The intel here shows that something is missing - New Cooper Lake Xeon processor + PCIe 4.0 SSD that doesn't support it

[williamcll]

Intel recently unveiled their new (still 14nm) line of Xeon scalable processors, supporting up to 28 cores with hyper threading and an even higher memory capacity

Quote

Intel today introduced its 3rd Gen Intel® Xeon® Scalable processors and additions to its hardware and software AI portfolio, enabling customers to accelerate the development and use of artificial intelligence (AI) and analytics workloads running in data center, network and intelligent-edge environments. As the industry’s first mainstream server processor with built-in bfloat16 support, Intel’s new 3rd Gen Xeon Scalable processors makes AI inference and training more widely deployable on general-purpose CPUs for applications that include image classification, recommendation engines, speech recognition and language modeling.

At the same time, they also debuted their new TLC enterprise SSD, which coincides with rumors that the company was working on PCIe 4.0 drives. It should be noted that even the announced processors cannot utilize this speeds.

Quote

Intel has yet to launch their first CPUs supporting PCIe 4.0, but other parts of the business are keeping pace with the transition: network controllers, FPGAs, and starting today, SSDs. The first PCIe 4.0 SSDs from Intel are based on their 96-layer 3D TLC NAND flash memory, slotting into Intel's product line just below Optane products and serving as Intel's top tier of flash-based SSDs. The two new SSD product lines are the first to fall into the D7 tier under the new naming scheme adopted by Intel in 2018. The P5500 and P5600 are closely-related products that differ primarily in their overprovisioning ratios and consequently their usable capacities, write speed and write endurance. The P5500 is the 1 drive write per day (DWPD) lineup with capacities ranging from 1.92 TB up to 7.68 TB, while the P5600 is the 3 DWPD tier with capacities from 1.6 TB to 6.4 TB. These serve as the successors to the P4510 and P4610 Cliffdale Refresh drives, and as such we expect some follow-on models to introduce the EDSFF form factor options and QLC-based drives that are still due for an update.

Source: https://videocardz.com/press-release/intel-announces-cooper-lake-3rd-gen-xeon-scalable-processors

https://www.anandtech.com/show/15860/intel-announces-d7-series-pcie-40-enterprise-ssds
Thoughts: While the SSD numbers are impressive. the New Float operation functions of the xeon seems very niche and only custom made programs can run it. Can't wait for the Linus review of the chip (if ever).

[Haro]

Cooper lake?

At least it's not nuke lake or something 

Thought they were on to something

[Arika]

10 minutes ago, TofuHaroto said:

nuke lake

I think i went there in one of the Fallout games

[5x5]

When your product is so advanced it's only supported on the competitions platform

[RejZoR]

Still 28 cores? Haven't they had 28 cores (56 threads) as max for several years now?

[porina]

55 minutes ago, williamcll said:

the New Float operation functions of the xeon seems very niche and only custom made programs can run it. Can't wait for the Linus review of the chip (if ever).

This isn't a "mainstream" server CPU so outside of the new data type, there isn't really anything to make you choose this. However, that new data type is a big deal for the machine learning bunch. Call it a niche if you want, it's a BIG area in itself. Now I'm not terribly interested in it myself, so I don't know in what use cases you'd want a CPU to do this work, as opposed to say GPUs. Maybe there is something about data sets that makes one or the other more optimal. As far as I'm aware, AMD CPUs don't have anything like this, so they would be incredibly inefficient for this type of work. They can do it, you can throw more cores at it, but use CPUs for their strengths.

 

For more, see https://www.anandtech.com/show/15631/intels-cooper-lake-plans-the-chip-that-wasnt-meant-to-exist-dies-for-you

 

The "mainstream" server CPU refresh lies with Ice Lake due later this year.

[Drama Lama]

Well next gen intel processors are going to have PCIe 4.0 an Intel wants at least to get money with their SSDs

[SpikeSpiegel]

Cooper Lake....Is it because of Sheldon Cooper?

 

 

[leadeater]

4 hours ago, porina said:

so I don't know in what use cases you'd want a CPU to do this work, as opposed to say GPUs

When your dataset is larger than any GPU VRAM on the market, that's the most common reason anyway.

[Dabombinable]

Unless Intel releases an adapter that changes 8x gen 3.0 lanes to 4x gen 4.0, this will remain solely usable on AMD's platforms.

[Senzelian]

7 hours ago, huilun02 said:

3300X + B550 mobo = unlocked platform with PCIE gen4 readily available for $275

 

Intel?

Intel:
Nananaanana, can't hear you!

[Haro]

1 hour ago, Senzelian said:

Intel:
Nananaanana, can't hear you!

If I don't see it 

It doesn't exist

[TheRandomness]

- Thread cleaned.

[Loote]

On 6/20/2020 at 11:45 PM, Dabombinable said:

Unless Intel releases an adapter that changes 8x gen 3.0 lanes to 4x gen 4.0, this will remain solely usable on AMD's platforms.

In many real applications gen 3.0 would not bottleneck this, the write speed is just barely over what it can do and as long as you read small random files it's going to be slower too. I'd be interested in seeing a comparison.

[mr moose]

We heard Intel decided it wasn't worth the R+D time to put pcie4 on their CPU's at the start of the year.  It should not be news to anyone that they would continue to develop it on other products even though they don't want it until their next round of CPU designs.

 

 

[vinit6694skr]

On 6/20/2020 at 3:16 PM, williamcll said:

Intel recently unveiled their new (still 14nm) line of Xeon scalable processors, supporting up to 28 cores with hyper threading and an even higher memory capacity

At the same time, they also debuted their new TLC enterprise SSD, which coincides with rumors that the company was working on PCIe 4.0 drives. It should be noted that even the announced processors cannot utilize this speeds.

Source: https://videocardz.com/press-release/intel-announces-cooper-lake-3rd-gen-xeon-scalable-processors

https://www.anandtech.com/show/15860/intel-announces-d7-series-pcie-40-enterprise-ssds
Thoughts: While the SSD numbers are impressive. the New Float operation functions of the xeon seems very niche and only custom made programs can run it. Can't wait for the Linus review of the chip (if ever).

Still 14nm ???

And still upto 28 cores ????

[AndreiArgeanu]

What's even the point any more, it just looks like the same old last gen xeons repackaged, perhaps with tweaked clocks. Compared to Threadripper and Epyc these Intel cpu's look weak. Like the 3990X, a cpu that's about 1/3 of the price of a single 8380HL and is "enthusiast" cpu could probably compete with dual socketed 8380HL and beat it easily.

I find it hard to believe that Intel has anything when it comes to performance and price compared to AMD's epyc. CPU's

[porina]

3 minutes ago, AndreiArgeanu said:

What's even the point any more, it just looks like the same old last gen xeons repackaged, perhaps with tweaked clocks. Compared to Threadripper and Epyc these Intel cpu's look weak. Like the 3990X, a cpu that's about 1/3 of the price of a single 8380HL and is "enthusiast" cpu could probably compete with dual socketed 8380HL and beat it easily.

I find it hard to believe that Intel has anything when it comes to performance and price compared to AMD's epyc. CPU's

You're totally missing the point. These are specifically designed for machine learning applications. AMD have nothing to go directly against these. Use the right product for the right use case.

[AndreiArgeanu]

2 minutes ago, porina said:

You're totally missing the point. These are specifically designed for machine learning applications. AMD have nothing to go directly against these. Use the right product for the right use case.

But AMD has deep learning and machine integration solutions?? Like what is it, that these cpu's purposely made for machine learning applications have that a standard cpu doesn't have? Is there something that I'm missing? Because AMD does have a page on Machine and Deep Learning Solutions that are equipped with Epyc

[porina]

4 minutes ago, AndreiArgeanu said:

But AMD has deep learning and machine integration solutions?? Like what is it, that these cpu's purposely made for machine learning applications have that a standard cpu doesn't have? Is there something that I'm missing? Because AMD does have a page on Machine and Deep Learning Solutions that are equipped with Epyc

I mentioned it already in the thread. Kinda annoying it isn't included in OP which leads to ignorance.

 

https://www.anandtech.com/show/15631/intels-cooper-lake-plans-the-chip-that-wasnt-meant-to-exist-dies-for-you

 

[AndreiArgeanu]

2 minutes ago, porina said:

I mentioned it already in the thread. Kinda annoying it isn't included in OP which leads to ignorance.

 

https://www.anandtech.com/show/15631/intels-cooper-lake-plans-the-chip-that-wasnt-meant-to-exist-dies-for-you

 

So the cpu is better at performing instructions that are commonly used in deep and machine learning applications? So technically unless you're using a server for deep and machine learning, AMD is still the leader when it comes to performance in other areas, that I can't really think of, and price to performance. Also is that why the price per cpu is like 13000$?

[leadeater]

26 minutes ago, AndreiArgeanu said:

So the cpu is better at performing instructions that are commonly used in deep and machine learning applications? So technically unless you're using a server for deep and machine learning, AMD is still the leader when it comes to performance in other areas, that I can't really think of, and price to performance. Also is that why the price per cpu is like 13000$?

Accelerated instruction sets give massive performance increases and additionally can make coding easier or in some cases actually make things feasible to do which were not before. Also in the designed system configurations these are for AMD will be far behind in performance even on feature parity instruction sets, EPYC is limited to 2 sockets where you can have up to 8 with these. 128 EPYC2 cores with 16 memory channels is not going to come close to 224 cores with 48 memory channels.

 

Additionally nobody pays list price, unless you are for some weird reason buying through a retail channel and just the CPU but that's so extremely edge case it's not even worth talking about. Some of the servers we have purchased cost the same or less than just the total combined list price of the CPUs in them but that's for the entire server including ram, storage, PSUs, the CPUs etc. $13000 is probably more like $6000-$8000 depending on your or your suppliers volume discount, pricing is based off of how much you buy and the more you buy the higher the % discount you get.

[Franck]

On 6/20/2020 at 6:23 AM, RejZoR said:

Still 28 cores? Haven't they had 28 cores (56 threads) as max for several years now?

Yes they have, for at least 3-4 years if i recall correctly.

 

On Hyper-V the 28 core scalable is pretty popular. Quite a good price / core / support. You just get 2 dual socket board and you get 112 cores 224 threads. This doesn't take much space and it give you the possibility of a good 30 VM running. This is more than enough for most businesses.