JEDEC Publishes DDR5 Standard - Launching At 4.8 Gbps, Better Power Efficiency

[TemplarGR]

Also AMD Apus do have 256MB EDcache, they don't depend on DDR4 that much.

Source? This is the first time i read something like this. If AMD APUs actually had 256MB of cache, they would have had much better performance in gaming than they do. Unless you refer to specific laptop models or something, but still, this is news to me.

[TemplarGR]

APU this, APU that. Just a reminder that APU = AMD marketing for CPUs with an iGPU.

Practically all Intel CPUs are "APUs" since nearly forever. They don't seem to be an endangered species.

Actually, no. Intel igpus are just discreet gpus on package. What AMD actually meant with APU was a homogenous chip with context switching. The original vision was for a common pool of memory and cpu/gpu cores acting in unison on it. It is very different.

[artivision]

Source? This is the first time i read something like this. If AMD APUs actually had 256MB of cache, they would have had much better performance in gaming than they do. Unless you refer to specific laptop models or something, but still, this is news to me.

Sorry speed writing, meant Intel.

[blueweb]

TemplarGR Ok, thanks for the details. I was sure the architecture would be different. It is not clear to me if APU actually implies a certain architecture instead of just the notion of CPU + GPU in the same package. And besides, the actual APUs, regardless of original vision, are not homogeneous, but still Zen + Vega, so isn't this just semantics?

[wizard69]

I hope they won't rush it. NO need DDR4 started at pathetic 2133 speeds and was nothing to write about until 3600MHz+ speeds.

Well supposedly DDR5 will be faster when it first hits the market. I'd rather see them rush it to the APU market though because the RAM is lower power and every little bit off bandwidth on AMD's APU's helps.

Aiming it for APUs is fruitless, at least on start. WHole point of APU solution is low price. Which kinda evaporates if you have to have expensive fast RAM.
Once market stabiilizes and 6000MHz speeds become normal, sure. By then you'll probably get 64GB per stick and ECC will be off-the-shelf normal...

The whole point of an APU is integrated graphics leading to more options for designers. APU's especially the next generation could have a huge impact on small form factor machines. Small form factor isn't always about price, at least not often, rather small is at times a better fit for the problem at hand..

BTW, there is far too big of a crowd of me-too morons o'n the board market. All generic copies, with the exception of a bord or two from Asrock and perhap Biostar etc, everything is generic.

I would love to see cheap SoC board with say soldered on 4800HS and soldered on LPDDR4 at 4266MHz quad channel.

The above I agree with 100%! However why not hook up DDR5 to that APU? Sadly the boards you do find with soldered on SoC are ususally special purpose in some manner or another.

ANd the same with a successor on LPDDR5. With all four DP ports on small , cheap but deadly mini-ITX or smaller board, without chipsed. Basically just with that the APU offers.

But no, we are to be forcefeed bazziliion same models with different name.

In many ways I'm agreeing with you here. There are better, more interesting, small form factor, ARM based boards out there compared to anything X86. That might be due to far cheaper SoC offerings but I often get the feeling that the x86 world is stuck in the past! There is far too much focus on legacy support. For the life of me I can't understand why any mini-ITX or similar small form factor board even bothers with SATA any more.

While I'm a little down on Apple when it comes to PC's at the moment; I'm really looking forward to seeing what the shipping Apple Silicon is all about. Anything legacy will likely be abolished on the new machines. Say what you will about Apple but they have the balls to drop old tech far sooner than the PC world every does. We still have PC oriented motherboards shipping with old style keyboard and mouse sockets, which just blows the mind.

[MrMorden]

Aiming it for APUs is fruitless, at least on start. WHole point of APU solution is low price. Which kinda evaporates if you have to have expensive fast RAM.
Once market stabiilizes and 6000MHz speeds become normal, sure. By then you'll probably get 64GB per stick and ECC will be off-the-shelf normal...

I'd certainly like ECC to be standard, but is there a reason to expect that? Intel shows no sign of giving up trying to charge a substantial premium for ECC support, and even more if you want a really high number of cores—say, eight and up. I'd be good with AMD's enabling everywhere and not officially qualifying the bottom-tier parts, but non-EPYC CPUs don't support RDIMMS, and trying to find 32GB UDIMMs is not easy. Maybe IBM can keep pushing down the entry point for POWER, or the ARM vendors will give us something between phone-level and a super-beefy server.

[wizard69]

AM5/Pcie5/ddr5 will be my next upgrade stop I think. We're approaching the point where the slowest part of the boot process is the bios loading.

When I built up my new desktop early this year, it was replacing a rather old unit and a newish laptop. The atop had died and i didn't have the stomach for an old 32 bit desktop machine. I went all AMD and as you know AMD had some issues with video drivers early this year. That resulted in more reboots than desired. Thankfully the reboots where so fast it wasn't the terrible event it would have been years ago. It is kinda funny but also shocking, that hardware has come so far so quick.

[wizard69]

The notion that APUs are destined for the low-cost segment (with all of the constraints that implies) doesn't need to be true in the future anymore.

It hasn't been true for a long while now. The so called laptop SoC that end up in these machines are fully capable of taking on a variety of jobs in industry. With the move to SSD's and multi core low wattage chips, a Small Form Factor machine today will beat the pants off a 5 year old desktop tower machine.

[blacknova]

With the move to SSD's and multi core low wattage chips, a Small Form Factor machine today will beat the pants off a 5 year old desktop tower machine.

I thought so too. In reality 8 year old 45w i7 still run circles around 1 year old 15w i7. Of course the main problem for newer CPU is a ultrabook's cooling solution, but the state remains - new 15w parts dedicated for long battery run are not a replacement for performance oriented solutions.

[milkylainen]

It is spelled HBM - High Bandwidth Memory.

Yes, I know. It was merely a typo, but thanks for correcting me.

HBM is generally a lot faster than DDR. They are basically the same memory, but HBM was designed as a stack of memory chips so one can have a very wide bus and get past the bottleneck. And we need HBM not only for GPUs, but for CPUs when we want to increase the core count. A picture says it all:

So much of a die in an AMD EPYC gets wasted on cache just to keep it fed, the design gets bend over to fit the needs of caches, software needs to be adjusted to take it into account, and the moment every core wants to access main memory the whole performance still goes to shit. All because DDR cannot keep up.

Well. Depends on what you mean by "not keeping up".
I agree that HBM2 provides more margin for even higher data rates and is a technically elegant solution.
HBM2 is a traditional trade:
Low pin count, higher frequency (high power), lower cost (DDR). For higher pin-count, lower frequency (low power) and high cost (HBM2).
Bandwidth is pretty similar within the same cost, as is latency.

The problem with HBM is that it's pincount is one magnitude higher. Hence need for an interposer and a local structure (preferrably).

It'll work fine for high end GPUs but HBM2 is going to be difficult for very large memory sizes with CPUs.
I don't know how you would suggest a modular x86 memory architecture for 64G+ RAM sizes?
The interposer structure would be extremely cumbersome. Cost would go through the roof.
I'm pretty sure it can be done, and cost will eventually drop.
But I don't think it will ever catch plain PCB-mounting DDRs in that aspect.