AUDIBLE FUCKING KEK INTEL

AUDIBLE FUCKING KEK INTEL

The Platinum 8180 costs $12000 a pop and it's getting stomped by a $4000 EPYC

Other urls found in this thread:

chiru.no/u/average_intel_owner.webm
thinkmate.com/inside/articles/amd-epyc-7000
hothardware.com/news/critical-flaw-in-intel-skylake-and-kaby-lake-hyperthreading-discovered-requiring-bios-microcode-fix
supermicro.nl/products/nfo/AMD_SP3.cfm?pg=SS
twitter.com/AMDServer
vocaroo.com/i/s1p98Qh2if6B
en.wikipedia.org/wiki/Pentium_F00F_bug
twitter.com/SFWRedditVideos

ok

YAMEROOOO

Also it was already obvious but memory bandwidth AMD demolishes.

...

DELET

Intel is stable

El ohe l

Are they? According to who? Because Intel has been fucking up a lot lately

E7 Xeons scale up to 8 sockets while EPYC is only 2 socket max.

yes, their Temps are stable @100°C

NOOOOOOOOOOOOOOOOOOO

And their scaling per socket is shitty for anything that needs to cross the QPI links.

>MOAR SOCKETS!
lmao. This needs a spurdo edit.

oh no. more useless 8p servers.

chiru.no/u/average_intel_owner.webm

ryzen already punted for the server market which is where the cake's at, explains the rushed release of x299 and dmg control

workstation epyc fucking when

thinkmate.com/inside/articles/amd-epyc-7000

LEL LEL LEL

hothardware.com/news/critical-flaw-in-intel-skylake-and-kaby-lake-hyperthreading-discovered-requiring-bios-microcode-fix

Those are racks, OEMs haven't released EPYC towers yet, supermicro has them upcoming.

supermicro.nl/products/nfo/AMD_SP3.cfm?pg=SS

It's absolutely nothing, retard

KEK
Being stable is easy with eyes closed

Do you compile OCaml on model 78 or 94, stepping 3? No?

Tweet to AMD here twitter.com/AMDServer I already sent them one asking them to make a workstation motherboard with decent power VRM and the ability to set the CPU multiplier. They're kind of clueless about the workstation market, they don't even understand people would buy 32 core epyc workstation if they provide a solution.

hmm, to buy a single intel cpu or 2 complete epyc systems?

hahahahahahahahahahahahahaha *inhales* hahahahahahahahahahahahahahahahahahahahahahahahahaha

Like the bricked atoms? freezing skylakes? disabled TSX? sata corruptions? usb sleep bugs? 2MB bugs?

Intel inside, errata outside

also FDIV bug

vocaroo.com/i/s1p98Qh2if6B

en.wikipedia.org/wiki/Pentium_F00F_bug

People seem to forget the F00F one

...

Harry, can you do an inventory on whether our company ran any business critical processes where the compiled code ran short loops of less than 64 instructions and used the AH, BH, CH or DH registers as well or their corresponding wider register (e.g. RAX, EAX or AX for AH) on a Skylake or Kaby Lake platform?

If we did then we might have silently gotten faulty data without ever knowing. Please report back on monday, thanks!

No problem! The patch was out in April, and it would have crashed the system when compiling, not months later, boss!
There is no difference between you and the fags calling out ryzen for having to have patches to get the memory working properly.

>The patch was out in April
Not on all steppings. No fixes for most the kaby lake motherboards.

It will be resolved. I'm old enough to remember windows 95 was a no-go on k6.

>No problem! The patch was out in April
Sorry, that's not satisfactory. The management and the whole analytics department has been using HP Skylakes since last summer. I need you to go back and verify ALL the data since then.
>it would have crashed the system when compiling, not months later, boss!
I have short patience with my employees lying to me Harry. I'm this close to firing you now.

Sorry to put all of this in your lap Harry, maybe if Intel would have notified us of the potential magnitude of this problem instead of trying to sweep it under the rug and hope no one noticed... Trust is a fragile thing Harry, remember that.

It's a good thing we picked Intel over opterons, that hypervisor bug would have bankrupted us, or that vesa local bus bug on piledriver, or the agp issues with athalons. Quit kidding yourself, boss.

>athalons

Harry, I'm tired of your excuses for not delivering. You've become a liability so I'm moving you to the basement and we're reclaiming your red stapler too. In addition I'm moving everything to Power8 + Hurd.

Hey boss, can I have Harry's red stapler? Also, I think we should be making up migration plans to Temple OS.

I like the cut of your jib. Let's order 50 Talos workstations.

Sure, make sure the TPS reports evaluting the move to Power8 + Hurd + TempleOS are on my desk by monday, and the stapler is yours. You will find that the migration will happen and it will cost no more than $1 million. Thanks.

I'm leaking the pics of you and Mike at the company retreat "playing around" in your fursuits on the pool table.

so the flagship Epyc literally beats the best silicon Intel has, bar none. that's kinda fucked up.

can't wait until GloFo moves to 7nm and Zen+ clocks reach KBL levels. shit's going to get ridiculous.

OY VEY
THIS CANT BE HAPPENING
DELET THIS GOY

excuse me i dont know much about servers but why do you need to get four intel cpus to beat two amd cpus

Intel's CPUs scale like shit. Their multi-CPU setups scale even worse.

As if there was any other option

Intel has just been caught with their pants down- they have an old design that they've been iteratively improving and then AMD came around with a brand new design.

There is now.

The interconnect between the processors is inferior so performance is negatively affected whenever the application has to share resources between processors. Zen has some amazing interconnect that has almost 1:1 scaling.

OY to the VEY

NOOOOOOOOOOOOOOOOOOOOO

B T F O

>2 32 cores barely beats 2 28 cores
wow it's nothing

please read op you fucking intelfag
>The Platinum 8180 costs $12000 a pop and it's getting stomped by a $4000 EPYC
>$12000 is getting stomped by $4000

>$12000 because it's 4S+ scalable
>the $4000 EBIN isn't
Fucking retard

Intel 4S is dogshit though.

>muh scalable
>scales like shit on dual socket let alone on quad socket

check the clock speed

At least it can scale to 4/8S
Some datacenters simply need as many cores as they can stuff into a single blade server, memory and IOs be damned.

why would you give a shit about >muh cores at this point
AMD is cheaper and faster

80% of the server market is 2S so I think AMD is making the right call here purely from a marketplace perspective.

I hope the stock drops on Q2 earnings so I can buy a fuckload of it. I'm pretty sure Ryzen 5 won't have changed anything, not even miners should have changed it that much.

t. cuck that didn't buy in at $2

Yes, it still haunts me. I wanted to buy too, but I was a poor student at the time. Now I'm a full blown wagecuck with some money to spare.

If it drops it won't drop much.
If ER is good it will go over $18 and you'll be in a even worse situation

delid dis

t. rajesh

...

they merely adopted the designated shitting streets
AMD was born in it, molded by it

...

PLZ BUY OUR CPU, GUYS!!!!

>Average Score: 283.13 GB/s
>Average Capacity: 162.00 MB
What are these metrics?

Made up numbers to make Intel look worse than it is

Good goy! Remember to buy our Xeon Scalable Family processors. They are proven and battle-tested!

If you're going to post about how badly it scales, you should include some information to deduce that. What's the corresponding 1S score?

While it's impressive that they're managing the beat Intel at those clocks, I have to say I was expecting higher clocks from EPYC given the supposed efficiency of Zen.

Uncore is not free and they want to fir into 180w TDP. Intel's 28c is 205w.

DELID DIS
>DELID DIS
DELID DIS
>DELID DIS

wait a minute...

>Literally no competition for 10 years
>Implying intel doesn't have a lineup of massive performance improvements to dish out as soon as they get some real competition
I'm going to enjoy buying up cheap intel stock and selling it for big bux.

Yes-yes, goy, we TOTALLY have a completely new super-incredible x86 uarch. Believe us.

>White CEOs
>ever
LMAO

is this... a...

Source: My uncle works at Nintendo.

So that's why Intel decided to create Skylake-X. They wanted to heat things up before releasing this new uber architecture that they totally have

>heat things up
Pun intended.

This idea only shows a complete lack of understanding of how the semiconductor business works. It's never a good idea to sit on performance improvements. If you can make a smaller chip that performs better, you always do so because it means you can use each wafer more efficiently and thus make more money. No competition means you can price those parts at a premium, so it's not as if you're losing any money by increasing performance year over year. In fact, if you can increase performance each year, you incentivize consumers to upgrade every year, or two years, etc.

There isn't a reason to sit on semiconductor tech, ever.