AMD Earnings

[cj09beira]

also another thing, sequentially amd doubled epyc's shipments, which is great, they also have had good instinct sales, it seems their virtualization tech is quite attractive 

[ZestyJalapeno]

They really still need to penetrate and bust out in the enterprise datacentre market, my company is still buying years old Xeons by the truckloads because Cisco only came out with EPYC-certified rack hardware last year

[Taf the Ghost]

1 minute ago, cj09beira said:

also another thing, sequentially amd doubled epyc's shipments, which is great, they also have had good instinct sales, it seems their virtualization tech is quite attractive 

Maybe, at least until Intel enters the market. 

[Taf the Ghost]

Just now, Dogeystyle said:

They really need to penetrate the enterprise datacentre market, my company is still buying years old Xeons by the truckloads because Cisco only came out with EPYC-certified rack hardware last year

The feeling I get is that Epyc only got hardware after the big players saw what Rome could do. That's when they bought into AMD's roadmap.

[cj09beira]

Just now, Taf the Ghost said:

Maybe, at least until Intel enters the market. 

ya the more i think about it the more i dislike the fact that intel is going to enter the gpu market, they will 1000% sure start abusing their market share and force all intel systems 

[RejZoR]

1 hour ago, leadeater said:

You know looking at that kind of thing is rather pointless right? Even from the same fab that can mean almost nothing, GloFo 12nm vs 14nm for example.

 

Node size numbers now days are about as meaning full as engine displacement, just a number.

There is no replacement for displacement. Still applies and always will.

[cj09beira]

1 minute ago, Taf the Ghost said:

The feeling I get is that Epyc only got hardware after the big players saw what Rome could do. That's when they bought into AMD's roadmap.

in the end epyc 1 even though its quite competitive will mostly be a platform to test before jumping over to rome, aka its a beta program, and i think thats a good thing as it means epyc 2 will be able to ramp harder and with more confidence behind it 

[k.m.p]

it seems that amd is far more stable as a company now compared to before ryzen, they've now had 6 quarters of profitability which hasn't happened in over 10 years, hopefully they keep this going with 7nm.

 

As Linus said in the wan show, as important as your product is, for the datacentre and server market you also need to show consistency, which by showing the 64 core epyc and maintaining backwards compatibility and upgradability until 7nm + they have, hence the growing epyc sales

[cj09beira]

Just now, k.m.p said:

it seems that amd is far more stable as a company now compared to before ryzen, they've now had 6 quarters of profitability which hasn't happened in over 10 years, hopefully they keep this going with 7nm.

 

As Linus said in the wan show, as important as your product is, for the datacentre and server market you also need to show consistency, which by showing the 64 core epyc and maintaining backwards compatibility and upgradability until 7nm + they have, hence the growing epyc sales

they have focused quite a bit into being consistent as much as possible, they are announcing the products for servers every 2 years, around december, every year around march for consumers, they also seem to have shared their plans with some of the bigger guys, having super computers already planed to use Milan is a pretty huge deal and means those that are seeing such plans are liking them quite a bit

[k.m.p]

15 minutes ago, cj09beira said:

they have focused quite a bit into being consistent as much as possible, they are announcing the products for servers every 2 years, around december, every year around march for consumers, they also seem to have shared their plans with some of the bigger guys, having super computers already planed to use Milan is a pretty huge deal and means those that are seeing such plans are liking them quite a bit

What do you mean by use milan?

I think a big part of them being consistent is having a architecture that can be built and improved upon, and very scalable to reach many markets ie the same dye in epyc as ryzen 3 

[cj09beira]

Just now, k.m.p said:

What do you mean by use milan?

I think a big part of them being consistent is having a architecture that can be built and improved upon, and very scalable to reach many markets ie the same dye in epyc as ryzen 3 

milan is amd's 3 version of zen, its its code name for the server market, (naples=zen1=epyc1, rome=zen2=epyc2)

ya they having chiplets will be quite good, it allows for lower costs, entering new nodes sooner and  as they start becoming better and better at binning chips and volume increases they will be able to sell higher clock parts with the same silicon just thanks to the better binning, though this will also mean that lower cost parts will also more consistently overclock worse than more expensive parts. 

[Doobeedoo]

As expected, it's going well for them so that's nice. This year will definitely be fun to see with Zen2 and Navi. 

[Stefan Payne]

5 hours ago, leadeater said:

You know looking at that kind of thing is rather pointless right? Even from the same fab that can mean almost nothing, GloFo 12nm vs 14nm for example.

 

Node size numbers now days are about as meaning full as engine displacement, just a number.

While it is true and you can argue that the process AMD used before is somewhat worse than Intel's, the new 7nm one is a real shrink and (fat) superior to what Intel has.

 

From what we know, the power consumption of Ryzen2 is almost halve that of Ryzen1 at the same clockspeed.

 

So its like comparing Intel 250nm to AMD 180nm or Intel 180nm to AMD 130nm. Its not even close for the CPU Core...

[Stefan Payne]

4 hours ago, Taf the Ghost said:

The feeling I get is that Epyc only got hardware after the big players saw what Rome could do. That's when they bought into AMD's roadmap.

I agree with that.

While Epyc was always pretty neat, the Rome was the thing that really took off and convinced people to go for the AMD Ecosystem.

Even, as I said earlier, Intel has Problems with their 10nm process...

AMD might have known that and went for a chiplet design to circumvent the Problems that the smaller sub 12nm node has...

4 hours ago, Dogeystyle said:

They really still need to penetrate and bust out in the enterprise datacentre market

Won't be happening too soon as many people responsible for the IT Department are just totally incompetent and don't know much about the Hardware that is around and/or only buy single source for whatever reason.

 

Louis Rossmann's competitor in the Youtube area (Eli), talked about that a couple of times and also only talks about "Intel"; "Intel" and "Intel", maybe nVidia. But rarely AMD...

 

So how can AMD gain a decent share in such an enviroment?

 

The only ones they can get are the biggest of the biggest, meaning Microsoft and Amazon for their Webservice, maybe some others and maybe some universitys.

 

The "normal" IT Departments in "08/15 Companys around the Corner" is not something they can get into...

 

Wich is why AMD should (and IIRC is) concentrate their efforts on "developing markets", where IT Departments and such things are new and they are open to everything they is...

That is not the Case in the US for the most part.

 

In Europe, especially Germany, it looks different. But Germany was always an AMD Stronghold...

[ZestyJalapeno]

13 minutes ago, Stefan Payne said:

 

Won't be happening too soon as many people responsible for the IT Department are just totally incompetent and don't know much about the Hardware that is around and/or only buy single source for whatever reason.

Not necessarily incompetent, just extremely conservative. AMD EPYC server ecosystems has existed in the marketplace for what? 1.75 years? It's pretty much a baby. Even if AMD can empirically prove that the hardware-software package is capable of 24x7x365 100.00% stability across whatever environment you deploy it in, it'll take years to rebuild that reputation.

[thorhammerz]

1 hour ago, Dogeystyle said:

Not necessarily incompetent, just extremely conservative. 

Shhhhh, didn't you get the memo?

 

Anyone found not offering their body and soul to our new overlord AMD and their new toys will be branded incompetent a heretic and exocommunicated .

[THraShArD]

I hate owning this stock. It's soo stressful.  Same with Nokia, both are so manipulated.

 

AMD is back where it was last week after being shorted to death

[mr moose]

9 hours ago, cj09beira said:

there is no replacement for displacement  

Electrons. 

 

 

[cj09beira]

6 minutes ago, mr moose said:

Electrons. 

 

 

no fair  

[leadeater]

4 hours ago, Stefan Payne said:

While it is true and you can argue that the process AMD used before is somewhat worse than Intel's, the new 7nm one is a real shrink and (fat) superior to what Intel has.

Superior how? We have like one proper demo under controlled conditions of only a few applications and it's only matching Intel 14nm not beating it by significant amount performance wise. Power efficiency is a fair bit better it seems but we have no idea how that scales, could be better at 4-4.5 Ghz then rapidly increases to worse than Intel as you approach 5GHz.

 

If you ignore the node names and sizes and purely look at the product which is better? Once you can actually judge if the product is actually better then you can start to try and make an assessment about which process is better, even then you'll have a hard time with that because the architecture has a big impact.

[leadeater]

4 hours ago, Dogeystyle said:

Not necessarily incompetent, just extremely conservative. AMD EPYC server ecosystems has existed in the marketplace for what? 1.75 years? It's pretty much a baby. Even if AMD can empirically prove that the hardware-software package is capable of 24x7x365 100.00% stability across whatever environment you deploy it in, it'll take years to rebuild that reputation.

And that's even if the vendor/product you are buying even has an AMD EPYC option. You can't just white box your own thing and add that to a controlled product stacked offering, that's typically impossible. If it's not supported it's not an option, if it's not an option then it's not an option. Businesses are buying less of straight servers and buying more hyper-converged systems where you can only buy from their list of options, none of them I'm aware of have AMD.

[Taf the Ghost]

2 hours ago, leadeater said:

And that's even if the vendor/product you are buying even has an AMD EPYC option. You can't just white box your own thing and add that to a controlled product stacked offering, that's typically impossible. If it's not supported it's not an option, if it's not an option then it's not an option. Businesses are buying less of straight servers and buying more hyper-converged systems where you can only buy from their list of options, none of them I'm aware of have AMD.

Lisa Su drops "accelerators" a lot when talking Epyc, and @leadeater point explains why. So many things are kind of turn-key setups from a SI. Google and a few of the big Chinese companies are the only ones just buying rack-mount systems that install all of their own customer software solutions.

[THraShArD]

This is great stuff here

 

AMD no longer has to pay millions in royalties to make 7nm CPUs and GPUs

 

https://www.pcgamesn.com/amd/amd-globalfoundries-wafer-supply-agreement

[AnonymousGuy]

It's all fine and good until you look at Intel's numbers and how across the board it's 12x more than AMD's numbers.   "oh you made 6B in 2018...cute, we made $75B".  To put another way: Intel sells close in SSDs alone what AMD does in their *entire* business.  If AMD makes no missteps (which is largely impossible since their business is dependent on TSMC) then at most they are an annoyance to Intel, not even real competition.

[cj09beira]

17 minutes ago, AnonymousGuy said:

It's all fine and good until you look at Intel's numbers and how across the board it's 12x more than AMD's numbers.   "oh you made 6B in 2018...cute, we made $75B".  To put another way: Intel sells close in SSDs alone what AMD does in their *entire* business.  If AMD makes no missteps (which is largely impossible since their business is dependent on TSMC) then at most they are an annoyance to Intel, not even real competition.

with zen 2, market share will start increasing faster, and hopefully oems will start supporting amd more (doubt it, at least in the consumer side), after zen 2 launches amd should have at least 6 months where they are by far the best option, then we will see how intel responds.

due to them selling more volume i dont expect them to reduce prices much if at all as it would make them loose more money, but at the same time it will make amd easily the better buy, will be interesting to watch