|
does it make sense to buy into the am4 platform at this point if you're just after pure price/performance and don't care too much about an 'upgrade path'? i'm seeing some pretty good ebay deals occasionally on the 5900x/5950x. i'm using an old as gently caress lenovo prebuilt with an i7 4770 at the moment with ddr3 so i'd be on the hook for the motherboard, cpu and RAM no matter what semi related: what's the minimum amount one should consider spending on a motherboard if they want it to be any good? i did own an AM4 PC with a 3700X for a bit (probably shouldn't have sold it lmao) and both boards I used with it, B450 and X570, were replete with weird issues. issues with the networking, issues waking from and going to sleep, issues with the USB, absolutely god-awful unusable noise on the analogue audio ports etc. i'd been using intel since my athlon 64 like 10+ years earlier and never had so many issues like that e: actually i didn't look at the 13600k before writing this - seems to beat the 5900x and trade blows with the 5950x in productivity, you can buy an old motherboard and DDR4 RAM and use them with it, the power consumption isn't that bad, and the platform is probably more robust (purely from my limited anecdotal experience). when did intel become the value option???? Generic Monk fucked around with this message at 19:08 on Oct 23, 2022 |
#
?
Oct 23, 2022 18:44
|
|
|
|
| # ? May 30, 2024 04:14 |
|
|
Generic Monk posted:does it make sense to buy into the am4 platform at this point if you're just after pure price/performance and don't care too much about an 'upgrade path'? i'm seeing some pretty good ebay deals occasionally on the 5900x/5950x. i'm using an old as gently caress lenovo prebuilt with an i7 4770 at the moment with ddr3 so i'd be on the hook for the motherboard, cpu and RAM no matter what When everyone talks about “buying into a platform” they’re saying they’re going to be doing upgrades for years to come or that they already have components that “fit”. If you want to purchase a completed, “sealed” box that has a certain performance at a certain price then you’re not buying “into a platform”, you’re buying a specific product. edit: ^^ outside of Intel having products like the 12100 that amd doesn’t have, that happened this week. hobbesmaster fucked around with this message at 19:11 on Oct 23, 2022 |
|
#
?
Oct 23, 2022 19:08
|
|
|
hobbesmaster posted:When everyone talks about “buying into a platform” they’re saying they’re going to be doing upgrades for years to come or that they already have components that “fit”. If you want to purchase a completed, “sealed” box that has a certain performance at a certain price then you’re not buying “into a platform”, you’re buying a specific product. i just liked how 'buying into a platform' sounded as a phrase when the platform in question has clearly reached its apogee. functionally though i would have to buy a whole new platform so i think the terminology works, even if the platform chosen is getting the last rites performed on it Generic Monk fucked around with this message at 19:29 on Oct 23, 2022 |
|
#
?
Oct 23, 2022 19:26
|
|
|
Generic Monk posted:when did intel become the value option???? When AMD went upmarket with Zen 3, and then Intel caught up on single threaded performance. The server market right now is that you buy Intel if you want value, and you pay up for AMD if you want the best performance. It's looking like desktop is there too now, with Intel turning the power up to 11 to stay competitive.
|
|
#
?
Oct 23, 2022 19:27
|
|
|
Generic Monk posted:semi related: what's the minimum amount one should consider spending on a motherboard if they want it to be any good? i did own an AM4 PC with a 3700X for a bit (probably shouldn't have sold it lmao) and both boards I used with it, B450 and X570, were replete with weird issues. issues with the networking, issues waking from and going to sleep, issues with the USB, absolutely god-awful unusable noise on the analogue audio ports etc. i'd been using intel since my athlon 64 like 10+ years earlier and never had so many issues like that The USB issues were persistent problems for years, but supposedly have now been solved. Most AM4 boards have cheap realtek ethernet adapters, which IMO are "fine" for wired ethernet. Yeah it's realtek, but wired ethernet is so simple and standard that even realtek doesn't gently caress it up. Wireless you want to pay attention to what network chip you're getting. OTOH a few X570s used an intel 2.5G ethernet chip that turned out to be buggy. Terrible analogue audio is a hazard of integrated mobo sound everywhere, and while some mobos are better than others it also depends on other components, the case, etc. If you care about sound you should be using digital out to an external DAC, or get a USB DAC. They're not expensive. But as you've realized, an AM4 build is not the value option right now unless you're getting a really good deal (ie buying from a friend not ebay) on used components. Generic Monk posted:when did intel become the value option???? When they spent 3 years stuck on an old process, had to compete for the first time in over a decade, and tried to coast on reputation until they lost their dominance in the enthusiast market.
|
|
#
?
Oct 23, 2022 19:40
|
|
|
Twerk from Home posted:It's looking like desktop is there too now, with Intel turning the power up to 11 to stay competitive. the 13900K is faster in 95W "eco" scenarios than the 7950X though? That's been the consistent finding, other than HUB who messed up their testing. like, basically AMD just stops scaling at a certain power level, and yeah if you limit it to that point but allow Intel to run free and keep scaling, Intel pulls more power. You can also cap Intel to the same level though, and at that point it's faster. Somehow this has been framed as a disadvantage for Intel, like, I guess it's better somehow that AMD stops scaling at a certain power level??? Maybe Intel should just have burned a power limit in at the same level as AMD stops scaling? the exception is super heavy AVX productivity loads like cinebench and blender, but fundamentally that's because the 13900K is a 7900X competitor (and priced as such, really) being stretched to compete with a 7950X. It does it, but, it needs more power to do it. But in gaming or other partial-load workloads, the 13900K actually is equally or sometimes edges out the 7950X in terms of efficiency. and remember this is Intel at a node disadvantage... which everyone was super keen on pointing out when comparing against M1. On a 5nm-tier node Intel should do quite a bit better, AMD just is throwing money at the problem and jumping onto 5nm as soon as it's ready.
|
|
#
?
Oct 23, 2022 20:04
|
|
|
also, while the XTU bug HUB hit may not have been their fault... they also didn't bother to talk to any other reviewers and ask if anyone else was seeing the same things. Or cross-check their CB R20/R23 results against any other reviewer at all. The power curve was turning obviously concave so it should have been relatively obvious that something was going wrong. while the bug may not have been their fault, I really wonder if the tables had been turned if they would have been a little more diligent about investigating a "potentially erroneous" result that benefited Intel, vs just rolling with it and going "that's our data, deal with it". Yeah they retracted that segment but the damage is done, the erroneous chart is already doing the rounds and people have already accepted that "zen4 is significantly more efficient than raptor lake", even when raptor lake actually edges it out in reality in eco-power scenarios. it's curious how with HUB these "random chance errors" always seem to benefit AMD and maybe it just comes down to them being more diligent about investigating the ones that don't favor their expected outcome while they just accept the ones that do. Same for their "editorial judgement calls" like RT and DLSS being downweighted or outright ignored as scoring factors, and so on. Like I legitimately cannot recall a time when HUB ever made a mistake in favor of Intel or NVIDIA. Ever. And in 3DCenter's cross-reviewer meta-reviews, their number consistently lean pro-AMD by about 5-10% over the all-reviewer (30+ reviewers) geomean. Maybe it's a Fox News Polling moment where they have a lean but it's an accurate number with a consistent lean, but, here we are with another retraction on inaccurate numbers leaning in the exact same pro-AMD direction their errors always do. Paul MaudDib fucked around with this message at 20:20 on Oct 23, 2022 |
|
#
?
Oct 23, 2022 20:10
|
|
|
Klyith posted:The USB issues were persistent problems for years, but supposedly have now been solved. The X570 board that I used had an intel adapter which I don't remember having too many problems with - trying to remember whether the sleep issues were resolved when I switched boards. I didn't cheap out on that board, it was like £150, but it still wasn't what I would call stable. Couldn't get more than a week of uptime without it bluescreening or loving up in some way. It also ran the CPU weirdly hot at idle compared to the previous one. For some reason I had decded to do a micro atx build that time around, and I found out in short order that mATX just seems to be a dumping ground for low quality parts. I think I had the only x570 motherboard that got made lmao upgrading is a want more than a need for me rn anyway so I'll have a look early next year once the market has shaken out a bit more
|
|
#
?
Oct 23, 2022 20:11
|
|
|
Alchenar posted:We're at the stage where everyone who bought AM5 is starting to complain about the regular crashes, so one more reason to wait a little and let the bios versions mature a bit. What's this now? I got a 7700x with an Aorus B board, friend has the same, though the got the X version of the board. Not a single crash at all.
|
|
#
?
Oct 23, 2022 20:23
|
|
|
Paul MaudDib posted:the 13900K is faster in 95W "eco" scenarios than the 7950X though? That's been the consistent finding, other than HUB who messed up their testing. I don't think this is true at all, do you have specific benchmarks in mind?
|
|
#
?
Oct 23, 2022 20:35
|
|
|
Arzachel posted:I don't think this is true at all, do you have specific benchmarks in mind? Der8auer did the benchmarks: https://youtu.be/H4Bm0Wr6OEQ He ran 90W i9 and 90W 7950x and in productivity they are pretty similar, with Ryzen usually being slightly faster. In gaming i9 is just slightly slower than uncapped i9, and way above Ryzen offerings. E: a crazy idea: Intel 13th gen capped at 90W and 4090 with 90% performance at ~300W should run stable on 650W PSU of decent quality.. alex314 fucked around with this message at 21:22 on Oct 23, 2022 |
|
#
?
Oct 23, 2022 21:09
|
|
|
Koskun posted:What's this now? I got a 7700x with an Aorus B board, friend has the same, though the got the X version of the board. Not a single crash at all. Apparently Asus boards are having issues. And by 'people complaining' I mean a number of people on Reddit who are obviously only a small proportion of users, but still highlighting there's a bit of risk.
|
|
#
?
Oct 23, 2022 21:14
|
|
|
I have a gigabyte board and haven't crashed a single time. I have updated the bios from stock though
|
|
#
?
Oct 23, 2022 21:17
|
|
|
alex314 posted:Der8auer did the benchmarks: Yeah but that's the opposite of what Paul's saying. All the power limiter tests I've seen have the 7950x ahead in multi-core until you go past the 150W mark where it stops scaling.
|
|
#
?
Oct 23, 2022 21:25
|
|
|
HUB straight up said AMD needs to cut prices because raptor lake skus compete a tier above their AMD counterparts in productivity and are as good if not sometimes better in gaming. So sure, they are completely shilling for AMD at all times lol. Nice to see some things never change here. On a more relevant note my 7700X and MSI B650M Mortar should be here at the end of the week and will pair it with the MSI 4090 Gaming Trio. Hopefully the board is as stable as everyone’s gigabytes boards. First AMD cpu in quite a while. B-Mac fucked around with this message at 23:27 on Oct 23, 2022 |
|
#
?
Oct 23, 2022 23:25
|
|
|
there have been better reasons to call them shills in the past but it's a bit silly here yeah
lih fucked around with this message at 00:17 on Oct 24, 2022 |
|
#
?
Oct 24, 2022 00:10
|
|
|
its not even a close contest here when 13600K is barely more expensive than a 7600X (lmao) while the former is even cheaper as a platform with budget mobos both are moot points to me since im not even remotely CPU bottlenecked by my icy cold 5600 driving a 3070 Palladium fucked around with this message at 02:23 on Oct 24, 2022 |
|
#
?
Oct 24, 2022 02:13
|
|
|
ugh quote isnt edit
|
|
#
?
Oct 24, 2022 02:22
|
|
|
Palladium posted:ugh quote isnt edit not with that attitude!
|
|
#
?
Oct 24, 2022 02:52
|
|
|
Paul MaudDib posted:On a 5nm-tier node Intel should do quite a bit better, AMD just is throwing money at the problem and jumping onto 5nm as soon as it's ready. Intel is staying "competitive" at a node disadvantage because it's a 50% wider architecture than Zen 4 for the Raptor Cove cores (and 25% wider for Gracemont, though it has A LOT more execution ports), and so you could also look at it as AMD is keeping parity+ via the node advantage for its less wide architecture, since IPC between Raptor Lake and Zen 4 are fairly close. When AMD gets Zen 5 to match architectural width with Intel, then we'll actually get to see how much the node advantage really gets them. Canned Sunshine fucked around with this message at 06:55 on Oct 24, 2022 |
|
#
?
Oct 24, 2022 06:50
|
|
|
lmao I can buy a Gigabyte Z690 AX D4 on amazon which is still cheaper than the cheapest B650 mobo, and the 13600K is head and shoulders above the 7600X really planning to move a lot of Zen 4 product this year AMD
|
|
#
?
Oct 24, 2022 13:27
|
|
|
would DDR4 compatibility have made a difference? fake edit: thinking about this a little, having an AM5 board that supported DDR4 would have undermined the implied promise of a long-lived AM5 board, because you can't keep it once you're finally ready to move into DDR5
|
|
#
?
Oct 24, 2022 13:34
|
|
|
gradenko_2000 posted:would DDR4 compatibility have made a difference? they just needed to send a shipment of crack to asrock HQ so their engineers could cook up some boards with both 2xDDR4 and 2xDDR5 slots it's been done before on earlier DDR transitions
|
|
#
?
Oct 24, 2022 13:51
|
|
|
ddr4 compatibility & reasonable motherboard prices would have been a big improvement to value but even with ram/board prices equal amd would still be losing, just by a much smaller margin. that would have made 'at least you can upgrade later on' way more appealing in comparison though
lih fucked around with this message at 14:04 on Oct 24, 2022 |
|
#
?
Oct 24, 2022 13:57
|
|
|
I think pricing and the board situation is a bigger issue than not having DDR4 support, Raptor Lake performs better and is cheaper to boot. Wouldn't be surprised if AMD does another marketing push around the X3D release with price cuts to the base lineup.
|
|
#
?
Oct 24, 2022 15:04
|
|
|
Anyone know of outstanding bugs in Ryzen 1? https://twitter.com/muizelaar/status/1584550803725234181
|
|
#
?
Oct 24, 2022 16:11
|
|
|
Subjunctive posted:Anyone know of outstanding bugs in Ryzen 1? Yeah, tons of them: https://support.amd.com/TechDocs/55449_Fam_17h_M_00h-0Fh_Rev_Guide.pdf Ryzen 1 is a very new architecture, a bigger single step change than anything else in more than a decade. Tons of these errata will completely ruin some specific application. Issue 1021 means that it doesn't obey the Java or C++ memory model, sometimes synchronized reads will get old data.
|
|
#
?
Oct 24, 2022 16:32
|
|
|
SourKraut posted:When AMD gets Zen 5 to match architectural width with Intel, then we'll actually get to see how much the node advantage really gets them. The next gen for both companies is a big swing. And Zen 5 won’t be up against an Intel 7 part anymore. Rumors this summer were that AMD wouldn’t get TSMC 3 wafers until Q4 2024. That would mean Zen 5 is either a TSMC 5 part that will compete against Meteor Lake’s Intel 4 based compute tile, or won’t launch for a long while where it will compete against Arrow Lake’s Intel 20A or TSMC 3 compute tile. Both companies potentially using the same fab on desktop will sure be interesting for comparisons though.
|
|
#
?
Oct 24, 2022 16:38
|
|
|
Ryzens TRNG getting stuck in a state where it just outputs 1s forever was a fun one I don't think that was even fixed, everyone just learned the lesson to use the OS CSPRNG instead of RDRAND repiv fucked around with this message at 16:45 on Oct 24, 2022 |
|
#
?
Oct 24, 2022 16:38
|
|
|
Now watch, we'll come back around to this years later when everyone has been using a bugged OS CSPRNG when we could have been using a much faster RDRAND instead that somehow accelerates the workload in question by 90%, a la the Linux KVM stubs fix from https://www.phoronix.com/news/AMD-NPT-GPU-Pass-Through SwissArmyDruid fucked around with this message at 15:42 on Nov 3, 2022 |
|
#
?
Oct 24, 2022 21:01
|
|
|
I asked the other day about trying to fix my friend's many bent pin 5800x3d after it ripped out with the cooler. I bent the pins back and installed it in my system, but it doesn't POST. I didn't see any broken pins while fixing it, so I'm not sure what might be wrong. Any ideas? How about pretend like it was never bent and fresh out the box. What else causes no POST? Hmm I'm using an MSI B450-A Pro with a 3600 that is using that smaller sized BIOS they had to release so that the 3000 series CPUs would work with it (I believe those had the MAX suffix, mine does not). I wonder if I needed to update the board or if it's even compatible at all? edit: shows it's compatible https://us.msi.com/Motherboard/B450-A-PRO/support#cpu but I need the latest beta BIOS. drat would that cause no POST if I hadn't updated? Silly me. KingKapalone fucked around with this message at 21:25 on Oct 24, 2022 |
|
#
?
Oct 24, 2022 21:15
|
|
|
KingKapalone posted:Hmm I'm using an MSI B450-A Pro with a 3600 that is using that smaller sized BIOS they had to release so that the 3000 series CPUs would work with it (I believe those had the MAX suffix, mine does not). I wonder if I needed to update the board or if it's even compatible at all? Yes, you need bios update 7B86vAH to run a 5000 series Ryzen in that mobo.
|
|
#
?
Oct 24, 2022 21:21
|
|
|
SwissArmyDruid posted:Now watch, we'll come back around to this years later when everyone has been using a bugged OS CSPRNG when we could have been using a much faster RDRAND instead that somehow accelerates the workload in question by 90%, a la the Linux driver fix from earlier this/last year.
|
|
#
?
Oct 24, 2022 21:41
|
|
|
Cygni posted:The next gen for both companies is a big swing. And Zen 5 won’t be up against an Intel 7 part anymore. Rumors this summer were that AMD wouldn’t get TSMC 3 wafers until Q4 2024. That would mean Zen 5 is either a TSMC 5 part that will compete against Meteor Lake’s Intel 4 based compute tile, or won’t launch for a long while where it will compete against Arrow Lake’s Intel 20A or TSMC 3 compute tile. If AMD didn't get TSMC 3, then wouldn't it probably move to N4P or an improvement on that? Apple would likely be on TSMC 3, which would open up N4P for others, though it's obviously an improvement on TSMC 5. And Intel 4 is basically still a 7nm process, if I recall? Though now Intel is more heavily integrating EUV.
|
|
#
?
Oct 25, 2022 01:20
|
|
|
SourKraut posted:If AMD didn't get TSMC 3, then wouldn't it probably move to N4P or an improvement on that? Apple would likely be on TSMC 3, which would open up N4P for others, though it's obviously an improvement on TSMC 5. nanometers are a meaningless, non-standardized measurement. what specifically are you measuring, how are you measuring it, what does it mean for the efficiency of the chip? you can just as easily say that TSMC N3, N4, N5 aren't 3nm, 4nm, 5nm either. I think Intel has shown that Intel 7 is pretty dang competitive still despite being "10nm," with more overall perf in raptor lake over zen 4, and there are some comparisons where they get surprisingly close on perf/watt.
|
|
#
?
Oct 25, 2022 01:52
|
|
|
SourKraut posted:If AMD didn't get TSMC 3, then wouldn't it probably move to N4P or an improvement on that? Apple would likely be on TSMC 3, which would open up N4P for others, though it's obviously an improvement on TSMC 5. The numbers are extremely arbitrary at this point. "Smallest feature size" is a stupid & pointless thing to measure; average transistor size / density is useful but they keep that a secret. Intel 7 is using FinFET & mostly EUV, which means it should be pretty similar the TSCM 5nm stuff. TSMC N3 will use GAAFET instead of FinFET. That's gonna be a big deal, and Intel won't have the equivalent until "Intel 20A" in 2024 or later (the recent rumors are that Intel will have very limited production on that in 2024).
|
|
#
?
Oct 25, 2022 01:53
|
|
|
BlankSystemDaemon posted:Software CSPRNGs can be changed, hardware ones embedded in CPUs can't be changed - which leads into the benefits of independent auditing of both the CSPRNG and its OS implementation. I kinda want a hardware RNG that's just a microgram of Americium from a smoke detector plated onto a CMOS sensor and hooked up to an FPGA. Depending on a few things, you could get possibly tens of megabytes per second of random noise, perfect for impressing your nerd friends when playing some absurdist variation of Shadowrun, where you need to roll 6^6^6D6 to determine if Satan is summoned to eat the world or not.
|
|
#
?
Oct 25, 2022 01:59
|
|
|
Dr. Video Games 0031 posted:I think Intel has shown that Intel 7 is pretty dang competitive still despite being "10nm," with more overall perf in raptor lake over zen 4, and there are some comparisons where they get surprisingly close on perf/watt. This might be a dumb question, but isn't this in large part because Raptor Lake's P-cores are 6-wide while the E-cores are 5-wide, compared to Zen 4's 4-wide architecture? So between that, Gracemont's execution port differences, and just cranking up the juice significantly at the high end of the performance curve to squeeze out a bit more, Intel still seems to be just about hitting the wall with what "Intel 7" can do without a significant architectural change.
|
|
#
?
Oct 25, 2022 03:34
|
|
|
Klyith posted:Yes, you need bios update 7B86vAH to run a 5000 series Ryzen in that mobo. Thank you. I got it working. Free 5800x3d for me! I'm coming from a 3600. What should I know as a part of the upgrade? I think this one will run hotter? I have a Scythe Mugen 5 rev.b. I'll have to run some benchmarks, but should that be fine? My BIOS settings got cleared and I don't remember what I had my fan curve at, but it's definitely louder sitting in Windows. Any recommended curve? I remember at Zen 2 launch that the power settings were weird and I ended up trying a few. I'm not sure if they switched when I installed the CPU, but it's currently on High Performance. I installed 1usmus custom Ryzen plans but might not have been using them. What do people use there? I probably need to update the chipset drivers?
|
|
#
?
Oct 25, 2022 04:04
|
|
|
|
| # ? May 30, 2024 04:14 |
|
|
KingKapalone posted:Thank you. I got it working. Free 5800x3d for me! I went from a 3600X to a 5800X3D (which required a BIOS update) and the only thing I think I've had to do is re-do my fan curves. I did end up upgrading my case for better thermal management, but that was also because I went from a 2070 Super to a 3080 12GB, so had more heat to manage in general.
|
|
#
?
Oct 25, 2022 04:34
|
|
