|
i thought the comment was fine/funny, maybe relax a bit and keep godwin's law on the shelf man
|
#
?
Jan 16, 2021 03:31
|
|
|
|
| # ? May 8, 2024 14:53 |
|
|
Man I do feel bad I was completely dismissive of the guy who played the Hitler card in the Intel CPU thread
|
|
#
?
Jan 16, 2021 03:36
|
|
|
I'm the guy in the meeting room who thought Pat was talking about Starbucks, not Apple. Seriously, what has Pat done? He is an EE who lead the 486 team, but frankly that was a loooong rear end time ago. Has he designed good child since then? Maybe my memory is bad, but back in the 486 days AMD was still trying to get by as a clone maker. AMD pushed the limits of the platforms by releasing chips that were slightly faster, like a 586 that worked on a 486 board. AMD released the K-6 with lower frequency but better IPC and it might have been less if a failure if more than 2 people coded for 3D Now!. AMD broke the GHz barrier with the K7 (still the best CPU of all time, in my opinion). I know Intel has had some good chips too (not you, Pentium 4), but I think Intel was also helped a lot by market share, mind share, anti competitive practices, and even unfair compiler tricks (at least one time). Maybe Pat is smart, I'm just not seeing a compelling reason to fawn over the guy holding a handful of Itanic CPUs. I used to run a dual Pentium Pro system, it was a turd. Not Wolverine fucked around with this message at 15:01 on Jan 16, 2021 |
|
#
?
Jan 16, 2021 05:57
|
|
|
IIRC Pat was the CTO for basically the entire era in which Intel did more than just micro-iterating on Sandy Bridge for a decade.
|
|
#
?
Jan 16, 2021 06:12
|
|
|
He’s got a good reputation if that’s worth anything. Better than Murthy or BK when they started. A lot of old timers/senior guys speak highly of him and I know a lot of people were lamenting his departure at the time. And if you believe the talk, he was hand groomed by Andy Grove or whoever to eventually take over. But Paul bungled that and Intel ended up going down the BK route that hosed everything up. So to some people Pat represents a major what if and the point where Intels once vaunted hire from within chain of succession broke But yeah this is all just talk and not necessarily evidence. I don’t remember my history well, was Larabee Pats thing as well?
|
|
#
?
Jan 16, 2021 06:17
|
|
|
Mr. Smile Face Hat posted:Then why did they get into the situation they're in now? This was all predictable and telegraphed years in advance. A couple of reasons: (1) They've been lead for a bit now by businessmen, not engineers. Said businessmen were not overly concerned that 10nm and below weren't working well because 14nm+++++++ was still making hilarious amounts of money. Seriously--as a company Intel has had some of their most profitable quarters ever on a process that by all rights should have been put out to pasture by a few years back. (2) #1 lead, directly or indirectly, to what appears to have been a far too lax attitude on fixing 10nm and below. If AMD had been reasonably competitive 4-5 years ago, things might be different, but it seems like leadership basically believed that they had all the time in the world to fix their poo poo, and hey, no rush--we're making bank in the meantime! (2.5) There was outright lying and a whole lot of misinformation / lack of attempting to get accurate information going on about the progress the nodes were making. (3) I honestly don't think anyone really expected Apple to get the M1 to work so well right out of the gate. If you'd have told me "yeah, the M1 crushes it in native benchmarks and poo poo specifically written to take advantage of it, but falls on its dumb face trying to emulate x86 poo poo" I'd have absolutely believed you without a second thought. To have managed not only a great hardware implementation but an extremely impressive compatibility layer at the same time is quite an accomplishment. I mean it's one thing to know that Apple has a ton of money and gets all sorts of advantages by owning their product stack top to bottom, it's another thing to see it executed so well. So, yeah. A combo of making a few bad technical decisions combined with a business-first mindset rather than a technology-first mindset, compounded by a loooooot of institutional arrogance. I don't think anyone there is unaware that they've been beaten on the technological front by both AMD and Apple at this point--the landscape today is considerably different than it was a year ago. They're still raking in cash on production volume, but I highly doubt they're unaware that that situation will not last forever without corrective action on their part. Mr. Smile Face Hat posted:Public Relations 101: Don't start off your new position with something that makes you look dumb, even if accurately quoted. (Referring to the "lifestyle company" quote, not the "good" one.) I have never heard an interesting, inspiring speech where every single sentence was so PR-perfect that it could stand on its own, taken entirely out of any surrounding context, and offer no one any issue. "We should be able to do this poo poo better than someone who doesn't even do it full-time" isn't really a particularly contentious concept. DrDork fucked around with this message at 10:23 on Jan 16, 2021 |
|
#
?
Jan 16, 2021 10:21
|
|
|
DrDork posted:I have never heard an interesting, inspiring speech where every single sentence was so PR-perfect that it could stand on its own, taken entirely out of any surrounding context, and offer no one any issue. "We should be able to do this poo poo better than someone who doesn't even do it full-time" isn't really a particularly contentious concept. Eh, it's still not great, and the reason has nothing to do with PR. The problem is the dismissive "lifestyle company" line. Intel's best and brightest engineers want leadership which understands the true state of things, and has a sane plan for dealing with it. They don't want bravado and bluster. They've probably had their fill. They've been watching that so-called lifestyle company do an incredibly good job of putting its CPU engineering teams in a position to succeed over the past decade, while their own management has done the opposite. So, some of them probably rolled their eyes a bit at that. Not saying it's a fatal mistake, Gelsinger has a lot of credibility inside Intel, but it does seem like a misstep.
|
|
#
?
Jan 16, 2021 13:16
|
|
|
WhyteRyce posted:I think he's highlighting that a lifestyle company is dunking all over them and their core business and phrased it in a way to show how ridiculous it is Yeah, the dude was the CEO of VMware, it’s not like he doesn’t know what a Mac is This isn’t news, even by “slow day at the Verge” standards BlankSystemDaemon posted:And he could've phrased it like that, or just better than how he did. lol
|
|
#
?
Jan 16, 2021 14:53
|
|
|
Kazinsal posted:IIRC Pat was the CTO for basically the entire era in which Intel did more than just micro-iterating on Sandy Bridge for a decade. I was looking at used E-ATX motherboards for a Supermicro chassis I have just for fun, and it's astonishing to me how it seems to match exactly on the curve that you would expect from Intel. I was also looking at the newest from the competition which has the same number of threads in a similar NUMA domain configuration (because of the chiplet design) but higher clock rates, and that just seems out-of-the-ballpark, so Intel shouldn't have rested as heavily on those laurels.
|
|
#
?
Jan 16, 2021 15:56
|
|
|
Is the intel memory controller also sensitive to single rank vs dual rank like zen3? Say I wanted 128gb RAM with an 11th gen CPU, should I take 4 x 32gb single rank which are hard to find, or could I do dual rank and not lose performance?
|
|
#
?
Jan 16, 2021 16:08
|
|
|
Kazinsal posted:IIRC Pat was the CTO for basically the entire era in which Intel did more than just micro-iterating on Sandy Bridge for a decade. He also did miracles at vmware, the guys resume ain't bad
|
|
#
?
Jan 16, 2021 17:37
|
|
|
Also, to me, Apple has been a serious silicon designer since the... A4? Obviously they're not at the scale that Intel is, but to me it's also dismissive since there are a lot of talented architectural designers at Apple, and right now that team is responsible for a very powerful SoC that is getting all the big headlines and positive reviews, while Intel is stuck on 14 nm++++++++++++++++++++++++++++ except for small volumes of mobile processors that are basically now stuck on 10nm. So I think it's a good hire and hopefully a "back to engineering focus" hire, but let's not poo poo on Apple because of Intel's several year mismanagement.
|
|
#
?
Jan 16, 2021 18:06
|
|
|
Not Wolverine posted:Seriously, what has Pat done? He is an EE who lead the 486 team, but frankly that was a loooong rear end time ago. Has he designed good child since then? Maybe my memory is bad, but back in the 486 days AMD was still trying to get by as a clone maker. AMD pushed the limits of the platforms by releasing chips that were slightly faster, like a 586 that worked on a 486 board. AMD released the K-6 with lower frequency but better IPC and it might have been less if a failure if more than 2 people coded for 3D Now!. AMD broke the GHz barrier with the K7 (still the best CPU of all time, in my opinion). I know Intel has had some good chips too (not you, Pentium 4), but I think Intel was also helped a lot by market share, mind share, anti competitive practices, and even unfair compiler tricks (at least one time). Maybe Pat is smart, I'm just not seeing a compelling reason to fawn over the guy holding a handful of Itanic CPUs. I used to run a dual Pentium Pro system, it was a turd. In the late 80s to mid 90s, Intel's main competition wasn't x86 clone makers, it was all the new RISC-ish designs that dominated high-end workstation/server markets and were threatening to move into consumer and small business segments as well. There were a lot of different options for "who's going to be the architecture of the next decade," from the POWER family (including a big market push from one Cupertino "lifestyle company" that has for some reason been heavily involved in technology for a very long time), MIPS, Alpha, and even internal competition from the i860/i960 (fun fact: Windows NT was originally developed for the i860, then shifted to MIPS; it was ported to x86 later). The fact that x86 not only stuck around on the strength of backwards compatibility, but managed to move into high-end workstations, servers, and supercomputers, says a lot about the skill of the people working on x86 designs in that era. Also, the Pentium Pro might not have been a great consumer CPU at launch, but it was a key part of that evolution. P6 was where Intel started to break down x86 instructions into RISC-like uops, which was a massive change in the design of the processor. If you chart the evolution of current Intel x86 designs, you can more or less draw a straight line from their current parts that ends at the first P6 chips (and completely bypasses netburst, lol). DrDork posted:Yeah, that is what I took from it: Apple's silicon design teams have clearly been anything but "second-class-citizens" for a long time now. They're one of the most cash-rich companies in the world and they've been willing to throw near-unlimited resources at becoming a serious player in the market.
|
|
#
?
Jan 16, 2021 18:25
|
|
|
Space Gopher posted:Apple's silicon design teams have clearly been anything but "second-class-citizens" for a long time now. They're one of the most cash-rich companies in the world and they've been willing to throw near-unlimited resources at becoming a serious player in the market. Their processor design teams are clearly very talented, yeah, I don't think anyone can argue against that considering what they've been able to put out. But they're second-class-citizens in the sense that they only exist as a means to an end for other products. If Intel was on 5nm by now with good volume as their roadmaps from 2014 or whatever said they should be, there's a real possibility Apple might not have bothered to invest as much as they have into a x86 replacement. Apple doesn't want to be a chip design company. They want to continue to be a luxury device company, and they'll put whatever makes the most sense inside said devices to meet their needs. Doing their own chip design makes sense to further that end, but makes little sense as an end unto itself unless Apple changes their mind and starts licensing their designs out.
|
|
#
?
Jan 16, 2021 18:58
|
|
|
There is more than to it than just designing CPU. Validation, chipset poo poo, I/O, integration, etc. They have been on a hiring spree to trying to grab the expertise and experience they need because they decided to go this route. All that is stuff you have to build up, stuff Intel already knew the recipe for and had but pissed away the advantage of
|
|
#
?
Jan 16, 2021 19:12
|
|
|
WhyteRyce posted:There is more than to it than just designing CPU. Yeah, turns out when you have basically unlimited money and are paying better wages than the competition you can collect all sorts of smart people to build up that ecosystem for you. It'll be real interesting to see what (if anything) Intel does to try to stop the talent bleed.
|
|
#
?
Jan 16, 2021 19:29
|
|
|
DrDork posted:But they're second-class-citizens in the sense that they only exist as a means to an end for other products. If Intel was on 5nm by now with good volume as their roadmaps from 2014 or whatever said they should be, there's a real possibility Apple might not have bothered to invest as much as they have into a x86 replacement. Apple ditched Intel, because Intel simply cannot compete when it comes to power/clock-efficiency, when contrasted with ARM - and Apple makes mobile devices, where power efficiency matters. The fact that Apple then threw enough money at the problem to hire people to design them a chip+software ecosystem that can beat Intel (well, with the possible exception of HPC software compiled using Intels own compiler), that's just icing on the cake that shows how Intel could do better. I hope Intel does do better, so we can have competition.
|
|
#
?
Jan 16, 2021 19:41
|
|
|
I know quite a few Intel storage division folks who had the option to go to SK in the purchase but most stayed, interestingly enough. Sounded like the signing bonuses to move were pretty dang good too!
|
|
#
?
Jan 16, 2021 19:44
|
|
|
Ika posted:Is the intel memory controller also sensitive to single rank vs dual rank like zen3? Say I wanted 128gb RAM with an 11th gen CPU, should I take 4 x 32gb single rank which are hard to find, or could I do dual rank and not lose performance? Does anyone even make single rank 32 GB sticks? Even my (granted, 3 years old) 16 GB sticks had to be dual rank.
|
|
#
?
Jan 16, 2021 19:46
|
|
|
Space Gopher posted:Apple's silicon design teams have clearly been anything but "second-class-citizens" for a long time now. They're one of the most cash-rich companies in the world and they've been willing to throw near-unlimited resources at becoming a serious player in the market. I don’t know that you get somebody like Jim Keller to build and staff up your chip operation if you’re not thinking of it as a “first-class” part of your business strategy
|
|
#
?
Jan 16, 2021 19:56
|
|
|
BlankSystemDaemon posted:Apple didn't ditch x86 because "Intel bad", even though Intel did rest on their laurels. You're right that they were never gonna go Intel/x86 for a phone SoC or even a tablet. But if Intel had kept ahead of TSMC and thus was currently chilling out on a performant sub-5nm process right now, there's a very real possibility that they'd be able to have provided laptop chips with similar power efficiency to the M1 and without the need for the (enormously expensive to develop, I'd have to assume) compatibility layer. I mean, path not taken and all that, but a lot of Intel's dominance over the years has been due to their node advantages. Without it...yeah, not doin' so hot. RISC on top of that is just a kick in the balls. BlankSystemDaemon posted:I hope Intel does do better, so we can have competition. AMD starting to crank prices as soon as they caught up to Intel was both entirely expected and yet disappointing. Given the chip scarcity worldwide, though, I'm not sure we're likely to see real price-based competition for consumer parts return for a while, even if 11th gen chips are legitimately competitive.
|
|
#
?
Jan 16, 2021 19:56
|
|
|
Indiana_Krom posted:Does anyone even make single rank 32 GB sticks? Even my (granted, 3 years old) 16 GB sticks had to be dual rank. At least one kit is listed under the single rank filter (Kingston HyperX Fury black DIMM kit 128GB, DDR4-3600, CL18-22-22 (HX436C18FB3K4/128) ), but that could also be a listing mistake since the details show single rank, x8, while dual rank kits show dual rank, x8.
|
|
#
?
Jan 16, 2021 20:07
|
|
|
DrDork posted:Yeah, turns out when you have basically unlimited money and are paying better wages than the competition you can collect all sorts of smart people to build up that ecosystem for you. Improve compensation would probably be the top thing to do with cutting out the politics that run rampant second. But yeah it's going to be tough when Apple is continually fishing in your pond for good HW engineers. Better compensation could put a real dent in that. priznat posted:I know quite a few Intel storage division folks who had the option to go to SK in the purchase but most stayed, interestingly enough. Sounded like the signing bonuses to move were pretty dang good too! It really depends on your situation. SKHynix has a headquarters elsewhere. They already have their own internal teams (ASIC, FW, etc.) that you could argue are better than Intel's. SKHynix uses replacement gate, Intel is the only one left on floating gate. There is a lot of redundancy. Apparently SKHynix promised that they intend to keep the Folsom site strong and staffed but looking at that overlap you have to wonder how long that will be the case. How is the merger going to work? I highly doubt they'd want two separate, independent teams who rarely talk or interact with one another. And once you move, you're in there and probably not as freely able to move around. Contrast that with staying at Intel, you can easily bounce around some of the various core competency groups that maybe feel a little more certain to you. Yeah, the signing bonus was nice. But you're giving up your Intel RSUs right? And I thought SKHynix doesn't have an employee stock program
|
|
#
?
Jan 16, 2021 20:23
|
|
|
DrDork posted:Their processor design teams are clearly very talented, yeah, I don't think anyone can argue against that considering what they've been able to put out. It’s going to really really matter for Apple when they start branching out into other wearables (glasses, etc) and poo poo like cars, which they apparently have never stopped secretly working on.
|
|
#
?
Jan 16, 2021 20:31
|
|
|
WhyteRyce posted:Improve compensation would probably be the top thing to do with cutting out the politics that run rampant second. But yeah it's going to be tough when Apple is continually fishing in your pond for good HW engineers. Better compensation could put a real dent in that. Yup I think that was the reasoning to stay. This is a Canadian satellite office and fairly small, and SK is going to share office space for a couple years. They’re ramping up their local hiring too which is great for the area. For the longest time Intel was only interested in hiring coops for 4/8 month terms and then hiring or not after that if they had graduated. Now the SK group seems interested in hiring more experienced folks too which bodes well.
|
|
#
?
Jan 16, 2021 20:31
|
|
|
Ok Comboomer posted:I don’t know that you get somebody like Jim Keller to build and staff up your chip operation if you’re not thinking of it as a “first-class” part of your business strategy I don't think the plan would work if they didn't put together a first class team, even if the business unit is second class in the master plan. But if you have a first class team with big wins, I don't see it staying second class for long.
|
|
#
?
Jan 16, 2021 21:17
|
|
|
BlankSystemDaemon posted:
I think I posted something like this in the Mac HW thread but what I respect Apple for in this is using their power to smack their entire team from silicon designers to compiler devs to the OS and hardware folks and align that entire vertically integrated capability to deliver. That’a the biggest roadblock I see for MSFT or others trying to do the same; FFS even with the Surface, it had laughable ACPI / low-level quirks that ownership and control of the ENTIRE loving PLATFORM should have prevented. Like... having one building where you could hear “hey, if you put that instruction in the CPU, my compiler work gets easier, the OS guy can do X and oh, it’ll make the HW designers’ lives easier if you add X and Y to the PHY so we can do one SKU for all our devices...” and the leadership to make that happen without in-fighting is pretty solid. And Jonny Ive is gone, so may something like the Touch Bar never happen again, so say we all. movax fucked around with this message at 21:42 on Jan 16, 2021 |
|
#
?
Jan 16, 2021 21:40
|
|
|
BlankSystemDaemon posted:Well, they've made advances now, so now they can fall back on their laurels, surely. Threadripper 3000 and Epyc are some wild hardware. 7nm lets Epyc squeeze in some big core counts at pretty decent clocks by server standards and Threadripper even clocks them high. The attractiveness of using old server parts has really dropped in comparison imo, the new stuff is clearly generationally better if you can afford it. TRX40 doesn’t do RDIMMs but Epyc does and the single-socket models are quite affordable (there is an 8C model that’s $450 MSRP). If you want gaming clocks and RDIMMs you have to go WRX80 though.
|
|
#
?
Jan 16, 2021 22:59
|
|
|
Tom's Hardware: Retailer Reveals Rocket Lake-S, Comet Lake Refresh Pricing TL;DR: i9-11900K: 604 USD i9-11900KF: 575 USD for comparison, the same retailer sells the i9-10900K for 555 USD i7-11700K: 455 USD i7-11700KF: 426 USD for comparison, the same retailer sells the i7-10700K for 398 USD i5-11600K: 293 USD i5-11600KF: 265 USD for comparison, the same retailer sells the i5-10600K for 282 USD ___ This is just my own personal addition to the story, but keep in mind that the Ryzen 9 5900X has an SRP of 549 USD the Ryzen 7 5800X has an SRP of 449 USD the Ryzen 5 5600X has an SRP of 299 USD Obviously there's still variance to these numbers, but it doesn't look all that good for Intel: even if we assume that people who are going to go with Ryzen 5000 are going to buy new boards just to avoid BIOS support issues, a B550 is still going to be cheaper than a Z590, so if the i5-11600KF is only just barely cheaper than the Ryzen 5 5600X for about the same performance, I don't see the value proposition in it. the i7 might be worth looking at relative to the 5800X, but the i9 is dead-in-the-water at a price that's competing with the 5900X for something that has gradenko_2000 fucked around with this message at 08:05 on Jan 18, 2021 |
|
#
?
Jan 18, 2021 08:00
|
|
|
gradenko_2000 posted:Tom's Hardware: Retailer Reveals Rocket Lake-S, Comet Lake Refresh Pricing Combine that news with this: https://www.guru3d.com/news-story/premium-z590-motherboards-to-become-brutally-expensiveone-mobo-even-passes-1500-eur.html ...and I'm starting to become very worried about the future of PC gaming. I'm seriously considering buying an FTW3 Ultra 3090, and I joked that getting the extended warranty on it, if it fails in 7-9 years, having it be that the 'equivalent' card performance/spending-wise then is the x50 part.
|
|
#
?
Jan 18, 2021 08:09
|
|
|
I wouldn't go so far as to say "I'm worried about the future of PC gaming" - a Ryzen 3100/3300X or Intel's i3-10100 (or the refreshed i3-10105) is still perfectly capable of running most games, at a price point that's usually well below 200 USD, and even a Ryzen 3600 / i5-10400 is going to let you drive just about anything up to the point where you're GPU-limited. If anything, the star of the show might well be the i5-11400F - if it can compete with the 5600X in gaming workloads even without overclocking* capability, that might be seriously good value for something that costs a third less than AMD's offering. ___ * what Intel should have done IMO even since 10th gen was to make a K-model for the i3 quad-cores.
|
|
#
?
Jan 18, 2021 08:52
|
|
|
I literally cannot see why anyone would buy an 11900K. An 11700KF is 30% less expensive, has the same core and thread count, and the 200 MHz difference in boost clocks can easily be gained with 30 minutes of fiddling about in the BIOS and Prime95. Anyone looking at an 11900K doesn't care about integrated graphics, so chopping off another thirty bucks going from 11700K to KF is a no-brainer.
|
|
#
?
Jan 18, 2021 09:00
|
|
|
When has the top bin segment ever made sense for Intel CPUs. It only does for 'price insensitive' consumers.
|
|
#
?
Jan 18, 2021 13:34
|
|
|
Speaking of companies poaching Intel engineers: https://www.bizjournals.com/portland/news/2020/12/31/microsoft-corp-lease-85-000-square-feet-hillsboro.html
|
|
#
?
Jan 18, 2021 13:35
|
|
|
Beef posted:When has the top bin segment ever made sense for Intel CPUs. It only does for 'price insensitive' consumers. This is different, the previous gen's i9 gave you two extra cores. No more. Maybe it was pointless for most games, but was of course good for productivity
|
|
#
?
Jan 18, 2021 14:24
|
|
|
Beef posted:Speaking of companies poaching Intel engineers: I saw that and wept for cornell road
|
|
#
?
Jan 18, 2021 16:29
|
|
|
Tiger Lake (Panther Canyon) NUCs announced... if I recall correctly, the AliExpress specials usually come in $200-$300 cheaper, right? Though, it'll probably be awhile until that happens since I think most of those are using some excess supply run of Skylake-era parts. This one (NUC11TNKV7 intrigues me because of dual NIC, but it'll probably be close to $1000.
|
|
#
?
Jan 18, 2021 19:35
|
|
|
movax posted:Tiger Lake (Panther Canyon) NUCs announced... if I recall correctly, the AliExpress specials usually come in $200-$300 cheaper, right? Though, it'll probably be awhile until that happens since I think most of those are using some excess supply run of Skylake-era parts. These are a harder sell vs a Mac Mini now. I'm curious where pricing lands, because I'd like to upgrade from a Haswell 4250U Nuc that I use as a home server, but I also don't want to spend much.
|
|
#
?
Jan 18, 2021 20:18
|
|
|
Twerk from Home posted:These are a harder sell vs a Mac Mini now. I'm curious where pricing lands, because I'd like to upgrade from a Haswell 4250U Nuc that I use as a home server, but I also don't want to spend much. They really are, what fresh hell. Like I’m at the point where I’d rather freshen up an old SFF optiplex for the handful of strategy games I still play on Steam that won’t play nice with Crossover and have M1 Mini take Lead Desktop role Edit: you could probably snag a more performant 4th gen i5 or an i7 for like $100 in used optiplex form, probably less if you go with USFF and forego any pretension of adding a dGPU. If you’re willing to spend $200-$400 you can snag 6th gen or later. trilobite terror fucked around with this message at 20:27 on Jan 18, 2021 |
|
#
?
Jan 18, 2021 20:22
|
|
|
|
| # ? May 8, 2024 14:53 |
|
|
I have to wonder if there's any market for NUCs. My brother got a current gen one for casual gaming on the TV but gave up on it after realizing how anaemic its GPU capabilities were. He idly considered hooking up an eGPU to it but at that point, you're better off getting a normal desktop
|
|
#
?
Jan 18, 2021 22:38
|
|
