Is Intel ready to get its ass reamed by 7nm EPYC(Rome) in late 2018, what's in the pipeline? Skylake-EP refresh...

Is Intel ready to get its ass reamed by 7nm EPYC(Rome) in late 2018, what's in the pipeline? Skylake-EP refresh? AHAHAHAAH

Other urls found in this thread:

intel.com/content/www/us/en/products/processors/xeon-phi/xeon-phi-processors/7290.html
anandtech.com/show/12032/amd-announces-wider-epyc-availability-and-rocm-17-with-tensorflow-support
rocm.github.io/install.html
twitter.com/SFWRedditImages

>ROCm 2.0 seems to aling with Q4 driver update
I smell win.

>Higher Order Reasoning

IT HAS BEGUN

No, Rome will stomp Intel and kill some server ARM vendors along the way.
Besides, it looks like Intel's datacenter shrink is slipping well into H2 2019.
It's about as bad as you can get.

If only we can import that to the retards.

Retards are not HIP-compatible.

yeah but where is partner vega, magic drivers and Zen+?

It's a datacenter roadmap.

>pcie4
I don't think this roadmap is legit

Why?

Compatibility with Naples.

PCIe is backwards compatible.
Slot Rome into the old board == PCIe3 speeds.
Slot it into the new one == PCIe4 speeds.

>7nm

oh god expect 60% manufactureing failure rate

Good point.

I seriously hope that 64 core with 256MB L3 is true, it would be slaughter.

Shrink?

MOAR COREZ
MOAR CACHE !!!!!1


t. AMD """"advancements""""

So the exact same thing Intel datacenter group does?

>it's happening!

So 40% yield for 16core 64MB L3 die?
Still respetable.
That's exactly what every hyperscaler wants?
Even Amazon will sever its contract with Intel for that.
A die shrink.
It probably is. CPC Hardware is about as valid as you can get.

I find it strange that this is the hill you want to die on
intel.com/content/www/us/en/products/processors/xeon-phi/xeon-phi-processors/7290.html

That's duct taped Silvermont cores with WIIIIIIIIDE FMA units attached to them.
Not exactly comparable.

anandtech.com/show/12032/amd-announces-wider-epyc-availability-and-rocm-17-with-tensorflow-support
Aight, exactly as on roadmap.

Good news? Stock crash incoming

Welcome to $AMD!

So the roadmap is confirmed? Intel is dead.

Yes, they've just delivered TF port mid-Q4.

But where are the Intel shills?

>someone who doesn't directly benefit from this rooting for a company

hmm really makes ya think

>now owning $AMD
?

But where is the AMD enterprise sales?

Awaiting platform.

post stocks fanboy

>AMD also wheeled out the Inventec P47 system that it announced at launch, with a single EPYC processor and four Radeon Instinct MI25 GPUs. In partnership with AMAX, 47 of these systems were put together into a single rack, capable of one PetaFLOP of single precision in a turnkey solution. From today, AMAX is now taking pre-orders for this rack, for delivery in Q1.

Not this time, (((friend))).

The state of ROCm is really funny or pathetic depending on how you look at it. Faster by using both the GPU and CPU, they said. As of right now it's slower than just using the GPU so it's totally pointless. This is why it's funny and laughable - and also pathetic.

Leveraging the combined power of the GPU and CPU is obviously a good idea in principle. And I really hope this goes somewhere. It's just not.. there or anywhere close to it right now.

So I guess that explains where the fucking servers have gone to, not that I expected anything different, fucking cloud niggers.
I had to deal with two sysadmins from 5000+ employee companies getting extremely pissy because they couldn't buy those 24 cores.

i would really appreciate it if you could delet this thread now

You can't because you don't have any. You're a poor consumer who has been corporately groomed to advertise for free.

With Dell championing epyc truly the end times cometh.

That's HSA you mongoloid.
Goddamit did they expect good availabity when AMD secured two niggerscalers day1?

All this using 2U racks and no meme proprietary mezzanine form-factors.

Given it is about as plug and play as you can get at this tier of hardware epyc is by far and away the most important zen derived component AMD currently make.

The ROCm Platform brings a rich foundation to advanced computing by seamlessly integrating the CPU and GPU with the goal of solving real-world problems.

rocm.github.io/install.html

You could... actually read what it is become calling names. You can also try it - right now - if you have the hardware handy.

ROCm is just HSA taken a small step further.

ROCm has about as much to do with HSA as CUDA: fuckall.

> xGMI interconnect for Vega 20 in 2018

fucking finally. I was among the people expecting Vega to do it. PCIe 4.0 confirmation is nice also, even if it likely only means a ~30% bump over the 12.5 Gbps transceivers in Zeppelin.

Expectation of Infinity Fabric 2.0/2.1?

Fuckall unless AMD pokes OSAT partners for advanced packaging.

...

Who will be responsible for Skynet? Intel, Nvidia, AMD, Google? or some combination of those.

IBM

So they'll double IF's socket>socket bandwidth?

Yes.