AMD is good. Happy Easter - He Has Ryzen - The Something Awful Forums

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > AMD is good. Happy Easter - He Has Ryzen

«‹›87 »

Twerk from Home: Jan 17, 2009; This avatar brought to you by the 'save our dead gay forums' foundation.

LibCrusher posted:

isnt their 14 just as good as everyone else's 10?

i'm honestly expecting that in a desktop power envelope, 14nm Rocket Lake is about to be evenly trading blows with Zen 3, and probably outperform in some situations, and that's 14nm vs 7nm.

laptops are a different game, but intel has 10nm there anyway

# ? Mar 1, 2021 04:00

Adbot: ADBOT LOVES YOU

# ? Jun 2, 2024 16:54

Laslow: Jul 18, 2007

alder lake is interesting at least. not that it�ll be great or anything, but because it�s weird. BIG.huge on x86? that�s certainly... something.

# ? Mar 1, 2021 05:27

Nomnom Cookie: Aug 30, 2009

LibCrusher posted:

isnt their 14 just as good as everyone else�s 10?

yeah

too bad for intel everyone else is on 7nm and going to 5nm by the time intel figures out their 10nm process

# ? Mar 1, 2021 05:40

Cybernetic Vermin: Apr 18, 2005

Nomnom Cookie posted:

yeah

too bad for intel everyone else is on 7nm and going to 5nm by the time intel figures out their 10nm process

this story was a pretty good summary in 2018, but in 2021 i think we really need to add to that that no one anywhere can apparently make enough chips to meet demand at any even semi-competitive process.

so intel may still have a huge advantage with their own fab capacity for the simple reason that everyone else is competing for slivers at tsmc. it is not even clear how much of that is expected capacity not being sufficient, or perhaps some poor yields or other issues actually cropping up.

# ? Mar 1, 2021 17:50

Private Speech: Mar 30, 2011; I HAVE EVEN MORE WORTHLESS BEANIE BABIES IN MY COLLECTION THAN I HAVE WORTHLESS POSTS IN THE BEANIE BABY THREAD YET I STILL HAVE THE TEMERITY TO CRITICIZE OTHERS' COLLECTIONS

IF YOU SEE ME TALKING ABOUT BEANIE BABIES, PLEASE TELL ME TO

EAT. SHIT.

Jim Silly-Balls posted:

nobody ever got fired for buying IBM, except in 2021 where you would get laughed out of a job

I mean RHEL is still basically NEGFFB IBM

but then that's an acquisition and they are doing their best to suck the life out of it

# ? Mar 1, 2021 18:03

Beve Stuscemi: Jun 6, 2001

dear nvidia. pleas make new video cards that are just 4 980tis bolted together, to avoid new chip fab shortages. also please hire me, this is a good idea I promise

# ? Mar 1, 2021 23:12

Twerk from Home: Jan 17, 2009; This avatar brought to you by the 'save our dead gay forums' foundation.

Jim Silly-Balls posted:

dear nvidia. pleas make new video cards that are just 4 980tis bolted together, to avoid new chip fab shortages. also please hire me, this is a good idea I promise

they did this already, don't you remember the Geforce 7950 GX2?

# ? Mar 1, 2021 23:22

poty: Jun 21, 2008; 虹はどこで終わるのですか？あなたの魂の中で、または地平線で?

Cybernetic Vermin posted:

this story was a pretty good summary in 2018, but in 2021 i think we really need to add to that that no one anywhere can apparently make enough chips to meet demand at any even semi-competitive process.

so intel may still have a huge advantage with their own fab capacity for the simple reason that everyone else is competing for slivers at tsmc. it is not even clear how much of that is expected capacity not being sufficient, or perhaps some poor yields or other issues actually cropping up.

apple doesnt seem to have any supply issues and they use the newest process where unexpected poor yields would show up

but yeah dont know who is getting the current 7nm capacity

# ? Mar 1, 2021 23:50

fart simpson: Jul 2, 2005; DEATH TO AMERICA

Twerk from Home posted:

they did this already, don't you remember the Geforce 7950 GX2?

i got one of these for free back in the day. he�ll yeah

# ? Mar 2, 2021 01:30

fart simpson: Jul 2, 2005; DEATH TO AMERICA

the drivers were terrible and stuff crashed all the time

# ? Mar 2, 2021 01:30

Samuel L. ACKSYN: Feb 29, 2008

Jim Silly-Balls posted:

dear nvidia. pleas make new video cards that are just 4 980tis bolted together, to avoid new chip fab shortages. also please hire me, this is a good idea I promise

# ? Mar 2, 2021 02:27

Laslow: Jul 18, 2007

someone gifted me a 4850x2 hand-me down, which got paired with my existing 4850, so i briefly had a triple crossfire setup. it worked ok with the 3 games that utilized it at all. they got replaced by a nvidia fermi, the king of power efficiency of its day.

# ? Mar 2, 2021 04:02

Woolwich Bagnet: Apr 27, 2003

in college for protein modeling we had a computer with 4 gtx295s in it that i put together. it was a beast at the time

# ? Mar 2, 2021 04:05

git apologist: Jun 4, 2003

Jim Silly-Balls posted:

nobody ever got fired for buying IBM, except in 2021 where you would get laughed out of a job

people still buy it OP

# ? Mar 2, 2021 07:24

ConanTheLibrarian: Aug 13, 2004; dis buch is late; Fallen Rib

Cybernetic Vermin posted:

this story was a pretty good summary in 2018, but in 2021 i think we really need to add to that that no one anywhere can apparently make enough chips to meet demand at any even semi-competitive process.

so intel may still have a huge advantage with their own fab capacity for the simple reason that everyone else is competing for slivers at tsmc. it is not even clear how much of that is expected capacity not being sufficient, or perhaps some poor yields or other issues actually cropping up.

if tsmc had 3 times the 7nm capacity it would still be over subscribed

after global foundries had to abandon their 7nm efforts i wrote them off, but theyre probably able to sell their 12nm crap at a premium now so they might end up with the cash they need to get back in the game

# ? Mar 2, 2021 14:52

Cybernetic Vermin: Apr 18, 2005

ConanTheLibrarian posted:

if tsmc had 3 times the 7nm capacity it would still be over subscribed

after global foundries had to abandon their 7nm efforts i wrote them off, but theyre probably able to sell their 12nm crap at a premium now so they might end up with the cash they need to get back in the game

yeah, it is all pretty interesting. obviously i know almost nothing of the business or technology, but it may have been a bit knee-jerk to think that the announcement to stick at their current node for the time being was bordering on a death sentence for global foundries. similarly intels situation is pretty funny-bad but one does notice that they are making plenty of hay continuing on 14nm, and i don't think it is *all* inertia.

not to predict that the newer, smaller and better shrink train has already derailed, but the picture may be getting more complicated pretty quickly.

# ? Mar 2, 2021 15:32

EIDE Van Hagar: Dec 8, 2000; Beep Boop

Cybernetic Vermin posted:

yeah, it is all pretty interesting. obviously i know almost nothing of the business or technology, but it may have been a bit knee-jerk to think that the announcement to stick at their current node for the time being was bordering on a death sentence for global foundries. similarly intels situation is pretty funny-bad but one does notice that they are making plenty of hay continuing on 14nm, and i don't think it is *all* inertia.

not to predict that the newer, smaller and better shrink train has already derailed, but the picture may be getting more complicated pretty quickly.

i think the m1 is a good indicator of where things are going, adding hardware accelerators for more for specific tasks.

moore�s law for a long time made hardware acceleration (except maybe GPUs and some similar vector/SIMD loads) quickly redundant as the CPUs 2 to 4 years down the road could do it all in software.

when that stops the only way to get increased performance in the same power envelope is to take frequently performed computations back into hardware to speed them up.

there are power and cost tradeoffs to adding area like that, but i think that�s the direction stuff is going.

depending on your workload is it really best to make a cpu on the nth shrink process? maybe you can handle that workload better with hardware acceleration and picking a more refined and power optimized version of the n-1 process, especially if the newest process is now taking 6-10 years to develop instead of 2-3.

EIDE Van Hagar fucked around with this message at 16:17 on Mar 2, 2021

# ? Mar 2, 2021 16:13

Beve Stuscemi: Jun 6, 2001

it�s hosed up how Apple bet the house on the power architecture and had to make up bs marketing garbage like the megahertz myth to make it not seem like a dumpster fire and now they�re back on a non-x86 architecture and it turns out that rules?

:rip:

intel

not really though because they�re so big they�ll find a way to make money and stick around.

# ? Mar 2, 2021 16:22

Private Speech: Mar 30, 2011; I HAVE EVEN MORE WORTHLESS BEANIE BABIES IN MY COLLECTION THAN I HAVE WORTHLESS POSTS IN THE BEANIE BABY THREAD YET I STILL HAVE THE TEMERITY TO CRITICIZE OTHERS' COLLECTIONS

IF YOU SEE ME TALKING ABOUT BEANIE BABIES, PLEASE TELL ME TO

EAT. SHIT.

well PPC was kinda garbage though

# ? Mar 2, 2021 16:51

Sagebrush: Feb 26, 2012

that was IBM making GBS threads the bed though right

like they promised a laptop-suitable G5 for years and never managed to make one that wasn't on the verge of melting down (factory liquid cooling lmao) and as a result the first intel apples were macbooks

# ? Mar 2, 2021 18:03

Twerk from Home: Jan 17, 2009; This avatar brought to you by the 'save our dead gay forums' foundation.

Private Speech posted:

well PPC was kinda garbage though

i dunno man, the Xbox 360 had some pretty good games

# ? Mar 2, 2021 18:09

FMguru: Sep 10, 2003; peed on;
sexually

Sagebrush posted:

that was IBM making GBS threads the bed though right

like they promised a laptop-suitable G5 for years and never managed to make one that wasn't on the verge of melting down (factory liquid cooling lmao) and as a result the first intel apples were macbooks

the problem was that apple was the fifth most important ppc customer (behind ibms own workstations and all three of that generations consoles) and they were the only ones asking for a g5 laptop chip. since apple had just survived a brush with bankruptcy and still had a very uncertain future. moto and ibm didnt want to commit the resources to build a nextgen laptop chip that had only one customer (that wasnt very high volume and wasnt a sure bet to still be in business by the time the chip was delivered)

# ? Mar 2, 2021 18:21

Perplx: Jun 26, 2004; Best viewed on Orgasma Plasma; Lipstick Apathy

steve on stage said they would ship 3Ghz powerpc chips soon, that never happened, could you imagine how angry that made him?
thats even worse than nvidia loving up a generation of laptops

# ? Mar 2, 2021 19:03

DamnGlitch: Sep 2, 2004

yeah I don�t get the impression IBM didn�t want to deliver a better product on the G5, but that they had hit an architectural culdesac that meant they werent able to achieve speed power consumption and temps that were acceptable.

Basically intel did the same thing 15 years later and now we have the m1.

# ? Mar 2, 2021 19:08

Nomnom Cookie: Aug 30, 2009

Cybernetic Vermin posted:

yeah, it is all pretty interesting. obviously i know almost nothing of the business or technology, but it may have been a bit knee-jerk to think that the announcement to stick at their current node for the time being was bordering on a death sentence for global foundries. similarly intels situation is pretty funny-bad but one does notice that they are making plenty of hay continuing on 14nm, and i don't think it is *all* inertia.

not to predict that the newer, smaller and better shrink train has already derailed, but the picture may be getting more complicated pretty quickly.

"my products are donkey poo poo but my competitors are so supply constrained that customers have no choice but to buy what i'm selling" is not where you want to be, long term. intel is doing fine right now, but it's still the same company that hosed up 10nm so badly they went from being a node ahead of everyone else to a node behind. if they can't fix their poo poo and make chips smaller than 14nm that don't suck rear end, they're going to start hurting bad when supply loosens up at TSMC

# ? Mar 2, 2021 19:09

echinopsis: Apr 13, 2004; by Fluffdaddy

290 i finally checked my ram speed

was set to auto and was moving at 2133 but it�s meant to run at 3200

but i tried it at 3200 and it wouldn�t post

went up a step at a time and 2800 seems to be the fastest it�ll post at 🤷‍♂️

# ? Mar 6, 2021 09:28

infernal machines: Oct 11, 2012; we monitor many frequencies. we listen always. came a voice, out of the babel of tongues, speaking to us. it played us a mighty dub.

try raising the voltage

# ? Mar 6, 2021 14:46

echinopsis: Apr 13, 2004; by Fluffdaddy

hmm common sense tells me never to go more than 3 menus deep in bios

# ? Mar 6, 2021 15:07

fart simpson: Jul 2, 2005; DEATH TO AMERICA

echinopsis posted:

hmm common sense tells me never to go more than 3 menus deep in bios

bump the voltage to like 17 volts or whatever imo

# ? Mar 6, 2021 15:25

infernal machines: Oct 11, 2012; we monitor many frequencies. we listen always. came a voice, out of the babel of tongues, speaking to us. it played us a mighty dub.

echinopsis posted:

hmm common sense tells me never to go more than 3 menus deep in bios

common sense should probably tell you not to gently caress with this stuff if you don't know what you're doing, but here we are.

presumably your ram has XMP, just use the highest profile and see if it posts

# ? Mar 6, 2021 15:35

infernal machines: Oct 11, 2012; we monitor many frequencies. we listen always. came a voice, out of the babel of tongues, speaking to us. it played us a mighty dub.

XMP will set the voltage along with the timings. if you don't have XMP, buy some ram that's less poo poo or face to learning about bullshit like tCAS, tRCD, tRP, & tRAS and a lot of trial and error

# ? Mar 6, 2021 15:43

hobbesmaster: Jan 28, 2008

turn on XMP, if it doesn�t post turn on XMP and give the the RAM 1.4V instead of 1.35V. if it still doesn�t post see if you can RMA it for something that does what it�s advertised to

# ? Mar 6, 2021 17:06

The Management: Jan 2, 2010; sup, bitch?

not posting is a feature.

# ? Mar 6, 2021 17:38

TerminalRaptor: Nov 6, 2012; Mostly Harmless

The Management posted:

not posting is a feature.

# ? Mar 6, 2021 18:59

jre: Sep 2, 2011; To the cloud ?

The Management posted:

not posting is a feature.

# ? Mar 6, 2021 19:23

Beve Stuscemi: Jun 6, 2001

I�m boastin about POSTin

# ? Mar 6, 2021 19:41

infernal machines: Oct 11, 2012; we monitor many frequencies. we listen always. came a voice, out of the babel of tongues, speaking to us. it played us a mighty dub.

loving epyc, yo

# ? Mar 8, 2021 16:38