|
Pryor on Fire posted:It's really not. Yes there are a few jobs in this field but most of them have gone away in the past 10 years. 90% of "video production" just happens on youtube or on phones nowadays. Well sure, most lovely Youtube videos for people that don't care to make their Drone video or Vlog not suck, but for those that care about their work a little bit, and with the 4K drone video market exploding, I don't think that's entirely accurate.
|
# ? Jan 10, 2017 19:53 |
|
|
# ? May 21, 2024 15:10 |
|
Gwaihir posted:Right, and the question remains, when are you seeing usage like that at home, heh? I think from their perspective it's mostly "it costs us little and there's actually a reason to, so why not?" I looked into whether this could cause any real serious problems, and it seems like its probably can't really. PCIe (at least the spec?) supports QoS, so network or audio devices should be able to take away as much bandwidth as it needs from the SSD. So probably no obvious problems, like audio skips or packet loss. Unless that's not actually implemented, or all the drivers suck. But at any rate, I think adding another 4 CPU lanes to consumer hardware is a very good idea, even if it's not an amazing win. There's a common device type that can use them, so why not.
|
# ? Jan 10, 2017 19:55 |
|
FaustianQ posted:There is always FD-SOI for GloFo and AMD, Of which they have 22nm, 12nm, and 7nm FD-SOI. I have no idea what the hurdles would be moving from finFET to FD-SOI or whether FD-SOI is even suitable for such things, but it's apparently a much less complex process if not more expensive in small volume. Yeah, good luck on getting the required transistor density with planar after going FinFET. That is gonna result in another Fiji-like situation with the die bumping up against that 600nm^2 size again. GloFo are going to try to do some stupid poo poo like... I dunno, V-TFET mixed with GAA and then we'll all be sorry.
|
# ? Jan 10, 2017 20:01 |
|
Does GloFo do business with any other company and product? If so Who/What?
|
# ? Jan 10, 2017 20:02 |
|
EdEddnEddy posted:Does GloFo do business with any other company and product? If so Who/What? They build the actual chips for a tons of companies, including Qualcomm and Broadcom.
|
# ? Jan 10, 2017 20:04 |
|
EdEddnEddy posted:Does GloFo do business with any other company and product? If so Who/What?
|
# ? Jan 10, 2017 21:23 |
|
Notably, chips that are nowhere near as bleeding-edge as what AMD is trying to do.
|
# ? Jan 10, 2017 21:39 |
|
Anime Schoolgirl posted:They make billions from chips (and not just CPUs) for low-end phones that still sell by the ton because smartphones get broken all the loving time Pretty much, nobody gives a poo poo who makes the 32nm 3 year old qualcomm SOC for a $50 retail android smartphone, so GoFlo obviously makes a ton of them. Video editing on anything HD takes god damned forever without fast storage, and when you actually have a lot of video/audio tracks and are doing effects and post-processing work, having a non-potato machine to do it on can make the difference between a video a week and a video every 2 weeks, just on the scrub/preview time alone.
|
# ? Jan 10, 2017 21:46 |
|
https://twitter.com/CPCHardware/status/818429683633127424 6C/12T is a confirmed thing, runs at about 65W, wonder why just two cores seems to make a large difference in power consumption, guess cache?
|
# ? Jan 10, 2017 22:02 |
|
I was able to pick up a mobo + 'cat core Athlon for $44 today ($40 off mobo with CPU purchase). Pared with the 150gb SSD and 4gigs of DDR3 I had lying around, plus a PicoPSU/case combo I grabbed for $20, I got a full HTPC system for less than the cost of a comparable Compute Stick. There are dozens of us out there, using the Athlon 5350 for things. I will probably try and offload this build to my niece in 4 months, but I wonder if Bristol Ridge would match this sort of niche?
|
# ? Jan 10, 2017 22:05 |
|
FaustianQ posted:6C/12T is a confirmed thing, runs at about 65W, wonder why just two cores seems to make a large difference in power consumption, guess cache?
|
# ? Jan 10, 2017 22:05 |
|
FaustianQ posted:https://twitter.com/CPCHardware/status/818429683633127424 This would be amazing just for miniITX cases. All of the i7 4C/8T run too hot for most super small form factor cases when using a small heatsink.
|
# ? Jan 10, 2017 22:10 |
|
Rabid Snake posted:This would be amazing just for miniITX cases. All of the i7 4C/8T run too hot for most super small form factor cases when using a small heatsink.
|
# ? Jan 10, 2017 22:11 |
|
Rabid Snake posted:This would be amazing just for miniITX cases. All of the i7 4C/8T run too hot for most super small form factor cases when using a small heatsink. That thing has gotta be clocked low-ish to get 6 cores into 65W, or AMD really has knocked one out of the park.
|
# ? Jan 10, 2017 22:17 |
|
Twerk from Home posted:That thing has gotta be clocked low-ish to get 6 cores into 65W, or AMD really has knocked one out of the park. Could simply be the difference between Samsung and GloFo
|
# ? Jan 10, 2017 22:18 |
|
Anime Schoolgirl posted:35/45w T processors exist for i7s, but they'll be a bit more expensive than the regular ones This has always driven me up the wall with i5's in particular, but then I'm guessing Intel is trying to prevent C2D shenanigans with people buying at i3 prices and overclocking to get i7 performance, etc. Won't be an issue on AMD processors, so unless you're concerned about overclocking, get the cheapest for the core count you want, get an X300/B350/X370 and overclock. Going to be weird capturing the feeling on 1998-2006 overclocking again.
|
# ? Jan 10, 2017 22:18 |
|
FaustianQ posted:This has always driven me up the wall with i5's in particular, but then I'm guessing Intel is trying to prevent C2D shenanigans with people buying at i3 prices and overclocking to get i7 performance, etc. Unfortunately desktop Skylake remains a better platform than desktop Kaby Lake regardless because mobo manufacturers stopped giving a poo poo and reenabled BCLK tomfoolery the month after the patch so lol
|
# ? Jan 10, 2017 22:22 |
|
Methylethylaldehyde posted:Pretty much, nobody gives a poo poo who makes the 32nm 3 year old qualcomm SOC for a $50 retail android smartphone, so GoFlo obviously makes a ton of them. Just working on my 1440P video from a GoPro goes immensely faster on my Raid-0 EVO 850 SSD setup than it does on Raid-0 2TB WD Blacks (duh) but even with 1GB/s transfer potential, I'm 100% saturating the drives with CPU time to spare. Sure you can do just fine with slow, but there is plenty of consumer CPU power that can utilize super fast storage options for sure.
|
# ? Jan 10, 2017 22:32 |
|
EdEddnEddy posted:Just working on my 1440P video from a GoPro goes immensely faster on my Raid-0 EVO 850 SSD setup than it does on Raid-0 2TB WD Blacks (duh) but even with 1GB/s transfer potential, I'm 100% saturating the drives with CPU time to spare. HD video editing is one of the few things I can think of that Joe home-user can really see the difference between an 850 EVO and a Intel 750 or other PCIe NVMe drive. Without making a SD preview/scrub placeholder file or other form of lower bandwidth witchery, any time you scrub through the video takes a longass time in HD.
|
# ? Jan 10, 2017 22:34 |
|
FaustianQ posted:Won't be an issue on AMD processors, so unless you're concerned about overclocking, get the cheapest for the core count you want, get an X300/B350/X370 and overclock. Going to be weird capturing the feeling on 1998-2006 overclocking again. I want to believe (pencil and spacer not included)
|
# ? Jan 10, 2017 22:40 |
|
eames posted:I want to believe
|
# ? Jan 10, 2017 22:43 |
|
lol, totally forgot about those pencil days and thank god never had a crunch moment
|
# ? Jan 10, 2017 23:42 |
|
I liked how if the HSF wasn't properly seated and functional the chip would melt down after a couple seconds, almost happened to once but I cut the power to the machine just in time.
|
# ? Jan 11, 2017 00:03 |
|
Regardless on how Vega turns out, looks like Zen is going to be a hit https://twitter.com/CPCHardware/status/818932115270209537 My guess is they can easily get a full product stack going with this kind of binning. Do 4/6/8 core models with hyperthreading and all the L3, then do second tier 4/6/8s with no hyperthreading and disabled L3 for the very bottom. Absolutely zero need to involve Bristol Ridge in the line up if volume is high enough, and this absolutely murders Intels low end poo poo. Discontinue in 2018 with Zen+ and Raven. SwissArmyDruid posted:Yeah, good luck on getting the required transistor density with planar after going FinFET. That is gonna result in another Fiji-like situation with the die bumping up against that 600nm^2 size again. I wanted to return to this for a second and ask what the density difference is like? Would it be possible to do Zen on FDX 12nm? Because I know if GloFo and wild claims are a thing in the fab industry, so grain of salt on this quote:GLOBALFOUNDRIES’ new 12FDX technology is built on a 12nm fully-depleted silicon-on-insulator (FD-SOI) platform, enabling the performance of 10nm FinFET with better power consumption and lower cost than 16nm FinFET. The platform offers a full node of scaling benefit, delivering a 15 percent performance boost over today’s FinFET technologies and as much as 50 percent lower power consumption. Makes it sounds like FDX 12nm is better than current 14nm finFET, even if by single digit percentages. I have no idea if GloFos 12nm could do real performance parts but it at least sounds promising for a different segment of chips, like say AMDs K12.
|
# ? Jan 11, 2017 06:28 |
|
SwissArmyDruid posted:Yeah, good luck on getting the required transistor density with planar after going FinFET. That is gonna result in another Fiji-like situation with the die bumping up against that 600nm^2 size again. Isn't FinFET actually worse density with the tradeoff of having better drive current and way less leakage?
|
# ? Jan 11, 2017 07:48 |
|
FaustianQ posted:My guess is they can easily get a full product stack going with this kind of binning. Do 4/6/8 core models with hyperthreading and all the L3, then do second tier 4/6/8s with no hyperthreading and disabled L3 for the very bottom. Absolutely zero need to involve Bristol Ridge in the line up if volume is high enough, and this absolutely murders Intels low end poo poo. Discontinue in 2018 with Zen+ and Raven. I really want to see what Zen can do because Intel's mainstream product stack is getting ridiculous: nothing really makes much sense in there between a $64 G4560 to a $330 7700K. I sure as hell won't buy a 4C/4T chip for $200+ in 2017.
|
# ? Jan 11, 2017 11:20 |
|
FaustianQ posted:I wanted to return to this for a second and ask what the density difference is like? Would it be possible to do Zen on FDX 12nm? Because I know if GloFo and wild claims are a thing in the fab industry, so grain of salt on this I will have to talk to some people and get an actual measurement, because there is a lot of GloFo yapping on the internet about blah blah blah, 12nm FDX beats 14nm FinFET, but no SEM images of 12nm FDX silicon to analyze. Which makes sense, I suppose, they're still in risk production. Whatever it is, the way that they are angling their inherited IBM FD-SOI process is for IOT garbage, which leads me to believe that whatever the end product, it will not avail itself to Summit Ridge. Maybe ULV Unfortunately, I can only tell you at this moment that the difference in density is > 1. (Writing this down for later: Samsung's contacted gate pitch is 78nm and absolute minimum metal pitch is 67nm.) SwissArmyDruid fucked around with this message at 12:02 on Jan 11, 2017 |
# ? Jan 11, 2017 11:52 |
|
If you're going to do anything but Zen for ULV, might as well go Cat cores, which do it way better. Isn't Power 8 on 22nm FD-SOI? EDIT: For substance, Ryzen launch confirmed for February, possibly pre GDC. http://videocardz.com/65571/amd-ryzen-to-launch-at-gdc2017 Which means SKU leaks soon, so soon, and pricing EmpyreanFlux fucked around with this message at 13:51 on Jan 11, 2017 |
# ? Jan 11, 2017 12:24 |
|
Beat me to it, makes sense with the x300/370 motherboards coming out late February. They're saying that it's going to be a real launch but I wonder whether they'll release the 6/4 core parts at the same time.
|
# ? Jan 11, 2017 14:08 |
|
FaustianQ posted:If you're going to do anything but Zen for ULV, might as well go Cat cores, which do it way better. That's obviously not the case, as Stoney Ridge is already out, and it is a construction core with a 15W TDP at the low end. Furthermore, cat and construction cores are 28nm designs, I find it hard to believe that AMD will retain any 28nm designs parallel to Zen and its derivatives going from 2H2017 onwards, and I believe what I remember of roadmaps is what makes me think this. edit: I was right, I had forgotten that Raven Ridge had parts as low a TDP as 4W. SwissArmyDruid fucked around with this message at 14:37 on Jan 11, 2017 |
# ? Jan 11, 2017 14:34 |
|
MaxxBot posted:I liked how if the HSF wasn't properly seated and functional the chip would melt down after a couple seconds, almost happened to once but I cut the power to the machine just in time. I did this on a Athlon 64 something or another. Burned up instantly , smoke and pop
|
# ? Jan 11, 2017 22:09 |
|
MaxxBot posted:I liked how if the HSF wasn't properly seated and functional the chip would melt down after a couple seconds, almost happened to once but I cut the power to the machine just in time. That and when pencils stopped working, I used old copper paint for window defoggers and some masking tape to connect bridges pain in the rear end sometimes but worth it for that FSB bump
|
# ? Jan 11, 2017 22:35 |
|
BOOTY-ADE posted:That and when pencils stopped working, I used old copper paint for window defoggers and some masking tape to connect bridges pain in the rear end sometimes but worth it for that FSB bump http://i.imgur.com/78YKk.gif
|
# ? Jan 11, 2017 22:41 |
|
penus penus penus posted:I did this on a Athlon 64 something or another. Burned up instantly , smoke and pop Great video from back in the day demonstrating the utility of thermal throttling: https://www.youtube.com/watch?v=UoXRHexGIok
|
# ? Jan 12, 2017 00:09 |
|
Supposedly there's been a leak that hints at Ryzen launch before AMD's session at GDC, so by March 3rd. https://www.anandtech.com/show/11031/amd-set-to-launch-ryzen-before-march-3rd-meeting-q1-target More like threadshitter, am i right!? vvv mobby_6kl fucked around with this message at 01:49 on Jan 12, 2017 |
# ? Jan 12, 2017 00:32 |
|
please call SMT "threadripper" please call SMT "threadripper" please call SMT "threadripper"
|
# ? Jan 12, 2017 01:08 |
|
Have they released the official benchmarks for the ryzen chipset yet? I've heard the leaked ones apparently skew the performance of the intel chips to make them appear slower than they really are in the comparisons.
|
# ? Jan 12, 2017 05:40 |
|
AbrahamSlam posted:Have they released the official benchmarks for the ryzen chipset yet? I've heard the leaked ones apparently skew the performance of the intel chips to make them appear slower than they really are in the comparisons.
|
# ? Jan 12, 2017 07:55 |
|
AbrahamSlam posted:Have they released the official benchmarks for the ryzen chipset yet? I've heard the leaked ones apparently skew the performance of the intel chips to make them appear slower than they really are in the comparisons.
|
# ? Jan 12, 2017 08:50 |
|
|
# ? May 21, 2024 15:10 |
|
Which is fine, because the gains from using quad-channel memory over dual-channel have never been statistically, price-performance, or realistically significant, unless you're doing some crazy-rear end compute poo poo. This approach by AMD is totally justified. (Only having dual-channel, I mean. We already knew that AMD was going to skew some benchmarks somehow, we just didn't know *how* yet.) Naples, on the other hand, I think I saw a thing showing that Naples will have 8-channel memory? Now there's a chip that will probably need it, if the way AMD wants to handle manipulating large data sets by default pans out.
|
# ? Jan 12, 2017 10:07 |