandwidth gap with a dGPUs GDDR/HBM.

If you pair the cheapest CPU with the most expensive memory kit just to use iGPU something is wrong with you. The slowest memory and proper GPU would give you much better results.

 

[grec]

If you pair the cheapest CPU with the most expensive memory kit just to use iGPU something is wrong with you. The slowest memory and proper GPU would give you much better results.

Obviously(well, given GPU prices recently slightly faster memory+APU has been a more cost effective approach to low end gaming for many but this would obviously be ott for the time being)

I said the benefits would come in having an extra 11 Vega CU's properly fed working on a given task alongside the 20-64CUs of the upcoming and current Vega dGPUs(Ryzen Mobile's 11CUs + Vega Mobiles upto 28CUs could be a very potent mix with the right memory). Many modern compute programs and tasks can easily share an embarrassingly parallel problem across both the iGPU and dGPU(s) but many tasks could be constrained by the speed of the Infinity Fabric and system memory bandwidth.

Ofcourse, THIS exact kit is abit too bleeding edge for most current use cases, but with next generation memory controllers it could be much more viable.