In its current M.2 state it seems pretty useless, but it seems really promising when it comes in DRAM form factors. You could have 8-16GB of 'fast' DDR4 memory, a larger pool of Optane (like 64-128GB), and then an SSD to load from.
Are there any videos, with conclusions, with a setup like that?
Angel Stewart
The DRAM form factor Optane doesn't exist yet, it's just road mapped. The M.2 drives create a good performance bump for mechanical drives, but that's all we've really seen so far.
It's honestly a concern that Intel won't license Optane support to AMD if it becomes a viable solution.
Cameron Price
What is the point of optane?
David Taylor
In its current state, it's an accelerator SSD that can be used to improve performance. It's likely this technology was effectively beta trialed in Apple computers, like much of Intel's tech is, so this solution does actually provide speed increases.
Chase Wood
can we ryzen up?
Brandon Carter
>I forgot that the 900p exists
Gabriel Garcia
Like Xeon, coal is still great
Adam Foster
optane works on AMD though look at the level1techs video
Ayden Ross
I'm not seeing how PCI-E/M.2 is DRAM form factor, but you do you.
Neat, we'll see if the DRAM form factor Optane does. I have a feeling if it gets tight, Intel won't let it work on AMD boards.
Jacob Evans
yeah, for barbecue
Alexander Hill
> won't license Optane support to AMD As far as I understand, it is only possible to "license" a boot UEFI module for Optane, otherwise it will be shown as a block device and therefore usable.
Evan Gutierrez
You can assign it to different drives now. No idea if it works with RAID or something like ZFS though.
Noah Scott
there's the cache optane, which are those 32gb m.2 drives, the 900p, which is a real storage drive, and the upcoming DRAM one
Sebastian Martin
It improves the part where SSD has been somewhat stagnant, being the random reads and writes. Also doesn't slow down at all no matter how full it is. So it's a pretty nice product, but it's retarded expensive for what it brings to the table. If it was 50% cheaper then it would be great.
Kayden Myers
next gen optane is going to be nice if intel keeps up
Wyatt Murphy
>when it comes in DRAM form factors except it doesn't
>at close to DRAM speeds Nobody even promised you that That's what I figured >make the new nand >uhh guys look we made this cool nexgen memory, come see it! >it's actually less bandwidth for 4x the price >uhh guys it's not actually an SSD, please buy our optane buffer to boost your hdd performance! >it's actually and SSD buffer for 4x the price >but what if you actually stick it into DRAM slot? >huh is it real? can we geat cheap terabytes of RAM? >yes goyim please buy our brand new xeon lineup that support sticking fucking SSDs into DRAM slots I mean, there is no way it will be even remotely useful if CPU is not aware of it. Most likely it won't even be connected to CPU but will be managed by chipset. It will be hell for programmers to utilize it. In fact, there's nothing stopping you from making 1 TB of swap on an nvme ssd right now (or even optane ssd)
Carson Martin
I didn't know people could make out a 6 microsecond latency difference
Levi Lee
6 microseconds is a lot actually, modern CPUs work on sub-nanosecond granularity.
Personally, I think it'll be a pretty impressive way to improve performance across an entire system. Contrary to I'm not seeing why Intel would make DIMM Optane not recognized by UEFI. Seems like one of the first things they'd do.
Nathan Long
I agree.
>Seems like one of the first things they'd do. It's the next logical step at least. Making an NVMe controller using 3D XPoint was pretty low hanging fruit, so that's why they did that first.
I wonder how much the next gen 3dxpoint memory will increase performance if they manage to make optane be half as fast as DRAM and not cost a fortune samsung would be in deep shit
Christian Walker
> Intel produces its own ram > teaming up to create their own dedicated GPU's > focus on AI How likely is it that we'll have a consumer GPU from Intel by 2025?
Dylan Sullivan
As long as they don't insist on making it x86 based like with the Xeon Phi, pretty good I think.
David King
>Making an NVMe controller using 3D XPoint was pretty low hanging fruit, so that's why they did that first. For sure, it's also a really safe option to try the technology outside of labs.
Agreed, and it'd be about fucking time this DRAM cartel got overthrown. No way should DDR4 cost the sort of money it does.
Realistically, even projecting the 900p's 2GB/s onto the future situation, what kind of performance hit would that even have for a modern system? Unless you're transferring huge amounts of data to the buffer, I can't see it slowing down every day tasks in any way, shape, or form.