|
counterpoint: 7800X3D is cool
|
# ? Jul 23, 2023 14:02 |
|
|
# ? May 30, 2024 12:05 |
|
Kazinsal posted:Up here in the great white north, a 7700X is $370 CAD and a 7800X3D is $590 CAD. It is over two hundred dollars cheaper to buy the 7700X and maybe lose an error bar's worth of frametimes. But what if I only care about poorly coded vr games
|
# ? Jul 23, 2023 14:04 |
|
I don't think it's possible to separate "slow AM5 uptake" from "low demand for PC poo poo in general". Like, AM5 being a more expensive platform plays a part, but we've just gone through a massive boom cycle where everybody and their dog bought a new PC. None of those people need a new upgrade yet.
|
# ? Jul 23, 2023 14:44 |
|
And a lot of people with early Zen CPUs just bought a 5800x3D instead of doing a full upgrade.
|
# ? Jul 23, 2023 15:47 |
ijyt posted:counterpoint: 7800X3D is cool
|
|
# ? Jul 23, 2023 16:17 |
|
Xakura posted:But what if I care about vr games This person provably doesn’t exist.
|
# ? Jul 23, 2023 16:52 |
|
Wibla posted:And a lot of people with early Zen CPUs just bought a 5800x3D instead of doing a full upgrade. This is almost exactly what I did (5600X3D instead) and I really can't see myself upgrading for another 3-5 years.
|
# ? Jul 23, 2023 18:42 |
|
njsykora posted:This person provably doesn’t exist. There are dozens of us! Dozens! For real though, if it weren't for DCS VR, and to a lesser extent MSFS, I wouldn't bother upgrading.
|
# ? Jul 23, 2023 18:53 |
|
Xakura posted:There are dozens of us! Dozens! i upgraded from a 12900k to a 7800x3d just for msfs basically, that upgrade would have made zero sense otherwise Hopefully the board I bought will be fine down the road since it was like $500
|
# ? Jul 23, 2023 20:10 |
|
Wibla posted:And a lot of people with early Zen CPUs just bought a 5800x3D instead of doing a full upgrade. *raises hand* I don't know what the percentage would be, but I think there are a fair number of enthusiasts or whatever term you want to use who are very averse to jumping on the start of a new generation of hardware, especially memory standards. I remember very clearly how early DDR2 just sucked, and didn't jump onto DDR3 or DDR4 until they were quite far into their lifetimes. I'm sure I'll end up with a DDR5 machine at some point, but upgrading to top-of-the-line DDR4 made way more sense to me.
|
# ? Jul 23, 2023 20:40 |
|
Fortunately there is enough pressure on the APU side that AMD must improve the Zen5 IMC to hit bandwidth targets if nothing else.
|
# ? Jul 23, 2023 20:53 |
|
CaptainSarcastic posted:*raises hand* That is the whole reason I haven't gone for a 7800X3D: never buy the first generation platform of a new memory standard if you can hold out. (It also helps that usually by the time the second generation CPU/motherboards come out for the "new" memory standard it will have lost most if not all of its price premium.) My existing system (Intel 9900K + 3080 Ti) has been a pretty good run coming up on 5 years, but I'm already deep into obviously CPU limited land so this system is pretty much tapped out for upgrade potential which is why I passed on a 4090 founders edition when I had plenty of chances a few weeks back. Next upgrade I'm going to need a complete new build from the ground up.
|
# ? Jul 23, 2023 21:46 |
|
The new AGESA is a good sign of things to come, I think. The IMC frequency (uclk) is still limited (though it got a small boost), but the higher supported memory clocks is nice to see. Hopefully Zen 5 will be able to hit the memory speeds we're seeing with the new AGESA in 1:1 mode.
|
# ? Jul 23, 2023 21:47 |
|
The discussion of silicon lottery a couple pages ago got me to test my 7900X. I manually set my Peerless Assassin to 100% fan speed and ran Cinebench single and multicore. On single core, I never got above 5.5GHz according to HWInfo and on multicore, it hit 5.1GHz before some cores throttled back to 4.9GHz (temps maxed at 93C before Cinebench finished 1 loop). AMD says boost clock up to 5.6GHz. If I had known to test this when I bought the CPU from Microcenter back in March, maybe I could've exchanged it for a better unit.
|
# ? Jul 24, 2023 02:30 |
|
Advertised boost clocks are always for single-core loads only. Multicore loads will have much lower boost clocks. GamersNexus found an all-core boost clock of around 5100 MHz with an Arctic Freezer II 360 cooling 7900X. Hardware Unboxed found 5100 MHz on one CCD and 5050 MHz on the other, also with a 360mm AIO. It makes sense that it will be slightly slower with a Peerless Assassin. Maybe there's some silicon lottery at play here too, but I wouldn't worry too much about it. You're probably losing a few percent of performance from that at most, and there's no guarantee you would've gotten a better CPU after a return. edit: Also yeah, I usually see max boost speeds of 50 MHz below the advertised speeds. I guess AMD just rounds to the tenth. Dr. Video Games 0031 fucked around with this message at 02:51 on Jul 24, 2023 |
# ? Jul 24, 2023 02:47 |
|
Josh Lyman posted:The discussion of silicon lottery a couple pages ago got me to test my 7900X. I manually set my Peerless Assassin to 100% fan speed and ran Cinebench single and multicore. What is the peak multiplier in HWINFO? On zen3 X3Ds AMD counts 44.5*100MHz as 4.5GHz so 55.5*100 is probably 5.6 as far as AMD’s marketing is concerned.
|
# ? Jul 24, 2023 02:49 |
|
Josh Lyman posted:The discussion of silicon lottery a couple pages ago got me to test my 7900X. I manually set my Peerless Assassin to 100% fan speed and ran Cinebench single and multicore. Are you actually logging data with a high sample rate, or just looking at the HWinfo window? You also need to be testing in a benchmarking setup -- a clean OS so nothing is happening in the background, checked that a single-thread workload is going to the preferred core, all options set to maximum performance, etc. I don't really think you've really established whether your CPU is a winner or loser in silicon lottery. Probably it's about average -- most CPUs are average and there are relatively few "losers" -- and you just don't have the test conditions right. And I would say, unless you want to really dig into a ton of boring nerd poo poo, don't bother. "Up to" = if you're within 100mhz, call it good. Dr. Video Games 0031 posted:You're probably losing a few percent of performance from that at most At most it's gonna be 1/55th and in any IRL use it's gonna be far less -- even games will have enough load on secondary cores to drop that max single-thread boost.
|
# ? Jul 24, 2023 03:33 |
|
Klyith posted:Are you actually logging data with a high sample rate, or just looking at the HWinfo window? You also need to be testing in a benchmarking setup -- a clean OS so nothing is happening in the background, checked that a single-thread workload is going to the preferred core, all options set to maximum performance, etc. My 7900X, even if just average, is definitely overkill for my uses which are basic computering and Diablo 4 which my 4070 is perfect for--I'm locked at 160fps at 1440p ultra with DLSS quality. I was originally planning to get a 13600K or 7700X but the 7900X Microcenter bundle was the same price for CPU/mobo/RAM so I took the free upgrade. I have some interest in trying Flight Simulator since I have a triple monitor setup and maybe Kerbal Space Program at some point, so I'm not exactly doing anything where an extra 100mhz on either single or multicore will make any difference. It's always nice to seeing higher benchmark scores of course, but since I'm not interested in liquid cooling my CPU, I suppose that's a bit of a red herring. Josh Lyman fucked around with this message at 05:24 on Jul 24, 2023 |
# ? Jul 24, 2023 05:22 |
|
sometimes i wonder if i should sell my 3950x and go for a cheaper and much faster in games 5800x3d now that i don't use all that cpu power regularly, but a couple times a year something really dumb happens that makes me glad i still have all the goddamn threads usually it's work related poo poo, but i'm playing stationeers with friends since a patch landed last week, and it's really nice to be able to run both the dedicated server and the game on my computer and still have the game perform normally i was worried because internet is littered with people complaining how vpses aren't beefy enough so as soon as you turn on ventilation in a big base everyone starts desyncing unless you go for a monstrously expensive one with many cores/ram, and how it's impossible to run the dedicated server and the game on the same machine and have a playable framerate, but i'm here just cruising along at 60-80% cpu usage and 90fps lmao thanks amd
|
# ? Jul 24, 2023 10:38 |
|
https://twitter.com/VideoCardz/status/1683402971936112640?s=20 X3D chips for laptops - potentially good for desktop gaming replacements
|
# ? Jul 24, 2023 12:34 |
|
ok i'll bite: how the hell does the mobile cpu naming scheme work? i see they're at least using a 7000 system that might actually correspond to the desktop generation this time around
|
# ? Jul 24, 2023 13:05 |
|
kliras posted:ok i'll bite: how the hell does the mobile cpu naming scheme work? i see they're at least using a 7000 system that might actually correspond to the desktop generation this time around https://www.msi.com/blog/understand-how-amd-name-their-mobile-cpu 7945HX3D 7 means it's released in 2023 9 means it's a Ryzen 9 (in terms of the product stack, as in competes with an Intel Core i9) 4 means it's using Zen 4 architecture 5 means it's the higher-tier model within this specific product segment (with the other possible option for this digit being '0', for the lower-tier model) HX means it's for the 55W+ TDP target (HS is ~35W, U is 15-28W, C is 15-28W but for Chromebooks, e is 9W / fanless variant of a U) 3D means it uses 3D-stacked cache
|
# ? Jul 24, 2023 13:16 |
|
alright, that's not too terrible aside from the form factor/tdp nomenclature actually
|
# ? Jul 24, 2023 13:27 |
|
Do they sell the AMD product name decoder ring at Microcenter?
|
# ? Jul 24, 2023 13:35 |
|
kliras posted:ok i'll bite: how the hell does the mobile cpu naming scheme work? i see they're at least using a 7000 system that might actually correspond to the desktop generation this time around The suffix 45 cpus are chiplet based designs like the desktop (and should be basically equivalent to desktop Ryzen 7000 in lower envelope). 40 are monolithic apus. E: and everything below 4 in the 3rd digit is rebranded from previous gens, because gently caress you AMD.
|
# ? Jul 24, 2023 16:02 |
|
https://twitter.com/taviso/status/1683484820817444866 Vulnerability in Zen 2 CPUs: https://web.archive.org/web/20230724143835/https://lock.cmpxchg8b.com/zenbleed.html
|
# ? Jul 24, 2023 16:16 |
|
That’s a fun one, up there with the Straight-Line Speculation thing from ARM.
|
# ? Jul 24, 2023 17:51 |
|
kliras posted:ok i'll bite: how the hell does the mobile cpu naming scheme work? i see they're at least using a 7000 system that might actually correspond to the desktop generation this time around They literally gave out decoder rings when 7000 series was announced for this purpose. Note that this does not preclude them doing wack-rear end poo poo like releasing Zen 3/3+ products in 2023 and them hoping that you see the "7" and forget that it doesn't work like Intel and fail to realize that it's not Zen 4: https://www.amd.com/en/products/apu/amd-ryzen-5-7535u SwissArmyDruid fucked around with this message at 20:50 on Jul 24, 2023 |
# ? Jul 24, 2023 20:44 |
|
Malloc Voidstar posted:https://twitter.com/taviso/status/1683484820817444866 lmao there's no fix for consumer CPUs, AMD only has pushed out patched microcode for Epyc. Tom's Hardware is saying AMD wasn't ready to announce and "didn't seem prepared". Desktop and mobile won't be until later this year. I think taviso hosed up and published early because AMD had patched the stuff Google runs. Just lmao. The bad news, I've seen at least a couple people saying this could be exploited by javascript. Dunno if that's true, others say not or it would be difficult. The good news, 30kb/s per core isn't exactly a fast way to dig through memory. Faster than some of the other Spectre-type exploits though. But way slower than Meltdown, which could read your memory in pretty much real time. But if you needed yet another reason to run Ublock with pervasive script blocking, here it is.
|
# ? Jul 24, 2023 23:13 |
|
project zero refers to the number of friends the security researchers there will have in the industry by the time they quit their jobs
|
# ? Jul 24, 2023 23:36 |
|
Klyith posted:lmao there's no fix for consumer CPUs, AMD only has pushed out patched microcode for Epyc. Tom's Hardware is saying AMD wasn't ready to announce and "didn't seem prepared". Desktop and mobile won't be until later this year.
|
# ? Jul 25, 2023 00:35 |
|
Malloc Voidstar posted:AMD broke the embargo for the vuln according to Tavis https://github.com/google/security-research/tree/master/pocs/cpus/zenbleed#timeline Yeah, I get that from the security researcher perspective the moment any patch is live is when they go public (because other people can look at the diff and steal credit). But there may be a difference of opinion on what counts as breaking embargo. AMD pushing out a patch for Epyc while staying radio silent probably indicates they thought they had more time. I dunno, it seems pretty haphazard compared to the Spectre/Meltdown disclosure where everyone waited until the appointed day and there was a four-way deal between Intel, the researchers, MS, and the Linux kernel team.
|
# ? Jul 25, 2023 01:01 |
|
AMD laptop chips have been pretty good - I recently picked up a 7940HS laptop with a 4090 for gaming on vacations it can go 9-10 hours on battery for desktop stuff (browsing etc.) and perform competitively with the latest intel chips for gaming. if anything, i'm kinda shocked at how poorly intel's laptop offerings are in terms of power consumption
|
# ? Jul 25, 2023 01:41 |
|
Speaking of 58003dxes. They go for about 300 cad here. I can survive a few months for a sale but how much do CPUs really go on sale for. If it's less than 20% I'd probably just pick it up.
|
# ? Jul 25, 2023 02:02 |
|
$300 CAD is already a discounted price. It might drop even further, but probably not by too much more than it already has. edit: Using PC Part Picker, I see it for $350 on Memory Express, which is the lowest recorded price for it by a good margin: If you can find it for $300, that's a very good price. Dr. Video Games 0031 fucked around with this message at 03:39 on Jul 25, 2023 |
# ? Jul 25, 2023 03:36 |
|
Dr. Video Games 0031 posted:$300 CAD is already a discounted price. It might drop even further, but probably not by too much more than it already has. Seconding this, buy it for 300 if you can find it. I was stupidly happy to get it for 350-400 a few weeks ago from Canada computers. Considering it was like 500+ at the start of the year, I didn't want to waste an opportunity for an upgrade. Eventually, these CPU's won't be findable. If you want it, get it now.
|
# ? Jul 25, 2023 16:43 |
|
I got the fancy 5800 mixed up with the 3d cache one. alas. Computer parts are always coming down in price though folx. I'm surprised by this scarcity nonsense. I think we're in a glut of hardware and the computer centres will blink first. A glut always follows scarcity.
|
# ? Jul 25, 2023 20:19 |
|
Toalpaz posted:I got the fancy 5800 mixed up with the 3d cache one. alas. Computer parts are always coming down in price though folx. I'm surprised by this scarcity nonsense. I think we're in a glut of hardware and the computer centres will blink first. A glut always follows scarcity. Eh, more likely that this will run out, and just not be available. Prior to that, the price should spike a little.
|
# ? Jul 25, 2023 20:53 |
|
The only reason people mention them running out is that it's your last chance for a good upgrade while not having to also upgrade your motherboard/ram
|
# ? Jul 25, 2023 21:13 |
|
|
# ? May 30, 2024 12:05 |
|
Klyith posted:lmao there's no fix for consumer CPUs, AMD only has pushed out patched microcode for Epyc. Tom's Hardware is saying AMD wasn't ready to announce and "didn't seem prepared". Desktop and mobile won't be until later this year. As far as consumer hardware: a vulnerability is a vulnerability but if you've got malware running on your bare metal you're already popped. The reason it's such a big deal for servers is it leaks across VM boundaries so you have to worry about someone else renting the same underlying server and spying on you. e: what on earth, is there a wordfilter on javascript? Harik fucked around with this message at 18:28 on Jul 26, 2023 |
# ? Jul 26, 2023 18:22 |