AMD '#Rekt son' reply to Intel's "four glued together" statement

[Sauron]

14 hours ago, RadiatingLight said:

TBH that ecosystem slide looks very unimpressive.

That's because they didn't use the logos twice like intel did. And for the enterprise, asus, gigabyte, dell and supermicro are all the ecosystem you need.

 

Besides this was more a response to the claim that amd has NO ecosystem somehow and that they have 0 compatibility with anything, which is obviously ridiculous.

[SpaceGhostC2C]

2 hours ago, ArcThanatos said:

AND asrock was on there twice!

Hey. at least they didn't put AMD's logo.

(which I guess they could, because I'm sure FirePro GPUs work fine on Xeon workstations )

 

 

25 minutes ago, XenosTech said:

No ryzen was designed for the server then scaled down... They hit a voltage wall because of GoFlo's process

I think by now we can all agree Ryzen was designed as a bunch of Commodore 64s zip-tied to a PCB.

[XenosTech]

1 minute ago, SpaceGhostC2C said:

I think by now we can all agree Ryzen was designed as a bunch of Commodore 64s zip-tied to a PCB.

Zip ties seem way too elegant... need something cruder

[Brooksie359]

13 hours ago, Hunter259 said:

I already explained why it's worse and you don't deny that it is so I'm confused to why you can't understand that it is worse in a technical standpoint. The product performs excellently and I have never made shots on it's performance. I just said Intel's claim is correct in that it is a "worse" way of achieving it.

From an engineering standpoint what amd has done is elegant. Anything that is engineered can be made better but often times isn't because the cost doesn't warrant it. The fact that AMD found a way to produce cpus with all the practical performance needed for the majority of tasks at much cheaper costs is impressive. It's simply a smarter more practical way of producing cpus than trying to produce incredibly large dies that are very expensive to produce and don't give a huge practical performance difference. I mean you talk about one being more technologically advance but that happens all the time in industry. There are methods that are more advanced but aren't used because they simply aren't practical. No business is going to simply make a better product because it's better. They are going to look at the cost difference and what you get for that cost and I'm pretty positive most companies would go with amd's method after doing the math.

[XenosTech]

1 minute ago, Brooksie359 said:

From an engineering standpoint what amd has done is elegant. Anything that is engineered can be made better but often times isn't because the cost doesn't warrant it. The fact that AMD found a way to produce cpus with all the practical performance needed for the majority of tasks at much cheaper costs is impressive. It's simply a smarter more practical way of producing cpus than trying to produce incredibly large dies that are very expensive to produce and don't give a huge practical performance difference. I mean you talk about one being more technologically advance but that happens all the time in industry. There are methods that are more advanced but aren't used because they simply aren't practical. No business is going to simply make a better product because it's better. They are going to look at the cost difference and what you get for that cost and I'm pretty positive most companies would go with amd's method after doing the math.

Even Nvidia is looking to do something similar in the future with their gpu's... I guess they don't wanna be caught with their pants down like Intel

[Brooksie359]

Just now, XenosTech said:

Even Nvidia is looking to do something similar in the future with their gpu's... I guess they don't wanna be caught with their pants down like Intel

Yeah the fact that amd can produce 32 core cpus with yeilds of quad core cpus is kinda amazing and incredibly smart on their part. I would not be surprised if others followed suit because AMD has shown it can be done without negatively affecting the preformance in the majority of tasks. From a business standpoint this is the type of design AMD needed to be able to compete with Intel and why they can sell an 8 core for 300 dollars.

[XenosTech]

5 minutes ago, Brooksie359 said:

Yeah the fact that amd can produce 32 core cpus with yeilds of quad core cpus is kinda amazing and incredibly smart on their part. I would not be surprised if others followed suit because AMD has shown it can be done without negatively affecting the preformance in the majority of tasks. From a business standpoint this is the type of design AMD needed to be able to compete with Intel and why they can sell an 8 core for 300 dollars.

Now all we have to do is see how the new mainstream 6 cores from intel perform compared to the current 6 cores from x99, x299 and the R5 1600

[Mira Yurizaki]

I think I'm more annoyed about this dude bro back and forth attitude between the companies.

 

I mean, I go look at Microsoft congratulating Sony on releasing things. Maybe it was meant as a half-hearted gesture, but it's certainly better than "LOL UR PS4 'PRO' IS NO MATCH FOR OUR 'TRUE 4K' SCORPION'.

 

Okay maybe Microsoft has taken pot shots at Sony.

[The Sloth]

4 minutes ago, M.Yurizaki said:

Okay maybe Microsoft has taken pot shots at Sony.

sony also has 

 

5 minutes ago, M.Yurizaki said:

I mean, I go look at Microsoft congratulating Sony on releasing things. Maybe it was meant as a half-hearted gesture, but it's certainly better than

look at the switch launch, Microsoft and Sony congratulated Nintendo. 

https://www.polygon.com/2017/3/3/14806574/nintendo-switch-twitter-sony-microsoft-arbys

[Mira Yurizaki]

3 minutes ago, nerdslayer1 said:

sony also has 

 

look at the switch launch, Microsoft and Sony congratulated Nintendo. 

https://www.polygon.com/2017/3/3/14806574/nintendo-switch-twitter-sony-microsoft-arbys

Oh, right, forgot about that.

 

But also Microsoft used to send cakes to Mozilla before they started doing a "release a major version of Firefox every month" thing. And Mozilla sent cakes to Microsoft in return for IE releases.

[MoonSpot]

1 hour ago, SpaceGhostC2C said:

I think by now we can all agree Ryzen was designed as a bunch of Commodore 64s zip-tied to a PCB.

What kind of fracking silver spoon were you born with?  Here in the real world we use twist ties; dem zip-ties are like a ducking buck each!

 

1 hour ago, Brooksie359 said:

...I would not be surprised if others followed suit because AMD has shown it can be done without negatively affecting the preformance in the majority of tasks. From a business standpoint this is the type of design AMD needed to be able to compete with Intel and why they can sell an 8 core for 300 dollars.

Its not a matter of being surprised or not, its a matter of survival or viability from product POV.  Server and big money can make whatever case they want and buy whatever they want, but the HEDT won't shoulder intels costs for long with this sort of offering from AMD.  The more generations intel doesn't adapt the more and more downward momentum they'll pickup (super unlikely to happen, they'd have to be a shade of stupid which we know they're not).  The only question is how many generations is intel committed to having to continue with monolithic dies, cause they can't do a 180 over night.

[Mira Yurizaki]

22 minutes ago, MoonSpot said:

What kind of fracking silver spoon were you born with?  Here in the real world we use twist ties; dem zip-ties are like a ducking buck each!

Meh, I just stack them and hope Gary from accounting doesn't trip over the setup. I picked a random name and position, don't hurt me plz.

22 minutes ago, MoonSpot said:

The only question is how many generations is intel committed to having to continue with monolithic dies, cause they can't do a 180 over night.

Well, they weren't above not doing that. Their first dual core was two single core dies on a package, ditto with their first quad core with two dual cores on a single package. And I want to say AMD took pot shots at them for that too.

 

But hey, as long as they can continue spitting out monolithic dies at their target success rate, they probably won't rethink their strategy.

[Curufinwe_wins]

1 hour ago, M.Yurizaki said:

Meh, I just stack them and hope Gary from accounting doesn't trip over the setup. I picked a random name and position, don't hurt me plz.

Well, they weren't above not doing that. Their first dual core was two single core dies on a package, ditto with their first quad core with two dual cores on a single package. And I want to say AMD took pot shots at them for that too.

 

But hey, as long as they can continue spitting out monolithic dies at their target success rate, they probably won't rethink their strategy.

Yeah I think part of the internal salt was how much AMD took shots at Intel for doing basically the Pentium-era equivalent of what AMD is doing now. 

[MoonSpot]

2 hours ago, M.Yurizaki said:

don't hurt me plz.

Why would I hurt ya?  Your innovative approach saves me a trip to the grocery stores produce isle to borrow twist ties.

 

2 hours ago, M.Yurizaki said:

Well, they weren't above not doing that. Their first dual core was two single core dies on a package, ditto with their first quad core with two dual cores on a single package. And I want to say AMD took pot shots at them for that too.

 

But hey, as long as they can continue spitting out monolithic dies at their target success rate, they probably won't rethink their strategy.

Well, did intel then have to go a full generation without being able to staple cores together?  I seem to remember them catching wind of things before hand, but that was soo long ago I don't really recall.  I suppose they could be aiming to do multi die for their 12/14/16/18 i9s and might be why there's no info(or wasn't any for a stretch)...don't know, I stopped paying attention to intels releases when they cooked up x299.

[Sunshine1868]

On 7/18/2017 at 8:26 PM, RadiatingLight said:

TBH that ecosystem slide looks very unimpressive.

I'm assuming you're looking at this through a gamer's eyes... this is an enterprise product, and those top three names on that slide own a VERY substantial majority of the enterprise compute market share. 

 

so it may be unimpressive to you, a gamer, but when a SysAdmin or CIO sees that AMD has partnered with these companies (which, for many years shuttered their AMD business or at least put it on the back burner), their ears perk up.

Edited July 21, 2017 by Sunshine1868
Fixed spelling of "shuttered"

[Sauron]

6 hours ago, XenosTech said:

Zip ties seem way too elegant... need something cruder

chewing gum

[Eaglerino]

Seems to me Intel will win the 'smack in the face' advertising. If you showed Intel's ad then AMD's melancholy response to a random consumer, they'd probably still pick Intel

[Trik'Stari]

18 hours ago, mr moose said:

 

The cow is not AMD though.   The cow simply represents amusing event, it is how AMD responds to it that determines how much "green" gets eaten. 

AMD is the cow, Intel is the Bluegrass, Nvidia is the green grass.

 

AMD is whipping Intel right now, but Nvidia remains unbothered because Vega is basically smoke and mirrors.

[Juanitology]

22 hours ago, RadiatingLight said:

TBH that ecosystem slide looks very unimpressive.

At least they didn't paste the same partner icon on it multiple times to make it look like there's more, like someone else who I don't want to metion cough-intel-cough.

[Jito463]

2 hours ago, Sunshine1868 said:

but when a SysAdmin or CIO sees that AMD has partnered with these companies (which, for many years shuddered* their AMD business or at least put it on the back burner), their ears perk up.

*shuttered  

_{(sorry, I couldn't resist)}

1 hour ago, Sauron said:

chewing gum

You don't waste good chewing gum!  Use the old stuff.  I'm sure I kept some under the couch.  Just need to soften it up a little.....

[mr moose]

3 hours ago, Trik'Stari said:

AMD is the cow, Intel is the Bluegrass, Nvidia is the green grass.

 

AMD is whipping Intel right now, but Nvidia remains unbothered because Vega is basically smoke and mirrors.

No, the cow is Intel.  Their slide was the equivalent of a cow trying to climb a tree.   My original post was about how AMD responds, and how their fan base responds as well. 

 

Cows aren't very good at climbing tree's, therefore if AMD or their fan base try to educate everyone about why the cow is trying to climb the tree or that they are bad at it (pointing out the obvious) is only going to validate the cow/lessen the hilarity.  Where as sitting back and letting people simply laugh at the cow and realise the better response is more mature PR.

[System Error Message]

23 hours ago, AnonymousGuy said:

 

This is the one.  Doesn't include IF to another chip on the package.  Going up 3.5x in latency depending on your workload is pretty well qualified as "inconsistent performance" as mentioned on the Intel slides.

not sure how i missed this but, each ryzen chip consists of 2 quads put together on the same die, hence the increase in latency, however i expect going to another CPU to be entirely the same latency as long as the controller used is the same controller. Im sure AMD guys already know that the normal way of CPU to CPU communication is "far" so infinity fabric was introduced for direct CPU to CPU communication. Think of it like tilera's mesh except over 4 CPU controllers. The latency is only there because there is a controller to send things through but thats still faster than using the bus/chipset for CPU to CPU like what both brands did in the past for their multi socket goodness.

 

That core i9 shows poor optimisation. Big singular dies of many cores? the latency does vary a little. Also to put things in perspective, logical core 1 is on the same core which seems to be 10ns. On the intel quad core and AMD ryzen, it takes 4x longer to communicate with any other core from the first and is even worse on the other CPUs. Ryzen takes 14x longer for core to core over the other further 4 cores as it goes through the controller (which is why the core2 lacked level 3 cache because of the higher latencies at the time).

 

The most important stat is missing from here, the lga1366 cpu at stock and overclocked as you can overclock the cache, so i wonder if that improves core to core on the 6 core as doing so i noticed huge boosts in physics simulation speeds.

 

140ns? you should see intel's 10Gb/s or better NICs. People make render farms and so on, seems like that higher latency doesnt affect the workload. Lets say you have to perform a multi threaded physics calculation. The code and data will be stored in the highest shared cache which has a much higher latency on any CPU than direct CPU to CPU communications. The cores are asynchronously fed the instructions and data into their pipeline and that pipeline has a delay which is the cycle time of the CPU * length of pipeline. So this graph is a moot point. 

[leadeater]

1 hour ago, System Error Message said:

The cores are asynchronously fed the instructions and data into their pipeline and that pipeline has a delay which is the cycle time of the CPU * length of pipeline. So this graph is a moot point. 

But it makes Ryzen look bad and fuels the optimization argument so just keep posting it every chance you get .

[XenosTech]

2 minutes ago, leadeater said:

But it makes Ryzen look bad and fuels the optimization argument so just keep posting it every chance you get .

Well at least some one gave him an answer, I certainly wasn't going to explain what inconsistent performance was. From that graph also intel's performance looks more inconsistent that's AMD's

[leadeater]

8 minutes ago, XenosTech said:

Well at least some one gave him an answer, I certainly wasn't going to explain what inconsistent performance was. From that graph also intel's performance looks more inconsistent that's AMD's

Well I can never pass up the opportunity to make a terrible joke or pun, I really am a terrible person lol.

 

But yea a graph of inter-core latency is utterly meaningless in the context of workload performance if you don't even know what the workload is and how it interacts with CPU resources. If there is never any data passing between cores or caches, other than from L2 to L1 in the same core, the latency could be 10 seconds and it wouldn't matter (stupid extreme used on purpose).

 

That's also ignoring things like OS processor queue like @System Error Message was talking about at a slightly lower level, hardware wise.