Corsair Postulates That DDR5 Memory Runs Hotter

[Chrispy_]

The next generation of memory also mandates the inclusion of on-die EEC which…

Im assuming that’s ECC?

and I don’t think the market segregation team at Intel is gonna like that.

\o/
Win-win; we get ECC and we get to flip the bird to Intel's shittiest department.

 

[Tetras]

The next generation of memory also mandates the inclusion of on-die EEC which…

Im assuming that’s ECC?

and I don’t think the market segregation team at Intel is gonna like that.

It's not equivalent to the kind of ecc support that workstations and servers have now so there will still be that split in the market.

 

[Valantar]

Yeah, was it @TheLostSwede who explained this so well at one point? From what I can remember, DDR5's on-die ECC protects against static in-RAM data errors, but not in-flight data errors, which (IIRC) are more common. So there's some additional protection, but not to the level of full ECC, which protects against in-flight errors as well.

 

[TheLostSwede]

For those interested.

DDR5 features | Interface IP | DesignWare IP | Synopsys

Technical Bulletin: A Few Key Features to Note About DDR5. Learn about DDR5 for servers, cloud, networking & consumer SoCs

www.synopsys.com

 

[Chrispy_]

Yeah, was it @TheLostSwede who explained this so well at one point? From what I can remember, DDR5's on-die ECC protects against static in-RAM data errors, but not in-flight data errors, which (IIRC) are more common. So there's some additional protection, but not to the level of full ECC, which protects against in-flight errors as well.

baby steps.

 

[MentalAcetylide]

Hotter running graphics cards, hotter running CPUs, and now possibly hotter functioning DDR5 RAM. Looks like things are heating up in the PC world. With these developing trends, I wonder how long before the cooling solution for the average PC ends up costing more than the PC itself.

 

[mechtech]

For those interested.

DDR5 features | Interface IP | DesignWare IP | Synopsys

Technical Bulletin: A Few Key Features to Note About DDR5. Learn about DDR5 for servers, cloud, networking & consumer SoCs

www.synopsys.com

Thanks

Quite a bit of changes. Looks like the cpu memory controllers will need some rework.

Also curious to see how much extra these will be over ddr4?

 

[FireFox]

Nvidia GPUs runs hot, AMD GPUs + CPU runs hot, Intel CPUs runs hot, Samsung SSD runs hot and now Corsair next Ram Gen will run hot.
Now I understand why the glaciers are melting and the planet is warming so fast

 

[watzupken]

One thing to note is that DDR5 also runs at lower voltages, 1.1V I believe, so that's about 10% less than DDR4.
Also buck converters have very high efficiency and they don't need to put a small one.
The chips are being designed with the on-die ECC, so maybe it won't be a big increase in heat, who knows. As they go into new nodes though, they will likely get some power savings(which is probably not going to be very big, but even small things help).

I think Corsair is just saying that to cover their ass in case they release a kit which thermal throttles, "it's not our fault! It's how DDR5 is"-kind of thing.

1.1V is JEDEC standard, and not likely to run at high clockspeed or tight latencies. For these overclocked RAM, I am expecting higher power consumption and heat output just by considering the clockspeed they are running. We don't have to speculate for too long before DDR5 becomes more common.

 

[TheoneandonlyMrK]

Hardly rocket science though is it, pAck in more performance, then add the power regulating circuit = more heat.

Obviously ffs.

 

[persondb]

1.1V is JEDEC standard, and not likely to run at high clockspeed or tight latencies. For these overclocked RAM, I am expecting higher power consumption and heat output just by considering the clockspeed they are running. We don't have to speculate for too long before DDR5 becomes more common.

JEDEC standard is for up to 3200MHz or 6400MT/s speeds. You don't really expect more than that for some time, right? Most initial modules will also probably be at JEDEC timings too. If we take a look at DDR4, they have a 1.2V and the tighter latencies are generally ran with 1.35V, maybe more for the more extreme kits.

The actual memory array clock has been kept constant for a long time too(200MHz, I think?).

 

[Chrispy_]

Do your own research, B die finder is a good tool to identify B die.

On your own link, the 4000+ have profile 1 at 1.45, and profile 2 at 1.35.

Give it up man - your comments are too easy to disprove. I see 1.35V across the board for the 4000 kit, but also the 4133 kit and the 4400 kit at 4266 speeds.

Patriot DDR4-4000

Patriot DDR4-4133

Patriot DDR4-4400

Also, Samsung stopped making B-die in early June last year.
Trident Z Royal Gold (isn't that supposed to be the best of the best cherry-picked old-stock Samsung B-Die?) is only 1.4V and so expensive it's not even funny.

F4-4000C16D-32GTRGA-(EOL) - Specification - G.SKILL International Enterprise Co., Ltd.

Trident Z Royal DDR4-4000 CL16-16-16-36 1.40V 32GB (2x16GB)

www.gskill.com

An example of 4266+ Kits currently on sale in the UK, not a single one of which is higher than 1.45V. No 1.5V in sight, no 1.6V anywhere at any speed, all at speeds vastly in excess of the 3800+ you're moaning about.

 

[Valantar]

1.1V is JEDEC standard, and not likely to run at high clockspeed or tight latencies. For these overclocked RAM, I am expecting higher power consumption and heat output just by considering the clockspeed they are running. We don't have to speculate for too long before DDR5 becomes more common.

DDR5 JEDEC standards currently go to 6400, and are slated to hit 8400 in time (just as DDR4 started at 2133, quickly went to 2400 and 2666, and then hit 3200 after a few years). So high clocks are definitely planned. Latencies for JEDEC specs are always loose - servers generally don't care much about those. But tightenting latencies don't require that much voltage, so if you can run 8400 CL[whatever] at 1.1V, then you can most likely run 8400 CL[somewhat less] at 1.2-1.3V. Of course, those speeds in early bins are likely to require mroe voltage, but that was the case for DDR4 as well - heck, finding (and running) anything above 3200 was difficult for quite a while. And 8400 is pretty fast, after all. Something like 6400 with low latency at 1.2V or thereabouts ought to become common pretty quickly once there's a market for these DIMMs.

 

[Chrispy_]

Chances are good that the initial run of >6400MHz kits with loosey-goosey timings can be slowed down to 6400MHz with much tighter timings.

Early Alder Lake and Zen4 DDR5 IMCs are likely to be less that stellar too as this will be their first swing at DDR5 - Who knows how many AGESA revisions it will take for Zen4 to reliably run at >6400MHz with tight timings!

I'd like to be optimistic but history has proven time and time again that the first gen of any new tech tends to be superseded quickly as it's basically a paying beta that spends the majority of its product cycle getting patched and fixed up.

 

[Valantar]

Chances are good that the initial run of >6400MHz kits with loosey-goosey timings can be slowed down to 6400MHz with much tighter timings.

Early Alder Lake and Zen4 DDR5 IMCs are likely to be less that stellar too as this will be their first swing at DDR5 - Who knows how many AGESA revisions it will take for Zen4 to reliably run at >6400MHz with tight timings!

I'd like to be optimistic but history has proven time and time again that the first gen of any new tech tends to be superseded quickly as it's basically a paying beta that spends the majority of its product cycle getting patched and fixed up.

True, though I wonder if they have been focusing a lot of reasources on IMC development to make the DDR5 transition better than the DDR4 transition was - AMD has gone from having pretty terrible IMCs to competing for the best ones out there in just a couple of years, and Intel has likely been sampling server DDR5 systems for a while now. Given the IMC changes required I still wouldn't expect first-gen DDR5 to outperform DDR4>4000, but I'm vaguely hopeful that it won't be a complete shitshow. I guess we'll see though.

 

[Chrispy_]

"Not a complete shitshow" would be nice. I'm prepared for the worst but hoping they exceed that low bar this time!

 

[eidairaman1]

It's not equivalent to the kind of ecc support that workstations and servers have now so there will still be that split in the market.

My board supports ecc

 

[Tetras]

My board supports ecc

Yeah, AMD may continue to support it with DDR5, but I was referring to Intel.

 

[The red spirit]

Haha so the RAM heatsinks will serve a purpose for the first time ever?

Forgetting Rambus?