AMD Zeppelin codename confirmed and perhaps 32 cores per socket

[csbin]

http://dresdenboy.blogspot.com/2016/02/amd-zeppelin-cpu-codename-confirmed-by.html


+ core_complex_id = (apicid & ((1 << c->x86_coreid_bits) - 1)) >> 3;
+ per_cpu(cpu_llc_id, cpu) = (socket_id << 3) | core_complex_id;






Skylake Xeon E7 28 cores


------------

update

In a discussion related to technology and market trends for the data center, Liviu Valsan (computing engineer at CERN) showcased slides which detailed AMD’s upcoming Opteron processors that feature their x86 Zen core architecture. The session not only provided details on AMD’s next generation processors but also mentioned Intel’s Broadwell-EP chips along with enterprise aimed storage and memory solutions.

 

[Arachnotronic]

Oh look another hit and run post from csbin showing how AMD is awesome!!


Threadcrapping and trolling are not allowed
Markfw900

 

[Boze]

Oh look another hit and run post from csbin showing how AMD is awesome!!

This doesn't prove AMD is awesome.

32 crappy cores still makes a crappy processor. Show me benchmarks.

Deeds. Not words.

 

[master_shake_]

server processor clocked at 2.6 ghz?

 

[MongGrel]

server processor clocked at 2.6 ghz?

Couple of old X5650's on a x58 dual CPU mobo would still blow it out of the water I bet.

But gopher it I guess.

 

[Bubbleawsome]

This doesn't prove AMD is awesome.

32 crappy cores still makes a crappy processor. Show me benchmarks.

Deeds. Not words.

Mass raytracing rendering. I'd love to get a hold on lots of cores for cheap.

 

[DrMrLordX]

Nice to see a name of a future product, even if the speculation about core count seems a bit iffy. It doesn't look like anything aimed at the consumer/enthusiast market anyway.

 

[Dresdenboy]

This doesn't prove AMD is awesome.

32 crappy cores still makes a crappy processor. Show me benchmarks.

Deeds. Not words.

They will clock at 1 GHz and suck 500W at IPCs lower than that of BD. I guess, you didn't read the REEL paper?

Couple of old X5650's on a x58 dual CPU mobo would still blow it out of the water I bet.

But gopher it I guess.

Sureley they will. I'd suggest shorting $AMD based on that.

Nice to see a name of a future product, even if the speculation about core count seems a bit iffy. It doesn't look like anything aimed at the consumer/enthusiast market anyway.

If the former leaked slides are real, then such a CPU with 4xDDR4 channels, HBM, Greenland GPU surely isn't aimed at HEDT. AM4 simply wouldn't support it.

 

[itsmydamnation]

Oh look another hit and run post from csbin showing how AMD is awesome!!

good to see you can play the ball...... oh wait.

 

[Sweepr]

Oh look another hit and run post from csbin showing how AMD is awesome!!

He forgot the winky emoticon this time.
Can you show us where does it say only Xeon E7 gets (up to) 28 cores csbin? Assuming you're not a bot.

Hopefully AMD is not throwing more cores to compensate for lower IPC/clocks.

 

[Dresdenboy]

He forgot the winky emoticon this time.
Can you show us where does it say only Xeon E7 gets (up to) 28 cores csbin? Assuming you're not a bot.

Hopefully AMD is not throwing more cores to compensate for lower IPC/clocks.

It might also be the case, that AMD wouldn't offer this max. amount of cores before 2018.

And the patch just shows, what the Linux kernel is getting prepared for, no roadmap SKU.

For some server tasks, memory hierarchy is more important than an avg. IPC >1.

 

[ShintaiDK]

Oh look another hit and run post from csbin showing how AMD is awesome!!

And based entirely on a rumour via a blog in the dreams of beating Intel.

 

[ShintaiDK]

He forgot the winky emoticon this time.
Can you show us where does it say only Xeon E7 gets (up to) 28 cores csbin? Assuming you're not a bot.

I dont think we will get an answer from someone who created more threads than posts.

 

[Dresdenboy]

And based entirely on a rumour via a blog in the dreams of beating Intel.

My blog entry clearly doesn't contain any signs of emotions. I think, you're projecting the inverse of your world view on pure data and statements. But you still behave exactly like expected.

 

[ShintaiDK]

My blog entry clearly doesn't contain any signs of emotions. I think, you're projecting the inverse of your world view on pure data and statements. But you still behave exactly like expected.

I was talking about csbin. But you obviously felt hit for some reason...

 

[coercitiv]

Oh look another hit and run post from csbin showing how AMD is awesome!!

Assuming you're not a bot.

I dont think we will get an answer from someone who created more threads than posts.

Thank you for your contribution. It makes the conversation significantly more fun and productive.

 

[Dresdenboy]

I was talking about csbin. But you obviously felt hit for some reason...

The OP doesn't deliver much information or opinion so that I've thrown away that hypothesis.

And never forget :
https://i.imgur.com/YuExZ.jpg

 

[ShintaiDK]

The OP doesn't deliver much information or opinion so that I've thrown away that hypothesis.

And never forget :
https://i.imgur.com/YuExZ.jpg

Just see the threads created.

That picture never gets old

 

[NTMBK]

Oh, the humanity!

 

[dark zero]

Seems that the trolls didn't got a sleep and managed to bash a paper from AMD only because they presents something that most might be true or false, but can defeat Intel on something.

More information are needed. I am skeptical about that. 32 cores at max. On dual socket. Seems that AMD is not going the highest end.

 

[NTMBK]

If AMD aren't going for big wide SIMD vectors on Zen, I can easily believe a 32 core server CPU. SKX is cramming 28 cores with 512-bit vector units (and presumably 512-bit data paths) into a socket, so 32 cores with 128-bit vector units seems plausible (even with the slightly inferior manufacturing process).

 

[NostaSeronx]

More information are needed. I am skeptical about that. 32 cores at max. On dual socket. Seems that AMD is not going the highest end.

It is only single socket.

Zeppelin = 8 cores per die * 4 dies on interposer * 1 socket => 32 cores / 64 threads.

Dual Socket = 8 cores per die * 4 dies on interposer * 2 sockets => 64 cores / 128 threads.

Not like Zep will ever release. 22FDX is pretty much being shoved down AMD's throats by GlobalFoundries in the closed off meetings. (They didn't do STM's 28nm FDSOI, and they wont do Samsung's FinFETs.)

 

[Dresdenboy]

Just see the threads created.

That picture never gets old

Maybe he tries to provoke you, knowing that you might jump on it.

Regarding the pic: what would happen, if the promised IPC jump appears, moar coars are added to a socket, and clocks at a typical socket TDP are at least at Abu Dhabi levels?

 

[ShintaiDK]

Regarding the pic: what would happen, if the promised IPC jump appears, moar coars are added to a socket, and clocks at a typical socket TDP are at least at Abu Dhabi levels?

If that was the case, AMD wouldn't keep it a secret and AMDs stock would explode.

Hypothetical scenarios stays hypothetical

 

[Sweepr]

32-cores to adress 80% of the server market, interesting.

As for Zen’s performance, Lisa Su stated that the CPU is expected to address an estimated 80% of the server market.

They probably don't expect it to beat the fastest Skylake-EP/EX (up to 28-cores).