|
ConanTheLibrarian posted:Why bother watching/reading content from fake rumour mills like Adored and wccftech when there are reputable, at least somewhat reliable leakers out there? Seems like a waste of time to me. People like to have some nerd fantasy material to tug themselves over, and it's understandable to be honest; it's exciting to think you're one of the first who knows about some new piece of tech, even if it turns out to be a load of old nonsense
|
#
?
Aug 6, 2020 16:59
|
|
|
|
| # ? Jun 5, 2024 00:23 |
|
|
Adored's navi predictions were also very funny, and he disappeared for a few months after getting everything wrong on that one
|
|
#
?
Aug 6, 2020 22:39
|
|
|
gradenko_2000 posted:https://www.youtube.com/watch?v=11NnOCQq-CQ You forgot to mention that it's an R3 4300G w/Vega 6, not the full fat Vega 8, and the Vega 6 is locked to 1700mhz. That's basically matching the GT 1030 in per ALU performance, which is...odd? That's slightly behind Navi in per ALU performance, so maybe that's just confirmation that Renoir is using a hybrid Vega-Navi (Vega ALU blocks, Navi everything else). https://mobile.twitter.com/GPUsAreMagic/status/1274024392469741569 It'd explain why AMD would have been in no hurry to replace it, Vega ALUs are much smaller than Navi ALUs, the ROPs and revamped Rasterizer are half the performance themselves. I mean, IIRC they already did mixmatch poo poo like this with Vega M GH/L which was Polaris with Vega IP blocks.
|
|
#
?
Aug 7, 2020 12:16
|
|
|
FuturePastNow posted:Adored's navi predictions were also very funny, and he disappeared for a few months after getting everything wrong on that one Was he right about anything? I don't recall him making any good calls. I completely stopped paying attention to him after he started claiming the 3600 would be 5GHz.
|
|
#
?
Aug 7, 2020 21:53
|
|
|
Even a stopped clock is right twice per day. I wouldn't give someone who makes enough guesses any credit when one of them turns out to be right, it's just coincidence.
|
|
#
?
Aug 7, 2020 22:21
|
|
|
EmpyreanFlux posted:You forgot to mention that it's an R3 4300G w/Vega 6, not the full fat Vega 8, and the Vega 6 is locked to 1700mhz. That's basically matching the GT 1030 in per ALU performance, which is...odd? That's slightly behind Navi in per ALU performance, so maybe that's just confirmation that Renoir is using a hybrid Vega-Navi (Vega ALU blocks, Navi everything else). An ARM-like composable blocks approach would also explain things like the RDNA < PS5 < RDNA2 situation. Edit: I knew their semicustom unit operated like this, from articles and interviews, but I didn't realize the blocks were so low level. The things I read made it sound more like "We can put you together a chip with x Bobcat cores and y Fiji cores", etc. mdxi fucked around with this message at 00:26 on Aug 8, 2020 |
|
#
?
Aug 7, 2020 23:51
|
|
|
What's wrong with Intel, and how to fix it: Former principal engineer unloadsquote:“Intel is very lucky AMD cannot get the volume, to be able to compete,” Piednoel. “If they were getting volume, the price difference would definitely cost Intel market share a lot more than what they are losing right now.”
|
|
#
?
Aug 9, 2020 07:26
|
|
|
Piednoel is a narcissistic blowhard though. Sure buddy, if a boy genius like you were in charge of Intel then none of the factors that constrain Intel’s range of options would exist. Like, if you were Intel’s leadership then what do you do? Their fab division is apparently broken, cleaning it out will probably put them even further behind schedule in the short term and there’s no guarantee of success in the long term (culture problems tend to be very persistent). All their new architectures are designed on nodes they can’t manufacture. Once it became obvious that 10nm and 7nm are clusterfucks they started backporting their designs to 14nm (likely around the 2017 timeframe) but that is a multi year proposition still (coming to fruition with Rocket Lake late this year/early next year). In the future they are going to focus on portability across nodes but the Intel Way was tying your architecture very tightly to the node (this was seen as a strategic advantage) and generally Intel nodes are quite different from the rest of the industry, so that doesn’t help you in the short term. Skylake obviously needs security work too, but again, all the new architectures are on nodes that can’t be manufactured and backporting those designs takes years. As frustrating as it is to enter year 5 of Skylake refreshes, Intel really is playing their hand as best as can be expected on the CPU side. America’s Smartest Boy over there should tell Intel what architecture, what node, how long he thinks it should take to get into production, and how much of it he thinks they can get produced. At the end of the day those are the questions that matter, and Intel has found a reasonably optimum solution given their predicament. It would be better if the predicament didn’t exist but the reality is it does. The actual crazy behavior is stuff like acquiring technologies and then killing them after a disappointing first gen, and the (previous) unwillingness to consider TSMC for certain product lines to bridge the gap on their manufacturing. Architecturally they are pretty much playing their cards as best they can. And yes, they could cut prices but why, when they’re selling everything they can make and their competitor is capacity constrained? Paul MaudDib fucked around with this message at 08:35 on Aug 9, 2020 |
|
#
?
Aug 9, 2020 08:12
|
|
|
Piednoel is a blowhard and he himself admits he has no recent information. Intel did indeed cut prices by issuing a new product line that's designed to go 1:1 with epyc. They will not cut msrp for existing products for a number of reasons. The fish rots from the head. Krazanich was a process engineer! Engineers as company managers isn't a panacea. Swan has been willing to do some needed things so he isn't bad. If he stops the dumbass acquisitions and entries into garbage markets it'll be good enough while he can rebuild intel manufacturing. Some financial discipline never hurts.
|
|
#
?
Aug 9, 2020 08:23
|
|
|
According to Tom's Hardware AMD is looking into big.LITTLE hybrid CPUs.quote:The method described in AMD's patent appears to allow the processor to independently sort out which type of thread should run on each cluster based on the instructions supported by the cores within. The threads could also shift between the cores based on utilization. For example, if the large core is underutilized, the processor would shift the thread to the small core (provided it supports the instructions). If the small core is over-utilized, the thread would shift to the larger core (again, provided it supports the instructions).
|
|
#
?
Aug 9, 2020 20:56
|
|
|
seems like an elaborate "gently caress you" to microsoft in the form of a patent
|
|
#
?
Aug 9, 2020 23:28
|
|
|
Oh, I see, making it so that the OS doesn't need to handle that scheduling, I see how it'd be a F-U to Microsoft.
|
|
#
?
Aug 10, 2020 00:49
|
|
|
Wouldn't Microsoft be glad to not have to worry about that complexity and push it down to the CPU vendor?
|
|
#
?
Aug 10, 2020 01:25
|
|
|
Seems like it'd be a big F-U to Mr. Linus Torvalds, who wrote Linux.
|
|
#
?
Aug 10, 2020 01:28
|
|
|
Either it's going to need some opt-in from the OS or the SMM firmware is going to be swapping LAPIC IDs on the fly so the OS doesn't take an enormous poo poo every time it tries to send an IPI to a core assuming that there's a specific thread on that core and suddenly oops nope there's something else there. Though swapping LAPIC IDs between different cores all the time has some other nasty impacts if what you thought was core 4 is actually now core 12 and core 12 is a .little core and whoops
|
|
#
?
Aug 10, 2020 02:02
|
|
|
Doesn't this mean AMD could make EPYC chips with half CPU dies and half TPU dies? Wouldn't this enable AMD to get better performance than AVX512 by using dedicated TPUs instead ikea? Like this seems perfect for Zen5 in 2023, it's a bit late for Zen4 though.
|
|
#
?
Aug 10, 2020 16:16
|
|
|
Subjunctive posted:Wouldn't Microsoft be glad to not have to worry about that complexity and push it down to the CPU vendor? Mofabio posted:Seems like it'd be a big F-U to Mr. Linus Torvalds, who wrote Linux.
|
|
#
?
Aug 10, 2020 16:22
|
|
|
Giant F U to god is what it is! I have no idea why this would be considered an insult to anyone
|
|
#
?
Aug 10, 2020 16:57
|
|
|
It's an F U to everyone who has been making heterogeneous core CPUs for years and didn't have the idea to patent this first. And I think nobody else bothered to patent it because doing that job will always be better done by the OS, which knows more about what the task needs than the CPU does. I doubt this will be a major thing on desktop CPUs.
|
|
#
?
Aug 10, 2020 17:17
|
|
|
Question for AMD goons: I got a 3700X to upgrade my 1700X and have a Thermaltake Contac Silent 12 cooler. Does anybody have any tips on how to re-attach the cooler after putting the new chip in and applying the thermal compound? It's got one of those stupid thumb clamp hooks and as far as I can tell it is absolutely goddamn impossible to install that thing without hopelessly smearing the thermal compound everywhere while trying to get the leverage to snap the frame onto the base. Not sure if this is even the right thread for this question but I dunno, hoping someone else with experience with this kind of thing may have their own way. Tried looking at guides online but it seems like everyone I've ever come across has zero problems getting this thing over their CPU, meanwhile I'm over here rolling the pipes in Arctic Silver 5 like a pig rolling around in mud
|
|
#
?
Aug 10, 2020 18:40
|
|
|
Tiny Tubesteak Tom posted:Question for AMD goons: I got a 3700X to upgrade my 1700X and have a Thermaltake Contac Silent 12 cooler. Does anybody have any tips on how to re-attach the cooler after putting the new chip in and applying the thermal compound? It's got one of those stupid thumb clamp hooks and as far as I can tell it is absolutely goddamn impossible to install that thing without hopelessly smearing the thermal compound everywhere while trying to get the leverage to snap the frame onto the base. Not sure if this is even the right thread for this question but I dunno, hoping someone else with experience with this kind of thing may have their own way. a: Don't worry about smearing the compound, the end result is to have a super-thin layer between the CPU & cooler. What you don't smear will mostly squeeze out anyways. b: Be much more worried about spring-clip retention things, especially if you need to use a tool to get it installed. Those things have high accident potential. I might recommend taking the mobo out of the case completely so you can work on it without obstruction.
|
|
#
?
Aug 10, 2020 18:59
|
|
|
Do you mean tools required to install? Or having to use tools to install it? I haven't used a tool in the past and afaik it isn't required to use one to install/uninstall, just the metal thumb clip. In fact, I can take the cooler off with the thumb clip no problem, it just seems to be getting it back on that I have all the problems
|
|
#
?
Aug 10, 2020 19:05
|
|
|
Tiny Tubesteak Tom posted:Do you mean tools required to install? Or having to use tools to install it? I haven't used a tool in the past and afaik it isn't required to use one to install/uninstall, just the metal thumb clip. In fact, I can take the cooler off with the thumb clip no problem, it just seems to be getting it back on that I have all the problems Many of those clip things have a slot where you can put a flat-head screwdriver and use it for extra leverage. Works great for forcing the clip down, but in doing so you're pushing down hard with a semi-sharp tool. One slip means you put a gash in your mobo. Yours may not have that though. You might also check that the CPU is fully inserted into the socket, if it's not that could make the clip much harder to get attached. Spring clips in general just suck.
|
|
#
?
Aug 10, 2020 19:21
|
|
|
Okay, so good news. The 3700X came and the stars must have aligned because that was the quickest and easiest CPU replacement I think I've ever done. I poo poo a couple bricks taking the 1700X out because when I went to pull the heatsink off, it pulled the loving processor clean out of the socket. I don't know if I was supposed to undo the socket latch before taking the heatsink off but I have no idea how I was supposed to reach the latch under the fins if that was the case. Luckily, nothing bad happened, the pins on the 1700X were totally fine and the 3700X went in no problem. BAD NEWS is, I STILL cannot POST this loving machine with four sticks of RAM installed. When I built this rig a few years ago, I heard up and down how Ryzen 1st gen had major problems running 4 sticks of RAM. I've been running 24GB in the interim because I need the higher amount and am willing to sacrifice the dual-channel speeds to get it. 2nd gen Ryzen came and went, and I still heard about the RAM compatibility issues, so I held off. So why, in the year of our lord 2020, can I still not POST this thing with four sticks of RAM and a freshly installed 3700X? I went through a decent amount of trouble to ensure that I had four sticks that were 1) on the QVL, 2) the same exact speed, and 3) the same exact CAS latency. One set is an EVGA 16G-D4-2400-MR CL15 set that I got from MicroCenter, the other is a HyperX Fury 2400 CL15 set. I can run the machine with one stick from each set, I can run it with literally any combination of three sticks of RAM, but gee willikers Batman, try to slot in that fourth stick and you don't even get a POST code on the digit display, you just go gently caress yourself. So now I'm pissed off, and I ragebought a 2x16GB Crucial Ballistix 3200MHZ set, and when it gets here on Wednesday, I'm going to fire this lovely rear end RAM into the sun with the force of a thousand angry gods. I guess this is the universe's way of telling me to pony up for a better set of RAM and stop trying to make the kit I have work. I have absolutely no idea how I can put three sticks in any configuration and boot, but the fourth stick shuts it all down. Flabbergasted. Blabberflasted. Just gonna run 16GB until Wednesday because I'm sweaty and already cut myself on the heatsink once, no need to go pushing my luck with a fresh set coming in. There isn't a Matisse QVL for the X470 Taichi but I've heard that there are no problems at all at 3200MHz and up, so fingers crossed? No idea. e: I want to point out that this issue isn't specific to the X470 Taichi. I had an MSI B350 Tomahawk before this and I couldn't POST with four sticks of RAM in that thing, either. barnold fucked around with this message at 22:48 on Aug 10, 2020 |
|
#
?
Aug 10, 2020 22:39
|
|
|
My 3700x doesn't post with the default voltages on 4 sticks, I have to lower my RAM volts by 0.1v to get it to post, after that I can take them up to DDR3666CL18. RAM timings are always sold as for 2 sticks, if you want to run 4 sticks you are going to be playing with your timings to get them to run but there is plenty of headroom. Also check your BIOS since that is what is mostly in control of your RAM timings so make sure it is the newest version and see what other people with your board are doing to get high overclocks on their RAM. 4 sticks of factory overclocked RAM are not going to run out of box. My big issue was the voltage was just too high for 4 sticks and was causing massive heat problems.
|
|
#
?
Aug 10, 2020 23:03
|
|
|
I had issues with POST and 4 sticks too. Resolved itself after I reset the bios and booted with all 4 installed. It took a long time to train on the memory but it's been fine since. Worth a shot. Weird thing is that I was able up restore my saved OCed bios settings soon after the successful POST with no problems. Same settings that were active when it wouldn't POST. Doesn't make any sense.
|
|
#
?
Aug 10, 2020 23:06
|
|
|
Tiny Tubesteak Tom posted:I poo poo a couple bricks taking the 1700X out because when I went to pull the heatsink off, it pulled the loving processor clean out of the socket. I don't know if I was supposed to undo the socket latch before taking the heatsink off but I have no idea how I was supposed to reach the latch under the fins if that was the case. For future reference, HSF removal always goes easier if the system is warm. Then, undo the fastener(s) and twist the cooler (as in: a clockwise/anticlockwise movement) a bit as you pull straight up. There's nothing magical about any of this; you're just thermally and then mechanically lessening the adhesion between the cooler and the IHS. After that, yes, you're supposed to flip up the lever on the ZIF socket. I generally clean the thermal interface material off the CPU before unseating it, as it's a lot safer to do then (you can't get TIM on the pins if the pins are still in the socket). But, as you've discovered, it's hardly the end of the world if the CPU comes out with the HSF/cooler.
|
|
#
?
Aug 10, 2020 23:25
|
|
|
Cezanne is showing up on SiSoft https://twitter.com/_rogame/status/1292844896374718470 It's basically Renoir, but Zen3. Also there is apparently better utilization of bandwidth, either some adjustment in the iGPU or the CPU needs to hit system memory less often. The iGPU is scores 6% better but has 22% less bandwidth. I have no idea if that would allow it to scale better at higher clock speeds but at least it's performance is harder to sabotage through OEM incompetence/malice? https://twitter.com/_rogame/status/1292850762708668416
|
|
#
?
Aug 11, 2020 03:58
|
|
|
 That clearance height limit. 
|
|
#
?
Aug 12, 2020 15:50
|
|
|
GRECOROMANGRABASS posted:
|
|
#
?
Aug 12, 2020 16:42
|
|
|
lmao
|
|
#
?
Aug 13, 2020 01:26
|
|
|
Have you heard the word of the 11 foot 8?
|
|
#
?
Aug 13, 2020 03:03
|
|
|
I don't understand why they ignore the blinking 'you are too tall for this' lights that's literally just in front of it.
|
|
#
?
Aug 13, 2020 08:53
|
|
|
I spent the next couple of hours watching bridge collision videos, thanks for that
|
|
#
?
Aug 13, 2020 09:34
|
|
|
Zedsdeadbaby posted:I don't understand why they ignore the blinking 'you are too tall for this' lights that's literally just in front of it. Most of the people who crash are in rented trucks, so they're also focused on traffic and managing the truck. They have no experience with needing to look for height warnings.
|
|
#
?
Aug 13, 2020 13:56
|
|
|
Klyith posted:Most of the people who crash are in rented trucks, so they're also focused on traffic and managing the truck. They have no experience with needing to look for height warnings. Also probably didn't actually know for certain the height of the truck. You'd think it'd be on a giant sticker or something on the dash but not in the moving truck I rented once!
|
|
#
?
Aug 13, 2020 14:16
|
|
|
Yeah. It happens all the time, but especially late August/early September every year on Storrow Drive in Boston, some dumbass college students will have rented a truck to move in and drive down Storrow because it's the easiest way to some places. There are repeated signs and swinging obstacles that will hit your vehicle if it's too tall, but nothing can stop people. Some pretty spectacular crashes can happen too, because a lot of the bridges are at angles horizontally and vertically and/or are curved.hobbesmaster posted:Also probably didn't actually know for certain the height of the truck. I don't think I've ever seen a rental truck that wasn't marked so you can see the height in the mirror.
|
|
#
?
Aug 13, 2020 14:24
|
|
|
The Gamer's Nexus "Ryzen isn't smoother" video sure does have the AMD subreddit in a tizzy. https://www.youtube.com/watch?v=1kK6CBJdmug Main reason I'm waiting for Zen3 is I'm betting the new CCX design shrinks or surpasses that gap, and 120+ 1440p gaming with the lowest input lag possible is about my only reason to upgrade.
|
|
#
?
Aug 13, 2020 22:07
|
|
|
Im pretty happy with my new 12 core Ryzen compared to my older 8700k. They did cost a similar amount, over 2 years I think. Bottom line, the power usage and heat production are half or less than the 8700k. Enough for me. I miss Quicksync.
|
|
#
?
Aug 13, 2020 22:40
|
|
|
|
| # ? Jun 5, 2024 00:23 |
|
|
I'm still on my "I refuse to buy Ryzen until they have a unified L3 cache" for reasons that are bullshit. Logically, I know it doesn't mean crap in day-to-day use, I've been running a 1600 (non-AF) for the past..... three months now? And I can't tell where it would actually make a difference, now that we're several years past the teething issues being ironed out, it's purely brainworm poo poo. Probably just using it as an excuse to wait for the launch of a new socket that will bring DDR5 with it.
|
|
#
?
Aug 13, 2020 23:02
|
|