|
Mr.Radar posted:The Scottish dude is done with AMD and won't release any more (public) speculation videos: Good.
|
#
?
Jul 23, 2019 18:26
|
|
|
|
| # ? Apr 29, 2024 16:50 |
|
|
Mr.Radar posted:The Scottish dude is done with AMD and won't release any more (public) speculation videos: haha yeah right, not until Zen 3 anyway.
|
|
#
?
Jul 23, 2019 18:38
|
|
|
Mr.Radar posted:The Scottish dude is done with AMD and won't release any more (public) speculation videos: And nothing of value was lost. edit: jfc, 41 minutes to say "gently caress that I'm out"?
|
|
#
?
Jul 23, 2019 18:50
|
|
|
It's been almost 2 weeks since my 3900X failed to ship and went on back-order. There's been no news at all since then. How likely is it that Zen3 will stick with socket AM4 and DDR4 RAM? I'm thinking of just buying a 3600 and purchasing a high-core count Zen3 chip next year. My i5-3570K is struggling a lot nowadays and I kinda want to get this new computer built. Has there been any word on when new stock should arrive?
|
|
#
?
Jul 23, 2019 19:15
|
|
|
Nalin posted:It's been almost 2 weeks since my 3900X failed to ship and went on back-order. There's been no news at all since then. How likely is it that Zen3 will stick with socket AM4 and DDR4 RAM? I'm thinking of just buying a 3600 and purchasing a high-core count Zen3 chip next year. My i5-3570K is struggling a lot nowadays and I kinda want to get this new computer built. AMD is promising socket compatibility with Zen3 on their roadmap. DDR5 will likely be a wet fart of performance when it first comes out, same as DDR4 on release, and I wouldn't worry about it too much.
|
|
#
?
Jul 23, 2019 19:21
|
|
|
Nalin posted:It's been almost 2 weeks since my 3900X failed to ship and went on back-order. There's been no news at all since then. How likely is it that Zen3 will stick with socket AM4 and DDR4 RAM? I'm thinking of just buying a 3600 and purchasing a high-core count Zen3 chip next year. My i5-3570K is struggling a lot nowadays and I kinda want to get this new computer built. https://www.nowinstock.net/computers/processors/amd/
|
|
#
?
Jul 23, 2019 19:40
|
|
|
SwissArmyDruid posted:edit: jfc, 41 minutes to say "gently caress that I'm out"? Hey it's ya boy ad revenue here to tell you that Youtube suuuuuuucks
|
|
#
?
Jul 23, 2019 20:13
|
|
|
Mr.Radar posted:The Scottish dude is done with AMD and won't release any more (public) speculation videos: The TLDR is: - blames AMD for not denouncing his leaks - blames fanboys for expecting him to do better - claimed that actually his leaks are not wrong, AMD just decided to release poo poo quality silicon this year and that 5ghz Zen 2 will be out next year totally vindicating him.
|
|
#
?
Jul 23, 2019 20:15
|
|
|
What a loving baby. I've never been tempted to watch one of his videos based on the rep he has. It's like the youtube equivalent of wccftech.
|
|
#
?
Jul 23, 2019 20:29
|
|
|
I feel like his channel began with reasonablish speculation and analysis and slowly drifted further and further from reality.
|
|
#
?
Jul 23, 2019 20:39
|
|
|
VostokProgram posted:I feel like his channel began with reasonablish speculation and analysis and slowly drifted further and further from reality. I feel kinda bad for the guy, he seems like he's got some personality problems (literally naming your channel Adored TV is just weird), the stress of it all is obviously getting to him, and he's obviously taking everything extremely personally. Watching someone who's obviously struggling a bit have a public meltdown isn't really a great feeling. Quitting reddit and laying low on twitter and producing some un-controversial content for a while would probably be a good idea for him. That said the guy just can't stop digging himself deeper. If he had just said "yeah, my leak was wrong, it happens", there would have been some snickering for a month and everyone would have forgotten about it. But I think this is Zen2 postmortem video number 4 or something just in the two months since the launch event - like jesus, he's been in damage control mode for like 6 months, and he still can't help digging himself deeper and deeper. He has a history of outright feuding with pretty much every other outlet (accusing PcPer of being on the take, going back and forth with HardwareUnboxed in videos and on twitter, probably a bunch of others I'm forgetting) and he was due for some schadenfreude back, now he's telling everyone that they shouldn't have taken his leaks so seriously, while in fact they were the ones taking the leaks with a grain of salt the whole time. And now he's pretty much going to take his ball and go home - you'll note this is Part 2, because Part 1 was a Patreon exclusive that I'm sure is another ego-stroker about how he was right this whole time. Guy needs to take a loving chill pill, drop the issue, and lay low for a while. As far as balance, it is a tough tightrope to walk. The AMD mediasphere loves the hype train poo poo, it's an easy route to subscriber numbers that some would consider.... "unnatural". There is a whole mediasphere of AMD-centric content producers that ride the relative enthusiasm of the AMD fanbase and general hatred of Intel/NVIDIA - "Moore's Law Is Dead", "not an apple fan", "good old gamer", "RedGamingTech", and most prominently Adored. That poo poo drives clicks like crazy and over time you become less of a neutral analyst and more of a partisan. After a while the general public gets tired of it, your clicks are mostly coming from the AMD mediasphere and if you piss them off then you've got nobody left, so it's full speed on the ayymd poo poo. This time I think he drank a little too much of the koolaid himself, and he's too egotistical to handle it reasonably. He kinda hoped to play it off as "close enough, things change", then "you can get there with overclocks", and it's just absolutely undeniable that the leak was full of poo poo and also absolutely undeniable that Adored can never ever bring himself to admit it. Paul MaudDib fucked around with this message at 22:01 on Jul 23, 2019 |
|
#
?
Jul 23, 2019 21:38
|
|
|
His claim that we got poo poo quality silicon makes no sense. Desktop wants the high leakage parts anyway, not server, so we are getting legitimately the best overclockers. Maybe the process will improve and they'll more often hit boost clocks across all cores (evidence points to best quality silicon hitting 4.55-4.65Ghz all core) but Zen2 is not loving hitting 5.0Ghz ever.
|
|
#
?
Jul 24, 2019 00:31
|
|
|
EmpyreanFlux posted:His claim that we got poo poo quality silicon makes no sense. Desktop wants the high leakage parts anyway, not server, so we are getting legitimately the best overclockers. Maybe the process will improve and they'll more often hit boost clocks across all cores (evidence points to best quality silicon hitting 4.55-4.65Ghz all core) but Zen2 is not loving hitting 5.0Ghz ever. It would be hilarious if months from now people are getting 4.7Ghz when all the people who bought at release can't. Fine WineTM
|
|
#
?
Jul 24, 2019 00:50
|
|
|
EmpyreanFlux posted:His claim that we got poo poo quality silicon makes no sense. Desktop wants the high leakage parts anyway, not server, so we are getting legitimately the best overclockers. Maybe the process will improve and they'll more often hit boost clocks across all cores (evidence points to best quality silicon hitting 4.55-4.65Ghz all core) but Zen2 is not loving hitting 5.0Ghz ever. Well, 5.0 was a single core. If good silicon is hitting 4.65 all-core I could see it hitting 5 GHz single core, as the process matures. Particularly if they do a Zen2+... Zen3 might be a 4-way SMT and that might imply server only and that might imply a Zen+ style stepping change / relatively vanilla shrink onto the 7+ node (although there is a massive amount of “mights” there obviously. And I do fully expect that people buying later in the cycle are going to get a much better shot at those 4.6 chips and so on. Same as Ryzen 1000, the bad silicon will start to close the gap on the good silicon. You’re not only paying to beta test AMD’s BIOS for them, you’re paying launch gouging prices to get the absolute shittiest chips AMD will ever release. Early adoption has its definite downsides. That said, as to his general thrust that AMD is saving the super great silicon for Epyc and he was right all along... no. His goalpost has been moving from 5 GHz stock, to 5 GHz OC, to “5 GHz maybe someday”. Maybe we do, maybe we don’t. But tbh 7nm seems kinda disappointing after all the hype... everyone was talking up how flat the voltage curve was gonna be and how great it would clock and it frankly barely seems to beat 12nm (although obviously at much greater efficiency). Paul MaudDib fucked around with this message at 02:02 on Jul 24, 2019 |
|
#
?
Jul 24, 2019 01:20
|
|
|
What was the link someone posted a bit ago about the zen 2 processors having one really good binned chiplet and one worse? I'd like to read more about that.
|
|
#
?
Jul 24, 2019 01:42
|
|
|
I wanna build something super small and as cheap as possible with maybe some RGB in the ram/cpu cooler, maybe fans. I was looking at the 3400G but it seems like it’s a poor value compared to a 3600+any GPU. One review I read said that a 2600+RX 570 nearly doubled the performance for slightly more than the APU. But it’s so small! It could fit into something like an Inwin Chopin. But it feels like a dead end.
|
|
#
?
Jul 24, 2019 02:07
|
|
|
Electric Bugaloo posted:I wanna build something super small and as cheap as possible with maybe some RGB in the ram/cpu cooler, maybe fans. I was looking at the 3400G but it seems like it’s a poor value compared to a 3600+any GPU. One review I read said that a 2600+RX 570 nearly doubled the performance for slightly more than the APU. If you've got a use for a lower power machine just do it. Maybe a HTPC or something? I wouldn't buy one for a desktop unless you were severely budget constrained or if it was a PC for a parent who just browses the web or whatever.
|
|
#
?
Jul 24, 2019 02:31
|
|
|
Anyone see that derbauer vid where he was measuring core voltages on Zen2? They idle at 0.2v. that's crazy https://www.youtube.com/watch?v=M5pHUHGZ7hU
|
|
#
?
Jul 24, 2019 02:35
|
|
|
Sidesaddle Cavalry posted:Version 2.0 for the basic integer stuff, and the general use extensions (i.e. multiply and divide, floating point) plus compressed instructions. Side note: base integer set for the 32-bit 16-register embedded style isn't there yet at v.1.9. Vector math and SIMD are the unfrozen mediocrity as karoshi posted. And that's just been the programming side of things. This is the exact kind of stuff that I would love a general CPU thread for (in addition to, not instead of the existing threads). I'm a hobbyist at best armed with some Crash Course videos and some wikichip articles. Thanks for sharing with a relative neophyte! I wish I had more to add 
|
|
#
?
Jul 24, 2019 02:53
|
|
|
Is there decent AMD laptops yet? Looking for middle range, as the budget laptop is always useless and full gamers/workstations are almost never of value.
|
|
#
?
Jul 24, 2019 03:07
|
|
|
ItBurns posted:It would be hilarious if months from now people are getting 4.7Ghz when all the people who bought at release can't. Fine WineTM 5Ghz 3950X 
|
|
#
?
Jul 24, 2019 03:08
|
|
|
Penpal posted:What was the link someone posted a bit ago about the zen 2 processors having one really good binned chiplet and one worse? I'd like to read more about that. I honestly think that was just people misunderstanding the boost behavior on the dual CPU die parts.
|
|
#
?
Jul 24, 2019 03:27
|
|
|
Burno posted:5Ghz 3950X They didn't say which May.
|
|
#
?
Jul 24, 2019 03:34
|
|
|
Cygni posted:I honestly think that was just people misunderstanding the boost behavior on the dual CPU die parts. People on Reddit were experimenting with per-CCX overclocking and most people were getting much better OCs on one CCD compared to the other. 
|
|
#
?
Jul 24, 2019 03:55
|
|
|
HERE'S A HOT TAKE FOR YA: If Zen 4 supports DDR5 and PCIe5, and it's in AM5 socket, AND THEY DON'T JUMP THE NUMBER TO ZEN 5, THEY HAVE FAILED US ALL If that all comes in Zen 5 naturally, then fine, sure, but Zen 4 would be the 5th generation 
|
|
#
?
Jul 24, 2019 05:02
|
|
|
NewFatMike posted:This is the exact kind of stuff that I would love a general CPU thread for (in addition to, not instead of the existing threads). I'm a hobbyist at best armed with some Crash Course videos and some wikichip articles. Believe me, I'm the neophyte compared to the great people I work with at Undisclosed Maker of Electronic Things, barely siphoning off what little knowledge I can make sense of in the short time I have until I go back to finishing school
|
|
#
?
Jul 24, 2019 05:23
|
|
|
The fact that the clock ceiling is so consistent reminds me of the Zen 1 launch where everybody thought it was the process but it turned out to be binning for top end SKUs like Threadripper. 4.1 GHz seemed unlikely at launch but got increasingly common towards the end of that generation. Getting another 200-300 MHz out of these CPUs would be a big deal for AMD now that the CPUs have almost reached single core performance parity.
|
|
#
?
Jul 24, 2019 06:32
|
|
|
Is there any chart that will show relative performance and RAM speed? Is DDR4 3000CL16 going a bit too slow?
|
|
#
?
Jul 24, 2019 07:04
|
|
|
ConanTheLibrarian posted:What a loving baby. I haven't really either, maybe a couple of minutes tops.. I have literally no idea why people refer to him, or he's become popular. Meaningless fluff in the form of overly long monologues. No thanks HalloKitty fucked around with this message at 07:43 on Jul 24, 2019 |
|
#
?
Jul 24, 2019 07:40
|
|
|
Prescription Combs posted:Anyone see that derbauer vid where he was measuring core voltages on Zen2? They idle at 0.2v. that's crazy I noticed that the cores go to super low frequencies too when idling compared to my previous CPU i7 4720HQ where they went to as low as 600 MHz iirc. At least according to task manager. Is the task manager reporting clocks incorrectly for Zen 2? Ryzen Master tells me those super low idle clocks but task manager shows 3+ GHz most of the time.
|
|
#
?
Jul 24, 2019 07:44
|
|
|
Is it possible that a game can get stuck on one of the lame cores and unable to live its best life? I'm just wondering why sometimes I load up iracing and the CPU frame time is really high. It usually sits around 7ms or whatever, but sometimes I load it up and it's at 15 and then goes back to normal if I quit and reload.
|
|
#
?
Jul 24, 2019 07:47
|
|
|
Alpha Mayo posted:Is there any chart that will show relative performance and RAM speed? Is DDR4 3000CL16 going a bit too slow? From reddit. https://docs.google.com/spreadsheets/d/1rqnHgPrG059fpOjArpc_nFSDVG1dsjw6y3o9ujf4hro/htmlview
|
|
#
?
Jul 24, 2019 08:05
|
|
|
Puddin posted:From reddit. What is this measuring? Alpha Mayo posted:Is there any chart that will show relative performance and RAM speed? Is DDR4 3000CL16 going a bit too slow? It differs from platform to platform and there aren't that many real-world comprehensive tests out there. For gaming on Zen 2, the difference between 3000 CL16 and something faster like 3600 CL17 or 3200 Cl12 seems to be about 0 to 5%ish, with most games closer to 0%. Note that anything over 3733MHz underperforms on Zen 2 because the Infinity Fabric slows down the bus to half the RAM speed, increasing latency. Basically it's worthwhile if it's relatively cheap to get something faster or lower latency. Note that Ryzen's memory controller only works with even-number CAS latencies, so CL15 and would perform identically to CL16 unless you manage to stably adjust the timing to CL14. E: This is with a 3900X and a 2080 Ti running at heavily CPU-bound frame rates. I suspect that the differences would mostly shrink or disappear with at GPU-bound frame rates, but I don't have any data to back that up. VVV That's a good point - TechPowerUp never does minimum/1% frame times or frame stability charts for anything and it's extraordinarily frustrating VVV Stickman fucked around with this message at 08:28 on Jul 24, 2019 |
|
#
?
Jul 24, 2019 08:19
|
|
|
Alpha Mayo posted:Is there any chart that will show relative performance and RAM speed? Is DDR4 3000CL16 going a bit too slow? https://www.techpowerup.com/review/amd-zen-2-memory-performance-scaling-benchmark/3.html https://lab501.ro/procesoare-chipseturi/amd-ryzen-3000-part-iv-ddr4-scaling-english-version Techpowerup doesn't have frame rate minimums which is the important part when benching memory and the Lab501 tests have a weird selection of memory speed/timing and seems to have some outliers so there's likely a fair bit of error margin. tldr: no it won't Cojawfee posted:Is it possible that a game can get stuck on one of the lame cores and unable to live its best life? I'm just wondering why sometimes I load up iracing and the CPU frame time is really high. It usually sits around 7ms or whatever, but sometimes I load it up and it's at 15 and then goes back to normal if I quit and reload. Yes, I've had that happen on a 2600 when a game decides to run on a SMT thread. Try manually setting core affinity to a physical core.
|
|
#
?
Jul 24, 2019 08:25
|
|
|
Paul MaudDib posted:Well, 5.0 was a single core. If good silicon is hitting 4.65 all-core I could see it hitting 5 GHz single core, as the process matures. Particularly if they do a Zen2+... Zen3 might be a 4-way SMT and that might imply server only and that might imply a Zen+ style stepping change / relatively vanilla shrink onto the 7+ node (although there is a massive amount of “mights” there obviously. No, I don't think it will. I think 4.7Ghz is where the design and process top out, just that we can agree this is early stage silicon and there are functional but lovely dies being used to meet demand, evidenced by the 3900X having one die that'll hit 4.55-4.65Ghz, while the other tops out at around 4.325-4.4Ghz. There are marginal gains to be made and I wouldn't be entirely surprised at a new stepping once the process is really mature, but we should expect 4.7Ghz at most. I also think Zen2+ won't be 7nmEUV, rather the 6nm process instead since it's way easier to port to, if Zen2+ exists. I doubt Zen3 is going to improve on this. Why waste time making a longer pipeline and a less dense process when the engineering effort with smaller processes makes it more expensive than merely going wider. 14nm++ will be the height of silicons clockspeed IMHO, I think chasing the Ghz dragon is a bad idea and Icelake is probably a good indication why. 18% gains* in IPC on top of Coffeelake indicate there is probably more headroom anyway in going wider than clockspeed will ever provide at a given voltage/power.
|
|
#
?
Jul 24, 2019 08:43
|
|
|
I've been a Ryzen late adopter twice, and maybe I've just gotten duds twice, or perhaps they bin the better silicon to the X chips only? 1. Ryzen 1700 that's stable at 3.8GHz/1.4v, can't go any higher no matter the voltage, even with a 360mm radiator. 2. Ryzen 2700 that's stable at 3.9GHz/1.35v, and also can't clock higher. Thankfully I don't really need the higher clocks, I'm GPU bound in the only game I play, and compiling software written in rust tends to use cores fairly well. I'm holding out for a 3950X and some kind of miracle so that I might be able to afford it.
|
|
#
?
Jul 24, 2019 09:02
|
|
|
The Ice Lake numbers only look good because it's two generations worth of incremental improvements rolled into one (rip Cannon Lake).
|
|
#
?
Jul 24, 2019 09:24
|
|
|
Alpha Mayo posted:Is there any chart that will show relative performance and RAM speed? Is DDR4 3000CL16 going a bit too slow? Getting faster ram won't hurt you, there have been benchmarks showing that some workload are memory bandwidth or latency constrained though its hard to say universally if it will matter for you. 3600 is the sweet spot for price/performance as the infinity fabric frequency is tied to the memory clock up to 3766, after that an IF frequency divider kicks in and you need to make a decision if your use case will benefit from more bandwidth or reduced latency because you can't have both after that point. Measly Twerp posted:I've been a Ryzen late adopter twice, and maybe I've just gotten duds twice, or perhaps they bin the better silicon to the X chips only? With Zen2, benchmarks seem to consistently show that a 3600 performs pretty much identically to a 3600x given sufficient cooling (bundled 3600 cooler can't keep up). It's hard to say if that's going to be universal for all X variants, but the 3600 seems to be the sweet spot and what I suspect most people will gravitate to. BangersInMyKnickers fucked around with this message at 14:14 on Jul 24, 2019 |
|
#
?
Jul 24, 2019 14:10
|
|
|
Mr.Radar posted:The Scottish dude is done with AMD and won't release any more (public) speculation videos: Deleted lol
|
|
#
?
Jul 25, 2019 00:30
|
|
|
|
| # ? Apr 29, 2024 16:50 |
|
|
Guy is turning into RedSydePhil.
|
|
#
?
Jul 25, 2019 00:37
|
|
