New around here? Register your SA Forums Account here!

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 money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Locked thread
JawnV6
Jul 4, 2004

So hot ...

Boiled Water posted:

Switching companies isn't that weird. In my line of work it's usually done because it's easier to move up (both in title and payscale) than internal promotion.
:confused: Yeah that's how you get from staff to senior staff. The principal architect of the CPU isn't hopping companies for that reason.

Adbot
ADBOT LOVES YOU

Malcolm XML
Aug 8, 2009

I always knew it would end like this.

JawnV6 posted:

:confused: Yeah that's how you get from staff to senior staff. The principal architect of the CPU isn't hopping companies for that reason.

yeah more than likely he finished up whatever he needed to do at AMD and then samsung showed him a giant pile of cash and marketable securities along with leading edge process tech and a huge profitable company backing him

Anime Schoolgirl
Nov 28, 2002

Malcolm XML posted:

yeah more than likely he finished up whatever he needed to do at AMD and then samsung showed him a giant pile of cash and marketable securities along with leading edge process tech and a huge profitable company backing him
That didn't work when Intel tried that

Durinia
Sep 26, 2014

The Mad Computer Scientist

Anime Schoolgirl posted:

That didn't work when Intel tried that

That's because at Intel he'd have about 1/10th the autonomy/decision making ability.

Yaoi Gagarin
Feb 20, 2014

Is it possible that, since he was apparently so important at AMD, he might be bound by some kind of contract or other legal stuff that prohibits him from working for Intel?

Col.Kiwi
Dec 28, 2004
And the grave digger puts on the forceps...

VostokProgram posted:

Is it possible that, since he was apparently so important at AMD, he might be bound by some kind of contract or other legal stuff that prohibits him from working for Intel?
Nah. Non-compete agreements that try and tell you you can't go work for a competitor after you've quit don't hold up in court. People still try and draw up such agreements but there is plenty of precedent for courts deciding they are invalid/unenforcable.

JawnV6
Jul 4, 2004

So hot ...

Col.Kiwi posted:

Nah. Non-compete agreements that try and tell you you can't go work for a competitor after you've quit don't hold up in court. People still try and draw up such agreements but there is plenty of precedent for courts deciding they are invalid/unenforcable.
Again this holds up for anyone under 20 years experience or so but we're not discussing a mere mortal. Kai-Fu Lee's transition had a rough 6 months where he was barred from certain technical aspects even if the non-compete was eventually invalid.

Durinia posted:

That's because at Intel he'd have about 1/10th the autonomy/decision making ability.
I'd presume it's something closer to this. That kind of guy doesn't go begging companies for work, he can probably afford to be picky about the team and project.

adorai
Nov 2, 2002

10/27/04 Never forget
Grimey Drawer

Col.Kiwi posted:

Nah. Non-compete agreements that try and tell you you can't go work for a competitor after you've quit don't hold up in court. People still try and draw up such agreements but there is plenty of precedent for courts deciding they are invalid/unenforcable.
This isn't some jerkoff sales guy, this is the guy who has detailed design knowledge of the entire product. If the chief chemist at Coke left to work at PepsiCo, there would be a valid lawsuit. This is the same deal.

Killer robot
Sep 6, 2010

I was having the most wonderful dream. I think you were in it!
Pillbug

Col.Kiwi posted:

Nah. Non-compete agreements that try and tell you you can't go work for a competitor after you've quit don't hold up in court. People still try and draw up such agreements but there is plenty of precedent for courts deciding they are invalid/unenforcable.

Depends a lot. First by state, second by just what the terms are. In my state, a non-compete is legally enforceable if and only if you're actually paid something for it. If they just put it in for nothing it's just to scare you, not that it stops anyone.

syzygy86
Feb 1, 2008

Killer robot posted:

Depends a lot. First by state, second by just what the terms are. In my state, a non-compete is legally enforceable if and only if you're actually paid something for it. If they just put it in for nothing it's just to scare you, not that it stops anyone.

In California, non-complete clauses are invalid entirely (except for some very narrow cases that don't apply here). In most states, it seems they are generally enforcible.

evol262
Nov 30, 2010
#!/usr/bin/perl

syzygy86 posted:

In California, non-complete clauses are invalid entirely (except for some very narrow cases that don't apply here). In most states, it seems they are generally enforcible.

It depends a lot. Really. In general, they're more enforceable the further east you go, but I wouldn't say they're "generally enforceable" in most states. It matters whether you acquired an advantage which you wouldn't have had without the job (client base, contact with company that poached, level of seniority/IP knowledge, etc)

SYSV Fanfic
Sep 9, 2003

by Pragmatica
My ethics and law professor said it the best. A non compete is enforceable until it isn't.

Ragingsheep
Nov 7, 2009
Someone is suing AMD over Bulldozer.

http://legalnewsline.com/stories/510646458-amd-faces-suit-over-alleged-misrepresentation-of-new-cpu

Anime Schoolgirl
Nov 28, 2002

fishing for settlements

Kazinsal
Dec 13, 2011


If the suit actually says "cannot perform eight instructions at once" then hoo boy, one AMD engineer explaining instruction-level parallelism throws this entire suit out.

If it's over the whole "eight physical independently functioning cores are not present in Bulldozer octo-core chips" then yeah they're probably going to need to find something a bit more elaborate to weasel their way out of it.

WhyteRyce
Dec 30, 2001

Anime Schoolgirl posted:

fishing for settlements

It stemmed from crappy and terrible marketing gimmick so whatever

Don Lapre
Mar 28, 2001

If you're having problems you're either holding the phone wrong or you have tiny girl hands.
https://www.youtube.com/watch?v=mlRWqqWay7c

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!

Kazinsal posted:

If the suit actually says "cannot perform eight instructions at once" then hoo boy, one AMD engineer explaining instruction-level parallelism throws this entire suit out.

If it's over the whole "eight physical independently functioning cores are not present in Bulldozer octo-core chips" then yeah they're probably going to need to find something a bit more elaborate to weasel their way out of it.

No, it's even dumber I think

quote:

The suit alleges AMD built the Bulldozer processors by stripping away components from two cores and combining what was left to make a single “module.”

That's not how they're made at all and AMD can pretty easily prove that. AMD is not stripping down and mashing two cores together, they specifically designed a core to behave like two physical cores.

SwissArmyDruid
Feb 14, 2014
We interrupt this little piece of bullshit idiocy from overly litigious dumbasses to bring you this good news: Glofo announces success with AMD on their FinFET process:

http://www.globalfoundries.com/news...on-amd-products

Thinks are looking up, gents. :toot:

edit: Here's a TechReport article: http://techreport.com/news/29287/glofo-successfully-builds-amd-chips-on-14-nm-finfet-lpp-process

SwissArmyDruid fucked around with this message at 23:57 on Nov 5, 2015

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!
So AMD might have some mixed 16nm and 14nm products? I thought both were effectively 20nm but FinFet so what's the design advantage for choosing one over the other?

Also Oh god the new default avatar is horrifying what the hell admins.

Anime Schoolgirl
Nov 28, 2002

FaustianQ posted:

So AMD might have some mixed 16nm and 14nm products? I thought both were effectively 20nm but FinFet so what's the design advantage for choosing one over the other?

Also Oh god the new default avatar is horrifying what the hell admins.
from the apple a9 chip comparisons, 16nm tsmc is more efficient, 14nm can afford to run hotter and uses less die size

Kazinsal
Dec 13, 2011

FaustianQ posted:

Also Oh god the new default avatar is horrifying what the hell admins.

Oh so THAT's why I've been seeing multiple people around with bonzi buddy avatars. I assumed a bunch of people had all collectively pissed off one strange person who decided to spend three digits on passive-aggressive avatars but forgot to put in red text.

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!

Anime Schoolgirl posted:

from the apple a9 chip comparisons, 16nm tsmc is more efficient, 14nm can afford to run hotter and uses less die size

Sounds like 16nm is better for K12, Zen mobile, and mobile GPU. 14nm sounds better for Zen desktop, desktop iGPU and dGPU.

I mean there has to be a reason AMD is willing to tap into both TSMC and GoFlo for upcoming products. Maybe GoFlo 14nm will give Zen the overclocking headroom to make up IPC shortfall.

Professor Science
Mar 8, 2006
diplodocus + mortarboard = party

FaustianQ posted:

Sounds like 16nm is better for K12, Zen mobile, and mobile GPU. 14nm sounds better for Zen desktop, desktop iGPU and dGPU.

I mean there has to be a reason AMD is willing to tap into both TSMC and GoFlo for upcoming products. Maybe GoFlo 14nm will give Zen the overclocking headroom to make up IPC shortfall.
it's probably just to avoid being supply constrained, same as apple

SwissArmyDruid
Feb 14, 2014
Agreed, except that I'd couch that more in terms of 'not getting hosed by manufacturing issues' at either foundry.

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!
Wouldn't an A10-9870 TSMC perform differently from a A10-9870 from GloFlo though? Or is the difference minuscule enough that it won't come into play? If so won't two similar products with "different" lithographies needlessly confuse people, or I am I thinking to hard about this?

SwissArmyDruid
Feb 14, 2014
I won't say that there's zero possibility. The process is not identical, there's always that chance. But the overall net effect seems to not have affected iphone 6 sales, (whose chips are also coming from split Samsung/TSMC) so :shrug:

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!
Grain of NaCl, but http://wccftech.com/amd-next-gen-bristol-ridge-fx-9830p-soc-leaked/

It's possible this may be on a 14/16nm processes contrary to the safe but odd bet that it's on 28nm. If so Then we might be getting our first good look at AM4 boards and accompanying processors in 4 months, and I wouldn't say more than 6. Still hoping for some HBM on Bristol now the gen1 has no supply issues, otherwise it's so close to Carizzo that I don't get why they'd bother.

PC LOAD LETTER
May 23, 2005
WTF?!

FaustianQ posted:

Wouldn't an A10-9870 TSMC perform differently from a A10-9870 from GloFlo though?
Any difference would be in terms of TDP and/or clockspeed. Per clock performance wouldn't change a bit. Chances are there won't be much if any difference either which way.

FaustianQ posted:

Still hoping for some HBM on Bristol now the gen1 has no supply issues, otherwise it's so close to Carizzo that I don't get why they'd bother.
I would assume if they're using a somewhat improved Carrizo that they won't be using HBM. For desktop it could be a decent enough low end machine. DIY'ers will want more performance than that so I doubt they will care about it.

I think you're going to be waiting quite a while before you see a CPU or APU with HBM on it. 2017-ish still seems reasonable.

Bloody Antlers
Mar 27, 2010

by Jeffrey of YOSPOS
I recently had an interesting experience with my 5 year old AMD FX 8350. I had been playing a lot of Far Cry 4 and everything would be fine unless I played for over an hour or so, when my system would abruptly crash. I updated drivers and everything like that, had a few shorter play sessions without issue, and then a week or two later had my next crash while running a fairly loaded VM and video encode batch. This build had been stable for years leading up to this, and there hadn't been any hardware changes since I added a Samsung 840 Pro SSD a while back, so I had the dreadful fear of random failing component set in.

I decided to open up the case and have a looksy. I routinely change my case air filters, so there wasn't much dust build up anywhere, and all the components were well-seated. I turned the computer on with the case open, and to my utter flailing horror noticed that my CPU fan was completely dead. Motherboard didn't seem to notice or care. I had been operating with just a heatsink for who knows how long. The HSF is an ARCTIC Freezer 7 Pro Rev 2, which apparently couples a very well-designed heatsink with a gutter whore of a fan. I ordered a replacement 92mm fan and everything has been :c00lbutt: since.

TLDR; 8350 @ 4ghz stock settings with aftermarket HSF survives operating with just the heatsink and does not complain about dead fan until pushed pretty hard.

1gnoirents
Jun 28, 2014

hello :)
lol http://hothardware.com/news/amd-slapped-with-class-action-lawsuit-over-deceptive-core-count-in-bulldozer-cpus

I'm just starting to feel bad now

My Rhythmic Crotch
Jan 13, 2011

I'm finding it hard to feel too bad for AMD regarding that lawsuit. I always thought it was too much of a stretch to call one bulldozer module two cores. They shoulda seen some form of blow-back coming from a mile away.

1gnoirents
Jun 28, 2014

hello :)
Oh I agree, and it was by far the most important piece of marketing for them imo as far as CPUs went for the average person, but, drat... some timing.

Paul MaudDib
May 2, 2006

TEAM NVIDIA:
FORUM POLICE
There's no legal meaning to "core" - Bulldozer does have 8 integer execution units, and the 286 had exactly 0 FPUs per integer core. Bulldozer can execute 8 completely separate threads at a time - as long as they're logical or integer operations, rather than a logically-blocking FPU load on every core.

Intel "logical cores" aren't real cores either. You can't run 8 simultaneous full-core threads on a 4790K any more than an AMD Bulldozer. At some point you overload the shared units.

At the end of the day there's no legal claim to be made, any more than there was to the old "performance equivalent" marketing ratings like Athlon XP 1800+ and so on.

Paul MaudDib fucked around with this message at 13:44 on Nov 9, 2015

WhyteRyce
Dec 30, 2001

I may not know how to define a core, but I know it when I see it

1gnoirents
Jun 28, 2014

hello :)

Paul MaudDib posted:

There's no legal meaning to "core" - Bulldozer does have 8 integer execution units, and the 286 had exactly 0 FPUs per integer core. Bulldozer can execute 8 completely separate threads at a time - as long as they're logical or integer operations, rather than a logically-blocking FPU load on every core.

Intel "logical cores" aren't real cores either. You can't run 8 simultaneous full-core threads on a 4790K any more than an AMD Bulldozer. At some point you overload the shared units.

At the end of the day there's no legal claim to be made, any more than there was to the old "performance equivalent" marketing ratings like Athlon XP 1800+ and so on.

Yeah but the difference is Intel specifically never called them cores. Well except for a real 8 core Intel of course that probably costs over $1000.

I get it, I really do, and precisely from their standpoint of consumers (how many times do you hear the phrase "but this has 8 cores and is cheaper..." that made sales), but the timing is just lame. They have too many real problems.

Lowen SoDium
Jun 5, 2003

Highen Fiber
Clapping Larry

Paul MaudDib posted:

There's no legal meaning to "core" - Bulldozer does have 8 integer execution units, and the 286 had exactly 0 FPUs per integer core. Bulldozer can execute 8 completely separate threads at a time - as long as they're logical or integer operations, rather than a logically-blocking FPU load on every core.

Intel "logical cores" aren't real cores either. You can't run 8 simultaneous full-core threads on a 4790K any more than an AMD Bulldozer. At some point you overload the shared units.

At the end of the day there's no legal claim to be made, any more than there was to the old "performance equivalent" marketing ratings like Athlon XP 1800+ and so on.

Intel, nor anyone else, has ever claimed that hyperthreading is 2x cores.

I don't think that this lawsuit has much of a chance of winning, but I personally think that AMD would have been more truthful if they had marketed their chips based on the number of modules rather than cores since they share more than just the FPU cores. They also shared the fetch, decode, and the L2 cache. Basically, the only things they have 2 of is the L1 chance and the integer units. In a lot of ways, it makes it more like hyperthreading unless it's 2 integer executions scheduled per module.

WhyteRyce
Dec 30, 2001

Didn't AMD at one point consider marketing their APU as an additional core or module?

Not Wolverine
Jul 1, 2007

That sheep tube looks super comfy, line it with some sherpa and I would totally rock that in cold weather
Does the shared fetch/decode matter? I was under the impression those operations took so little time it will hardly effect CPU performance. Really, I can see this as two 1/2 or 3/4 cores. I guess it sucks if your doing a lot of floating point math, but it's a ridiculous lawsuit.

Adbot
ADBOT LOVES YOU

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!

Crotch Fruit posted:

Does the shared fetch/decode matter? I was under the impression those operations took so little time it will hardly effect CPU performance. Really, I can see this as two 1/2 or 3/4 cores. I guess it sucks if your doing a lot of floating point math, but it's a ridiculous lawsuit.

The width of the floating point was also increased from Steamroller up so they behaved more like independent cores as well, IIRC. Will still get bogged down if there are two many floating point operations, but so won't an Intel CPU, the difference is degree.

Really, I could see the biggest fallout from this case being a redefining of what counts as an x86-64 core, in which case it might open doors.

  • Locked thread