|
BangersInMyKnickers posted:Buldozer was an absolute joke, but Zen is pretty good if you're will to trade off clock speed for more cores and Zen2 looks like its going to close the gap real fast intel is meeting them in the middle
|
# ? May 24, 2019 09:53 |
|
|
# ? Apr 27, 2024 12:16 |
|
BangersInMyKnickers posted:Buldozer was an absolute joke, but Zen is pretty good if you're will to trade off clock speed for more cores and Zen2 looks like its going to close the gap real fast
|
# ? May 24, 2019 13:55 |
|
Budzilla posted:I remember a while back there was an Intel slide implying that 7nm from 'other leading foundries' was ~10% better compared to their 10nm. That doesn't matter because Intel won't have an 10nm part this year. They're stuck on 14nm for a long while. 10nm intel is never coming out, we have to too toward 7nm intel for them to do a node jump.
|
# ? May 24, 2019 14:42 |
|
wargames posted:10nm intel is never coming out, we have to too toward 7nm intel for them to do a node jump. Meanwhile TSMC is already starting N7+ volume production https://twitter.com/chiakokhua/status/1131931279602069504
|
# ? May 24, 2019 18:26 |
|
wargames posted:10nm intel is never coming out, we have to too toward 7nm intel for them to do a node jump. Also 7nm won’t increase clock speed, just reduce power and increase density, IPC should improve but we might never see 6Ghz.
|
# ? May 24, 2019 19:23 |
|
Perplx posted:Also 7nm won’t increase clock speed, just reduce power and increase density, IPC should improve but we might never see 6Ghz. Bring back Prometeia.
|
# ? May 24, 2019 19:51 |
|
Perplx posted:Also 7nm won’t increase clock speed, just reduce power and increase density, IPC should improve but we might never see 6Ghz. I'm honestly kinda fine with this--I'm still rocking a 5820k, and honestly it's plenty fast enough for anything I've ever needed it to do. I would really appreciate being able to eek out a few extra hours from my laptop, though, so any power reduction is quite welcomed.
|
# ? May 24, 2019 20:04 |
|
Perplx posted:Also 7nm won’t increase clock speed, just reduce power and increase density, IPC should improve but we might never see 6Ghz. anything that reduces power on intel part would be great. Also if they can fix their security while they are at it thanks.
|
# ? May 24, 2019 20:06 |
|
DrDork posted:I'm honestly kinda fine with this--I'm still rocking a 5820k, and honestly it's plenty fast enough for anything I've ever needed it to do. I would really appreciate being able to eek out a few extra hours from my laptop, though, so any power reduction is quite welcomed. tbh an official non-ES 1660v3 is already down to $250, I'm actually hoping that this might be the straw that finally gets server farms to dump their Haswell-E in bulk and those prices nosedive even further.
|
# ? May 24, 2019 20:34 |
|
MaxxBot posted:Meanwhile TSMC is already starting N7+ volume production 5nm volume in Q1 2020, loving wild.
|
# ? May 24, 2019 21:00 |
|
Huh. If Taiwan is so far ahead, why are Intel and AMD processors the norm?
|
# ? May 24, 2019 21:40 |
|
Instructions take 2 characters in Mandarin so their CPUs are half the speed they should be.
|
# ? May 24, 2019 21:52 |
|
surf rock posted:Huh. If Taiwan is so far ahead, why are Intel and AMD processors the norm? AMD is using TSMC for their new CPUs launching very soon, Intel uses their own fabs.
|
# ? May 24, 2019 21:59 |
|
surf rock posted:Huh. If Taiwan is so far ahead, why are Intel and AMD processors the norm? There's a lot that goes in to CPU performance other than just node size. Another factor to consider is that listed node size does not always directly translate 1:1 between fabs the way you'd think, because not everything in a chip is actually produced at the smallest size the node is capable of running at. In the past TSMC's nodes have generally been considered behind Intel's by one or two, eg their 10nm was producing actual results roughly equivalent to Intel's 14nm node. We shall see if this continues, though. But, yeah, they produce a poo poo ton of chips for applications other than general purpose desktop CPUs--ARM CPUs, for example, massively out umber Intel ones in global application, because they make up the brains of your phones, switches, IoT, etc. So in many ways they are the norm.
|
# ? May 25, 2019 02:03 |
|
Another factor that has traditionally been different is that Intel optimizes for clock speed and TSMC for low power and high density. This made Intel’s version of a node the best (by a very large margin too) if you wanted to make a hot chip which ran real fast. I don’t think those process design priorities have truly changed, but it’s quite possible they don’t matter as much anymore. It’s so easy to achieve far too high a power density to effectively cool a chip that even Intel has been on a decade+ long trend of improving performance by making their processor cores more power efficient (thus allowing them to extract more performance from whatever power density limit they choose to design against). Having the industry leading Ids(on) logic transistor in your cell library doesn’t matter as much when industry wide design trends favor using smaller, more efficient transistors.
|
# ? May 25, 2019 02:58 |
|
DrDork posted:a very good post AMD and Intel for a long time were the only ones who could produce x86 ISA CPU's and x86 CPU's that are saleable around the world and they are the de facto standard in the PC world for a variety of factors. These days AMD can't make their own CPU's anymore (they sold/spun off their fabs into GF because Bulldozer was a huge flop that financially almost sunk the company) so they have TSMC and GF do it. Historically there were other x86 compatible CPU designers and makers (VIA, Cyrix, IDT/Centaur, etc) but they weren't able to keep up with the development and products of either Intel or AMD and have either disappeared, got bought out (pretty much all the later VIA x86 CPU's were really Centaur designs for example, I think VIA got the whole Centaur team in a buy out), or focused instead on other markets (ie. embedded, the Geode SOC's produced by Nat. Semi were really Cyrix MediaGX (which was really just a 5x86!!) SOC's for instance until AMD bought the line from them) when they couldn't compete in the desktop/server/laptop space anymore. AMD did sell their 1st gen Zen design to a Chinese company a while back so technically there still are non-AMD/Intel producers of modern high performing x86 CPU's but they're not allowed to sell it outside of China per the legal agreements. Besides the obvious gigantic technical and financial issues of building a x86 CPU that can compete with Intel/AMD's latest and greatest the legal and patent issues of doing so make it just about impossible for any other competitor to pop up again like they did in the 80's and 90's.
|
# ? May 25, 2019 06:28 |
|
surf rock posted:Huh. If Taiwan is so far ahead, why are Intel and AMD processors the norm? In addition to what's already been said, TSMC is a pure semiconductor foundry - they manufacture chips that their customers design. They don't design anything in-house. It's one of their core business principles to never compete with their customers, because if they did they'd never be trusted with any trade secret chip designs again.
|
# ? May 25, 2019 17:07 |
|
priznat posted:I’m shocked that they figure the current type CEM connectors are good enough for those rates. Maybe with a 2” trace (on megtron 6)!! For real. We're getting into serious territory for serdes though given 28gbaud and 56gbaud is par for the course now in ethernet implementations, it might just be reasonable for high end servers and motherboards.
|
# ? May 25, 2019 17:37 |
|
K8.0 posted:Instructions take 2 characters in Mandarin so their CPUs are half the speed they should be.
|
# ? May 25, 2019 17:39 |
|
Well this is new https://adoredtv.com/ Announced here: https://www.youtube.com/watch?v=gDeTruZOC8Y This guy really doesn't like Intel TBH, but it's pretty interesting nonetheless.
|
# ? May 26, 2019 14:58 |
|
Adored is a significantly worse version of wccftech and the only reason you should consume his content is if you've run out of parody articles to read and are hungry for more.
|
# ? May 26, 2019 15:36 |
|
I don't watch him, but what has he put out that's so inaccurate? E - nm, I listened to three minutes of that and that's enough. GRINDCORE MEGGIDO fucked around with this message at 16:24 on May 26, 2019 |
# ? May 26, 2019 16:22 |
|
Intel's first to 5ghz all cores.
|
# ? May 26, 2019 17:25 |
|
i bet they held back top bin 9900K examples for a while, 8086K all over again. also this:
|
# ? May 26, 2019 21:32 |
|
Intel announces inverse bitcoin: pay money and manually solve math problems at 5 GHz to generate electricity and cool your PC.
|
# ? May 26, 2019 22:51 |
|
Vintage computing, a room full of people doing a lot of math problems by hand
|
# ? May 26, 2019 23:49 |
|
K8.0 posted:The big stuff is within your window. Zen2 in a few months and then Intel hopefully combining a significant process leap with some architectural improvements. After that it should all be incremental for a while. Is this the Intel thing that was coming up? It sounds pretty impressive. edit: it's fun to compare a top current desktop processor against my six-year-old laptop processor. Is there a Zen2 processor I could add to the comparison? I don't know what the numbers mean but I like that they're mostly very different. surf rock fucked around with this message at 00:14 on May 27, 2019 |
# ? May 27, 2019 00:04 |
|
surf rock posted:Is this the Intel thing that was coming up? It sounds pretty impressive. No, that intel chip is just an older model at a higher speed. Its not right saying its factory overclocked, but its functionally almost the same.
|
# ? May 27, 2019 01:29 |
|
Malcolm XML posted:For real. We're getting into serious territory for serdes though given 28gbaud and 56gbaud is par for the course now in ethernet implementations, it might just be reasonable for high end servers and motherboards. Well, AMD just shrugged and casually dropped PCIe 4.0 in Ryzen 9x.
|
# ? May 27, 2019 10:37 |
|
GRINDCORE MEGGIDO posted:I don't watch him, but what has he put out that's so inaccurate? lol
|
# ? May 27, 2019 14:31 |
|
I remember how mad he got when (think it was Hardware Unboxed?) who threw cold water on these largely because the pricing made little sense if the performance was going to be as expected, as there would be no reason to undercut Intel so aggressively and throw revenue away. Welp, we got a site to launch!
|
# ? May 27, 2019 14:36 |
|
I wanna know the TDP for the 9900KS. Gonna cook pancakes on top of my PC
|
# ? May 27, 2019 19:38 |
|
There are plenty of reviews out there running 5 GHz 9900k's, though some of them were on incorrect bios settings and never corrected them. The ks is going to be somewhere between stock and those overclocked results. I have no trouble cooling my 5GHz 9900k on stock voltages.
|
# ? May 27, 2019 21:29 |
|
BangersInMyKnickers posted:more cost effective to buy more 2U dual-socket AMD servers even if you need twice as many as 4U quad socket intels. Maybe if you're just looking at the server hardware, but once you factor in the extra switches, PDUs, software licenses, storage, hardware failures and management overhead?
|
# ? May 28, 2019 05:59 |
|
Fabricated posted:I wanna know the TDP for the 9900KS. Gonna cook pancakes on top of my PC Intel is probably going to say it's 95w.
|
# ? May 28, 2019 06:04 |
|
While Intel's definitely got egg on their face for re-treading the 9900K, I really can't wait until Buildzoid looks over the X570 boards and we also get some FLIR imagery of the X570 chipset and why it needs active cooling.
|
# ? May 28, 2019 08:12 |
|
Its already been said by Steve and Buildzoid that heat is an issue if you run M.2 RAID off of the PCIe 4.0 buses on the X570 chipset but otherwise probably won't be a problem. Heat-wise you're looking at like 10-15W max, which is a lot for a chipset these days but not all that much really, so nothing really interesting there to stare at with a FLIR cam either.
|
# ? May 28, 2019 09:38 |
|
It's okay, guys, I still think Intel's..... okay! I mean, like as not, Thunderbolt is still an Intel-only thing until USB 4, and that's not gonna show up until something 2022 at the earliest. I'm probably buying an LG Gram with Ice Lake for my brother still, when those show up. Speaking of Thunderbolt in the intervening period: I, for one, am very glad to see that Intel is doing slightly less stupid things with regards to not hanging Thunderbolt exclusively off the PCH. Though their testing methodology with Gen11 graphics raises some eyebrow-raising questions. SwissArmyDruid fucked around with this message at 12:24 on May 28, 2019 |
# ? May 28, 2019 11:24 |
|
SwissArmyDruid posted:It's okay, guys, I still think Intel's..... okay! I mean, like as not, Thunderbolt is still an Intel-only thing until USB 4, and that's not gonna show up until something 2022 at the earliest. As long as Intel's 7nm products hit in 2021 and are competitive Intel will be fine.
|
# ? May 28, 2019 13:54 |
|
|
# ? Apr 27, 2024 12:16 |
|
PC LOAD LETTER posted:Its already been said by Steve and Buildzoid that heat is an issue if you run M.2 RAID off of the PCIe 4.0 buses on the X570 chipset but otherwise probably won't be a problem. Asus has said it still runs hot even running a single SSD off the PCH.
|
# ? May 28, 2019 15:47 |