Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us $3,400 per month for bandwidth bills alone, and since we don't believe in shoving popup ads to our registered users, we try to make the money back through forum registrations.
«245 »
  • Post
  • Reply
Eyes Only
May 20, 2008

Do not attempt to adjust your set.


Is it an actual shortage or is Intel just unable to keep up with unexpected growth in the market?

Adbot
ADBOT LOVES YOU

Winks
Feb 16, 2009

Alright, who let Rube Goldberg in here?


Eyes Only posted:

Is it an actual shortage or is Intel just unable to keep up with unexpected growth in the market?

The slight increase in consumer chip prices are likely because of the production of the 9000 series chips, the others are because of demand. They've been increasing their 2018 revenue forecast by a few billion over the year.

Seamonster
Apr 30, 2007

IMMER SIEGREICH


Mr.Radar posted:

Even Intel is forecasting AMD will take about 20% of the server CPU market next year. Depending on how well 7nm Epyc ends up performing, how well AMD can keep up with demand, and how long Intel's supply shortages last that could easily go higher. And while the server CPU market isn't a very large portion of the total CPU market it's a very profitable sector for Intel so losses in that sector will disproportionately impact Intel's bottom line.

Good. Because no way in hell is AMD worth ~30bn without a significant market share increase in the enterprise space.

Combat Pretzel
Jun 23, 2004

No, seriously... what kurds?!

Rumors are going around that AMD is going to release an X499 chipset. Around CES time. I sure hope there's something revolutionary coming with it, because otherwise it's loving stupid to do that out of step with the Threadrippers.

Khorne
May 1, 2002

Goonstone Champ x2

Seamonster posted:

Good. Because no way in hell is AMD worth ~30bn without a significant market share increase in the enterprise space.
The entire market is significantly over valued. AMD was the only company I know of that was under valued until recently, and it's mostly because they were sucking at everything for a few years and CPUs for far longer than that.

They face huge issues in so many spaces right now, but they're going to kick rear end and take names in the server CPU market if they see adoption and can supply the demand. The problem is they've had good products before that didn't sell that well relative to how they beat the competition on price and performance. I hope it's different this time, and it seems it will be, but no one can see the future.

Khorne fucked around with this message at Sep 9, 2018 around 02:36

Paul MaudDib
May 2, 2006

"Tell me of your home world, Usul"


Eyes Only posted:

Is it an actual shortage or is Intel just unable to keep up with unexpected growth in the market?

Intel had expected to move production onto 10nm a long time ago and didn't expand 14nm enough to keep up with growth. iirc Z390 was supposed to be on 14nm and they've backed off that because there's just no capacity available and now it's just a firmware update for Z370.

Given the prolonged delays on 10nm I'd be curious whether Intel starts taping out low-priority stuff (eg things like chipsets or Atoms) onto 14nm or 7nm to free up their internal capacity, but they've traditionally been hesitant to do that over IP concerns. It would obviously look real bad for them too, but appearances aside the question is whether shortages would be worse and/or how quickly they think 10nm will be operational.

Paul MaudDib fucked around with this message at Sep 9, 2018 around 04:13

Christobevii3
Jul 3, 2006


I am unable to get a Dell Precision T7920 warrantied at work with a 12Core Xenon. They're trying to give ma a dual 20 core replacement at 2.0Ghz base clocks to satisfy due to no availability of parts for the last month.

spasticColon
Sep 22, 2004

In loving memory of Donald Pleasance

Khorne posted:

There's a chance zen2 won't hit until q3/q4 2019. I just wanted to buy a new computer.

Same here. I'm still on a oc'd i5-2500K and it's starting to show it's age in newer games.

spasticColon fucked around with this message at Sep 10, 2018 around 02:02

pixaal
Jan 8, 2004

All ice cream is now for all beings, no matter how many legs.

Toilet Rascal

Christobevii3 posted:

I am unable to get a Dell Precision T7920 warrantied at work with a 12Core Xenon. They're trying to give ma a dual 20 core replacement at 2.0Ghz base clocks to satisfy due to no availability of parts for the last month.

Are they throwing in additional Windows Server licences to cover the added cores 8 cores? I don't know what the clock on your 12 core is, but I assume it's higher than 2ghz, which is also likely a no-go.

Paul MaudDib
May 2, 2006

"Tell me of your home world, Usul"


Paul MaudDib posted:

Intel had expected to move production onto 10nm a long time ago and didn't expand 14nm enough to keep up with growth. iirc Z390 was supposed to be on 14nm and they've backed off that because there's just no capacity available and now it's just a firmware update for Z370.

Given the prolonged delays on 10nm I'd be curious whether Intel starts taping out low-priority stuff (eg things like chipsets or Atoms) onto 14nm or 7nm to free up their internal capacity, but they've traditionally been hesitant to do that over IP concerns. It would obviously look real bad for them too, but appearances aside the question is whether shortages would be worse and/or how quickly they think 10nm will be operational.

On this note...

quote:

Intel is encountering tight 14nm process production capacity in-house, and is looking to outsource part of its 14nm chip production to Taiwan Semiconductor Manufacturing Company (TSMC), according to industry sources.

Intel intends to give priority to its high-margin products mainly server-use processors and chipsets amid its tight 14nm process capacity, and therefore plans to outsource the production of its entry-level H310 and several other 300 series desktop processors to TSMC, the sources indicated.

Intel has seen its overall 14nm chip supply fall short of demand by as much as 50%, the sources said. Outsourcing has become the only and appropriate choice for Intel since the company is unlikely to build additional 14nm process capacity, the sources noted.

TSMC is already a contract manufacturer of Intel for SoFIA-series handset SoC chips and FPGA products, and makes Intel's baseband chips for use in the iPhone, the sources said.

Motherboard makers expect the tight supply of Intel's 14nm chipsets to ease by the end of 2018.
https://www.digitimes.com/news/a20180910PD210.html

Munkeymon
Aug 14, 2003

Motherfucker's got an
armor-piercing crowbar! Rigoddamndicu𝜆ous.



Nap Ghost

Khorne posted:

The entire market is significantly over valued. AMD was the only company I know of that was under valued until recently, and it's mostly because they were sucking at everything for a few years and CPUs for far longer than that.

They face huge issues in so many spaces right now, but they're going to kick rear end and take names in the server CPU market if they see adoption and can supply the demand. The problem is they've had good products before that didn't sell that well relative to how they beat the competition on price and performance partially because their competition was ratfucking them. I hope it's different this time, and it seems it will be, but no one can see the future.

FTFY

NewFatMike
Jun 11, 2015




Comedy option that's all-too real in the clown shoes timeline: contract to GloFo.

They did say they wanted to focus on 14 and 12nm

pixaal
Jan 8, 2004

All ice cream is now for all beings, no matter how many legs.

Toilet Rascal

NewFatMike posted:

Comedy option that's all-too real in the clown shoes timeline: contract to GloFo.

They did say they wanted to focus on 14 and 12nm

I honestly thought that is exactly what had happened. GloFo dropping 7nm would make so much sense if that was true.

Harik
Sep 9, 2001

From the hard streets of Moscow
First dog to touch the stars



Plaster Town Cop

Paul MaudDib posted:

On this note...

https://www.digitimes.com/news/a20180910PD210.html

quote:

entry-level H310 and several other 300 series desktop processors to TSMC,
That's a bizarre mistake for a tech rag to make.

I ended up being out of town all weekend, sorry for the post&run. Re: tire fire:
Taking a 20% hit on their main cash cow has got to hurt, and no amount of enthusiast toys can make up for it. Doubly so when their biggest buyers have to go elsewhere because you just can't supply them. How long can they get away with blaming it on their ex-CEO, and when do we start seeing more C-level heads roll?

In AMD news, the abandonment of 7nm has led to another amendment to the WSA. WFFCtech so probably need another confirmation, and they don't know what the changes are. I'd be surprised if that decision didn't cause a material change in the relationship worthy of a renegotiation.

And goddamn AMD just run some cheap DRAM IP through Glofo already. The market will gobble it up and you won't be paying a tax on every real chip you make. Your "costs" are going to be lower than everyone else because you get to subtract the double-penalties out.

ColTim
Oct 29, 2011


I wonder how much of the bottleneck stems from increasing the core counts (and die dize) on the bread and butter CPUs rather drastically in the past couple generations.

Combat Pretzel
Jun 23, 2004

No, seriously... what kurds?!

What's the point of having a WSA, if GloFo doesn't even want to support the desired node size.

Cygni
Nov 12, 2005

raring to post



Didn't see this mentioned, AMD launched more of the Zen+ stack today. 2700E and 2600E (45w limited parts), 2500X (4/8), and 2300X (4/4)

https://www.anandtech.com/show/1334...2500x-and-2300x

lllllllllllllllllll
Feb 28, 2010

Now the scene's lighting is perfect!


quote:

AMD’s Ryzen 5 2600E and the Ryzen 7 2700E are the company’s first eight and six-core CPUs featuring a 45 W default TDP. The new processors enable PC makers to build small form-factor desktops that do not need high-performance cooling. The chips are clocked at 300-400 MHz below their 65 W counterparts. At the same time, the Ryzen 5 2600E and the Ryzen 7 2700E still have the cache as the higher rated parts, but will not support Precision Boost Overdrive.
That sounds kinda cool for the people unvervolting (me). 8 cores in a 45W package is quite something. The CPU could be idling almost at room temperature.

Paul MaudDib
May 2, 2006

"Tell me of your home world, Usul"


Cygni posted:

Didn't see this mentioned, AMD launched more of the Zen+ stack today. 2700E and 2600E (45w limited parts), 2500X (4/8), and 2300X (4/4)

https://www.anandtech.com/show/1334...2500x-and-2300x

Single-CCX too, curious how well that'll play out.

I'm told the intra-CCX latency is extremely low, like lower than a ringbus even, so on paper they should be a lot better in latency-sensitive situations like gaming. OTOH, people playing around with disabling cores and APUs don't seem to make much of a difference.

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!


Late releases like these seem to kind of indicate a more May-July release for Ryzen 3000 IMHO.

Craptacular!
Jul 9, 2001

Fuck the DH


Those E chips are pretty awesome for my DVR/transcoding box. My super cheap A320,board maxes out at 45W, but More Cores Are Better for transcoding and I probably wouldn’t have to lean on GPU acceleration and could just encode in software while using a GT1030.

Combat Pretzel
Jun 23, 2004

No, seriously... what kurds?!

I find that temperature offset stuff on the Threadripper (and Ryzen?) pretty irritating.

Desuwa
Jun 2, 2011

I'm telling my mommy. That pubbie doesn't do video games right!

Combat Pretzel posted:

I find that temperature offset stuff on the Threadripper (and Ryzen?) pretty irritating.

I've found most windows software already handles it, the only one that hasn't yet is msi kombuster's on screen display. Same for my motherboard which is how I'm controlling the fans. Linux software doesn't yet but they do tend to lag behind.

Mr Shiny Pants
Nov 12, 2012


Combat Pretzel posted:

I find that temperature offset stuff on the Threadripper (and Ryzen?) pretty irritating.

Yep, mine says 76 degrees on Linux. I am not even sure I need to subtract 27 degrees off of it, but it ran over a 100 degrees in the summer and I can't quite imagine that being right.

Mr Shiny Pants fucked around with this message at Sep 12, 2018 around 06:35

PC LOAD LETTER
May 23, 2005
WTF?!

Slippery Tilde

Paul MaudDib posted:

I'm told the intra-CCX latency is extremely low
The inter-CCX latency really matters for high(er) thread count/NUMA stuff. For anything else lowering the inter-CCX latency will be nice but not a really big deal.

Which is why they can currently get away with it being so high as it is and still get fairly good performance in desktop application stuff.

Combat Pretzel
Jun 23, 2004

No, seriously... what kurds?!

Mr Shiny Pants posted:

Yep, mine says 76 degrees on Linux. I am not even sure I need to subtract 27 degrees off of it, but it ran over a 100 degrees in the summer and I can quite imagine that being right.
The whole temp stuff seems weird. Package temperatures on my TR seem to be more correct, without the 27°C offset. But when doing random low effort crap, like watching a video and using a browser, it randomly tends to jump 8-10°C and then comes back down. But when playing a game, which loads the CPU, it's steady and goes slowly up (when the water cooling is trying to get to steady state).

Paul MaudDib
May 2, 2006

"Tell me of your home world, Usul"


PC LOAD LETTER posted:

The inter-CCX latency really matters for high(er) thread count/NUMA stuff. For anything else lowering the inter-CCX latency will be nice but not a really big deal.

Which is why they can currently get away with it being so high as it is and still get fairly good performance in desktop application stuff.

I said intra-CCX (within a CCX), not inter-.

Within a CCX, Ryzen actually has lower latency than Intel does, so at least on paper a single-CCX design would be a lot better for gaming, but like I said, people have played with it and it doesn't seem to help all that much and I don't really get why

Anime Schoolgirl
Nov 28, 2002

~perfect~
battlebrother





a simpler explanation is that they're having real trouble finding dies that are partially hosed on both sides

like if there are two/three bad cores on one CCX but the other CCX is fine, you'd probably want to go with what's faster in microcode, even in game performance it doesn't actually translate to much

Anime Schoolgirl fucked around with this message at Sep 12, 2018 around 00:19

Desuwa
Jun 2, 2011

I'm telling my mommy. That pubbie doesn't do video games right!

I was playing with PBO and I think I might just leave my 2950x at stock or with only a very modest boost, like 20W just to feel like I gained something. The boosts don't seem to be worth pushing it since I'm not constantly loading all cores, doubly so because if I load my UPS too much its fan will turn on and that thing is loud.

Combat Pretzel posted:

The whole temp stuff seems weird. Package temperatures on my TR seem to be more correct, without the 27°C offset. But when doing random low effort crap, like watching a video and using a browser, it randomly tends to jump 8-10°C and then comes back down. But when playing a game, which loads the CPU, it's steady and goes slowly up (when the water cooling is trying to get to steady state).

The temperature reported by the package isn't from a single sensor but instead takes the maximum of a bunch of different sensors that aren't individually exposed. So it's not your entire chip getting 10C hotter, it's just one small part, which is entirely normal with modern CPU boost algorithms. What's happening is there's tons of power and thermal headroom at idle so when a task does come in a single core will run flat out until it's done with as much power as a single core can handle. You hit another steady state, where the energy leaving the cores equalizes with the energy entering the water, very quickly, and below that temperatures can swing wildly.

SwissArmyDruid
Feb 14, 2014



Anime Schoolgirl posted:

a simpler explanation is that they're having real trouble finding dies that are partially hosed on both sides

like if there are two/three bad cores on one CCX but the other CCX is fine, you'd probably want to go with what's faster in microcode, even in game performance it doesn't actually translate to much

"we can't field this downmarket part because yields are too good!"

Why do I have this intense feeling of deja vu?

Potato Salad
Oct 23, 2014

Nobody Cares



Tortured By Flan

SwissArmyDruid posted:

"we can't field this downmarket part because yields are too good!"

Why do I have this intense feeling of deja vu?

They *sniff* grow up so fast.

SwissArmyDruid
Feb 14, 2014



Put a sock in it, you know it was a rhetorical question. =P I cut my teeth on P5s and K6s.

PC LOAD LETTER
May 23, 2005
WTF?!

Slippery Tilde

Paul MaudDib posted:

I said intra-CCX (within a CCX), not inter-.
Whoops sorry.

ufarn
May 30, 2009


Combat Pretzel posted:

The whole temp stuff seems weird. Package temperatures on my TR seem to be more correct, without the 27°C offset. But when doing random low effort crap, like watching a video and using a browser, it randomly tends to jump 8-10°C and then comes back down. But when playing a game, which loads the CPU, it's steady and goes slowly up (when the water cooling is trying to get to steady state).
I had to set my minimum processor state to something like 9x% in the power settings, because Ryzen seems to struggle to find a plateau. Having it jump above my fan curve trigger ever minute was driving me mad.

Klyith
Aug 3, 2007

GBS Pledge Week


ufarn posted:

I had to set my minimum processor state to something like 9x% in the power settings, because Ryzen seems to struggle to find a plateau. Having it jump above my fan curve trigger ever minute was driving me mad.

That seems awfully wasteful, look into your fan control to set some hysteresis on the CPU fan (and any others that are based on CPU temp). Even an air cooler has plenty of thermal mass, the spike in temperature from a short burst of full clockspeed is more about conductivity. So a 1/2 second delay before the fan reacts to the CPU isn't really hurting anything.


Also if using BIOS fan control, build your fan curve with that +20 offset accounted for -- I was also having this problem until I realized that I needed to start spinning up at 60C not 40. That despite the temp display in the BIOS itself using the non-offset temp.

ufarn
May 30, 2009


Klyith posted:

That seems awfully wasteful, look into your fan control to set some hysteresis on the CPU fan (and any others that are based on CPU temp). Even an air cooler has plenty of thermal mass, the spike in temperature from a short burst of full clockspeed is more about conductivity. So a 1/2 second delay before the fan reacts to the CPU isn't really hurting anything.


Also if using BIOS fan control, build your fan curve with that +20 offset accounted for -- I was also having this problem until I realized that I needed to start spinning up at 60C not 40. That despite the temp display in the BIOS itself using the non-offset temp.
I set mine to 55C. I use my CPU for video playback with hardware accelleration disabled because my GPU is ancient, but there's definitely something off about my cooling. I'm using a D14 cooler with Noctua paste which should do the job, but something still ain't right.

Only thing I've got left to try out is bump the minimum fan speed for the case fans in ASUS's "quiet" profile with the AI Overclocking Profile set to "Auto" for now.

It's not so much a solution as a fix, but the fan hysteresis was driving me insane.

ufarn fucked around with this message at Sep 12, 2018 around 11:09

Drakhoran
Oct 21, 2012



So if anyone here bought a 32 core Threadripper and will at least occasionally use it for gaming, you really should update your drivers.





Benchmarks with new Vs. old drivers here.

Truga
May 4, 2014


Lipstick Apathy

I have it on pretty good authority that epyc2 is going to be 8x8 cores and also a +1 something else. Is there anything that +1 thing could be besides a gpu?

SwissArmyDruid
Feb 14, 2014



ARM security co-processor? I don't think AMD ever stopped flogging those as an option, yeah?

SwissArmyDruid fucked around with this message at Sep 14, 2018 around 13:55

Adbot
ADBOT LOVES YOU

Mark Larson
Dec 27, 2003

Interesting...

SwissArmyDruid posted:

ARM security co-processor? I don't think AMD ever stopped flogging those as an option, yeah?

Is that basically what an Apple T2 is?

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply
«245 »