|
spasticColon posted:Unfortunately, my Sandy Bridge board is a P67 series. I might sell the RAM though. Everything else is my system is so old I don't think I could resell it for much. Does used DDR3 RAM (Two different 2x4GB kits) fetch a good price? Ram doesn't go for much (maybe $15-25), but working P67 motherboards still sell for $50-80 and 2500ks for $40-50. The 1070 is still fine for 1080p/60, though you might want to turn down a setting or two in demanding/poorly optimized games like AC:Origins.
|
![]() |
|
![]()
|
# ? May 18, 2024 18:46 |
|
Stickman posted:Ram doesn't go for much (maybe $15-25), but working P67 motherboards still sell for $50-80 and 2500ks for $40-50. Nobody should bother with used Intel CPUs and mobos anymore. Good DDR4 is so cheap now, and the 6 core gen1 or gen2 Ryzen on mobo combo deals wipes the price/perf floor out of them an actual warranty and without the questionable reliability of an 8 year old used mobo.
|
![]() |
|
Palladium posted:Nobody should bother with used Intel CPUs and mobos anymore. Good DDR4 is so cheap now, and the 6 core gen1 or gen2 Ryzen on mobo combo deals wipes the price/perf floor out of them an actual warranty and without the questionable reliability of an 8 year old used mobo. Nobody should, but there's still plenty of people willing to throw a $70 bandaid on an old system rather than $250 to upgrade. And to be fair, even though the upgrade is a much better value it is still quite a bit more expensive. VVVV Good point - the math definitely works out differently in different places, too! VVV Stickman fucked around with this message at 02:21 on Jun 13, 2019 |
![]() |
|
Stickman posted:Nobody should, but there's still plenty of people willing to throw a $70 bandaid on an old system rather than $250 to upgrade. And to be fair, even though the upgrade is a much better value it is still quite a bit more expensive. Anecdotally, in Australia it cost me $578 AUD to upgrade to a 2600/B450 Tomohawk/16GB LPX DDR4 3400 on Monday. That was the cheapest option, even after shopping all of the potential options including those with 3+ week shipping and sketchy RMA stories. So $150 for a 4.5GHZ 2500k (and a $30 Evo 212+) with 16GB DDR3 1600 is a bit of a no-brainer. I actually managed to get a spare gigabyte z77 working after all - the fucker still lives and still runs at 4.5 - and I'm considering putting the lot on FB marketplace with a GTX770 I have lying around for a few hundred bux. I could probably cover the cost of a 3600x if I put it in a case, but I only have a 450W PSU spare. Is it a smart purchase? Probably not, but for half the cost of a new barebones system, someone is probably gonna pay it.
|
![]() |
|
Yeah, an overclocked 2600k is as fast as a stock R7 1700 or R5 2600 in a lot of games. Of course, you can OC the Ryzens too. But there is a reason Sandy Bridge havers never stop smuggin'. https://www.gamersnexus.net/hwreviews/3412-intel-i7-2600k-revisit-2018-benchmarks-vs-9900k-ryzen-more Cygni fucked around with this message at 02:33 on Jun 13, 2019 |
![]() |
|
Yeah, we're at the point where 4/4 is starting to cause problems in some games, but 4/8 is generally still good enough for gaming. It's definitely next on the chopping block, though.
|
![]() |
|
In hindsight I wish I would have gone with a 2600K but it was over $100 more and I was reading conflicting info about how hyper-threading/SMT could cause a performance hit in some games. But back in 2011 most games were still poorly optimized and single-threaded. Now more games are multi-threaded but still poorly optimized. Honestly the only upcoming AAA games I'm excited for on PC are Doom Eternal and Cyberpunk 2077 which is why I'm thinking about stepping down my upgrade to a 3600 or 3600X and a B450 motherboard.
|
![]() |
|
Yeah, I put that $100 into more RAM and top of the line aircooler. Hell, I didn't even get an SSD not because I couldn't afford one but because the only stuff out there reasonably priced was strictly used as OS drives back then.
|
![]() |
|
mdxi posted:I noticed that ASRock released the updated BIOS for my motherboard, so I did some messing around with underclocking/undervolting. I've always run everything stock, so I wanted to learn how this all worked before the 3X00s dropped. Thanks for reminding me to look: I've been waiting for an update for my B450 Pro4 since BIOS version 3.20 screwed up passthrough by not allowing me to reboot Windows VM cleanly. The page mentions installing "AMD all in 1 with VGA driver ver:18.50.16.01_WHQL" before flashing. I'm using Linux to virtualise, with a Ryzen 1700 and an Nvidia GPU. Sorry for the dumb question, but there's no need for me to install an AMD driver before flashing BIOS, is there? I didn't do that last time. ![]()
|
![]() |
|
mdxi posted:I noticed that ASRock released the updated BIOS for my motherboard, so I did some messing around with underclocking/undervolting. I've always run everything stock, so I wanted to learn how this all worked before the 3X00s dropped. Thanks for reminding me to look: I've been waiting for an update for my B450 Pro4 since BIOS version 3.20 screwed up passthrough by not allowing me to reboot Windows VM cleanly. The page mentions installing "AMD all in 1 with VGA driver ver:18.50.16.01_WHQL" before flashing. I'm using Linux to virtualise, with a Ryzen 1700 and an Nvidia GPU. Sorry for the dumb question, but there's no need for me to install an AMD driver before flashing BIOS, is there? I didn't do that last time. ![]()
|
![]() |
|
That's only relevant if you have a Ryzen APU (that is, onboard graphics). The 1700 doesn't have an onboard GPU, so you should be fine. See http://asrock.pc.cdn.bitgravity.com/TSD/Display%20recovery%20SOP.pdf
|
![]() |
|
You don't need to install that.
|
![]() |
|
TheFluff posted:That's only relevant if you have a Ryzen APU (that is, onboard graphics). The 1700 doesn't have an onboard GPU, so you should be fine. See http://asrock.pc.cdn.bitgravity.com/TSD/Display%20recovery%20SOP.pdf Stanley Pain posted:You don't need to install that. Yeah. It was a dumb question, looking back to this morning ![]() Thanks for confirmation though. To continue the spirit of the thread, the 1700/B450 combo is a total bargain for using as a casual virtualisation host. I'll apply that update tonight and play with new clock settings. Thanks mdxi
|
![]() |
|
Stickman posted:Yeah, we're at the point where 4/4 is starting to cause problems in some games, but 4/8 is generally still good enough for gaming. It's definitely next on the chopping block, though. The only titles that really really have problems with 4C4T are Ubisoft garbage with tons of DRM (FC5/AC:O) and BF:V (which just has optimization problems in general, it's lost 30%+ performance since launch). These titles are literally just optimization problems, the Ubi stuff runs fine on a 1.6 GHz laptop processor in the consoles and BF:V lost its head engine guy a month before launch and haven't been able to replace him. While DX12 titles are starting to favor Ryzen more, if they do things right the 4/4 processors are not terrible in these games either. DX12 is a general gain in CPU performance, it's not some curse that affects low-core-count processors. 4/4 should still push 90fps+ average / 60fps lows in those titles, that's still a perfectly fine gaming experience. The 7000 series was definitely a bit of a rip but the 7600K is pushing 2.5 years old at this point and Skylake is a year older than that. You expect i5s to have a bit faster upgrade cycle, ~3 years of solid gaming performance is about what you get. And you can still push another year or two out of it in most games if you really need to. Hardware Unboxed and others seem to have a little bit of an unreasonable expectation of what you get from the value-oriented SKUs, 3-4 years is decent and the i7s (or 5820K) were always the recommendation if you intended to go for a longer upgrade cycle. But really the 3600 is going to be the default recommendation for the price bracket for a while, unless Intel drops something really compelling. No sense paying $150 for 4/4 when you could pay $200 and get 6/12. Intel really needs to readjust their whole lineup here, who knows if they're going to, but if they don't do an 8700K-style realignment they're going to get pushed out of the DIY market very shortly. Paul MaudDib fucked around with this message at 17:40 on Jun 13, 2019 |
![]() |
|
Paul MaudDib posted:The only titles that really really have problems with 4C4T are Ubisoft garbage with tons of DRM (FC5/AC:O) and BF:V (which just has optimization problems in general, it's lost 30%+ performance since launch). These titles are literally just optimization problems, the Ubi stuff runs fine on a 1.6 GHz laptop processor in the consoles and BF:V lost its head engine guy a month before launch and haven't been able to replace him. People aren't paying $150 for 4/4 though, the i5-9400F is 6/6 and a $150 part, right?
|
![]() |
|
Twerk from Home posted:People aren't paying $150 for 4/4 though, the i5-9400F is 6/6 and a $150 part, right? Locked 6/6, but yeah. I was thinking more of the 8350K/9350KF.
|
![]() |
|
Paul MaudDib posted:BF:V ... it's lost 30%+ performance since launch How?
|
![]() |
|
Combat Pretzel posted:Wait, what? ![]()
|
![]() |
|
The Frostbite engine's source code is basically the Tome of Vile Darkness for programmers.
|
![]() |
|
Combat Pretzel posted:Wait, what? ![]() DICE's engine guy (who literally wrote the API for Mantle/Vulkan) and their project manager got tired of EA's poo poo and bailed to start Embark Studios a month before BF:V launched, and I'm guessing everyone who was anyone at DICE went with them. The game launched completely broken and they had an absolutely skeleton crew fixing it while a second studio worked at building a whole different battle royale mode that played fairly differently onto it for the next 6 months. So the technical state of the codebase has just gotten worse and worse. They didn't release basically any new content for that entire time and the player base has just been absolutely hemorrhaging. DICE/EA refuse to release player numbers but they've started deleting game modes to try and consolidate the remaining player base. Up until they did the first major map release a month ago the game was basically collapsing in on itself, it may have finally struck bedrock but up until then I'd say it was actually a solid contender for Anthem-style removal from benchmarking suites. I'm back playing BF1 and apart from the smaller modes not starting anymore it's great. There's actually still community servers with active admins to ban hackers. Hacking is a MASSIVE problem in BF:V since there are no rental servers and DICE basically doesn't do fairplay bans. Paul MaudDib fucked around with this message at 18:11 on Jun 13, 2019 |
![]() |
|
It’s not just Ubi/BF5, though. A lot of other newer AAA titles like Hitman 2 and Shadow of the Tomb Raider are starting to have major frame pacing issues on 4/4 processors. Even if they’re not to the point of completely bottlenecking the GPU, hitching and stutter is pretty terrible for the experience.
|
![]() |
|
Stickman posted:It’s not just Ubi/BF5, though. A lot of other newer AAA titles like Hitman 2 and Shadow of the Tomb Raider are starting to have major frame pacing issues on 4/4 processors. Even if they’re not to the point of completely bottlenecking the GPU, hitching and stutter is pretty terrible for the experience. That's Ryzen though. The APUs have much less cache than the Zeppelin-based processors and get hit hard by Zen's weak memory controller. Effectively this manifests as much lower IPC on APUs compared to the main Ryzen lineup. The 2400G underperforms the 1500X very significantly (6% slower despite 100 MHz higher clocks), and the 2200G underperforms the 1300X very significantly (9% slower despite 100 MHz higher clocks), despite higher clocks and single-CCX they just can't make up for the loss of cache. And that's on the same uarch generation. This is why I've always advised against people buying Raven Ridge with the intention of upgrading to a dGPU down the road. You are still trapped with that lovely CPU performance unless you upgrade that too. It's better to just suck it up and jump to a 460 or something (they're about $50 on ebay). Meanwhile the 7600K/8350K and 7700K still lay waste to the 1500X and 1300X, let alone Raven Ridge. Especially if overclocked (they're sitting on an easy 20% OC headroom over the stock all-core boost). This is specifically a problem with Ryzen, and specifically a problem with Raven Ridge. The Zeppelin hexas/octos have higher IPC than the Zeppelin quads, and the Zeppelin quads have higher IPC than the APUs, because of the cache. And notionally, there is still a place for 5+ GHz 4/4 if it were priced correctly. Just not at $200. It would have been fine as an i3 this generation (if it hadn't been priced like an i5), it would be fine as a Pentium next generation at say $100 (and AMD will probably release one as an update to the 1300X/1500X eventually as well). Not everyone needs the 3900X to play their COD games. ![]() ![]() (compare 2200G/1300X and 2400G/1500X, and vs the 7600K/7700K) ![]() ![]() (compare 1500X against 8350K OC. Just pretend the 8350K is a 7600K because TPU doesn't have OC results for the 7600/7700K in their 2400G charts and they are basically the same processor) Paul MaudDib fucked around with this message at 18:43 on Jun 13, 2019 |
![]() |
|
quote:It’s not just Ubi/BF5, though. A lot of other newer AAA titles like Hitman 2 and Shadow of the Tomb Raider are starting to have major frame pacing issues on 4/4 processors. Even if they’re not to the point of completely bottlenecking the GPU, hitching and stutter is pretty terrible for the experience. Yeah I had all sorts of problems on my old 6600k before I moved to a 2600X. Not just BF:V but basically any new game unless it was an indie.
|
![]() |
|
A slightly more recent article pitching a 7600K against a 1600: https://www.techspot.com/review/1859-two-years-later-ryzen-1600-vs-core-i5-7600k/ Those recent game results make it seem like developers have just stopped prioritizing 4/4 CPUs.
|
![]() |
|
Llamadeus posted:A slightly more recent article pitching a 7600K against a 1600: https://www.techspot.com/review/1859-two-years-later-ryzen-1600-vs-core-i5-7600k/ That's kind of the article I was obliquely responding to. So I mean out of those games, at 1080p/overclocked we have:
It is possible to just disagree with a reviewer's conclusions ![]() When it comes down to it, Steve really is basing his conclusion primarily on AC:O having problems. That's the big one where Ryzen is playable and the 7600K is not. And that's with him frontloading the test with the 3 worst titles he could find for the 7600K. (I guess you could also argue that BF:V is a problem because of the gap between min and average, despite having a min that is almost at 60 fps anyway, but that's another title that is just a mess for optimization, and Ryzen is still wandering around by 40 fps too.) Paul MaudDib fucked around with this message at 19:12 on Jun 13, 2019 |
![]() |
|
4/4s still seem viable to me for 60fps gaming, some family members of mine are using stuff like i5 6500's and have no problems. But they are to the point where i would really recommending killing every single running process/bloated windows service you can. If you gamed in the single core era, you will know this routine well.
|
![]() |
|
I'd agree that Ubi problems are an outlier, except more often than not they seem to be the games people want to play, at least in the part picker thread.
|
![]() |
|
Cygni posted:4/4s still seem viable to me for 60fps gaming, some family members of mine are using stuff like i5 6500's and have no problems. But they are to the point where i would really recommending killing every single running process/bloated windows service you can. Exactly, I wouldn't recommend a quad-core for a new midrange/high-end build, and it was overpriced even back during the 6000/7000 series, but at the same time its death has been greatly exaggerated. It still holds 60fps minimum (or real close) in almost all titles and that's the budget standard. A lot of times it's still outperforming the 1600. Probably most times, outside of a pointed editorial from HU. There is a place for 4/4 and 4/8 especially at 5 GHz, and I'm positive that AMD will get around to introducing one once they've filled demand for the enthusiast parts. Again, it's just that the place is more like $75-100, and not $200 like Intel want to charge. ItBreathes posted:I'd agree that Ubi problems are an outlier, except more often than not they seem to be the games people want to play, at least in the part picker thread. Yeah, no accounting for taste ![]() If that's your game then you gotta optimize around it, true. Paul MaudDib fucked around with this message at 19:28 on Jun 13, 2019 |
![]() |
|
The APU point is good, but I have seen other similar reports for Shadow of the Tomb Raider and Hitman 2. SotTR, disabling cores/threads on a 6850k and 1700x. ![]() I am curious why the 1600 performed so poorly in techspot's benchmark, though - their numbers for the 7600k line up pretty well with the 4/4 6850k, but the 1600 isn't even close to the 6/12 1700x (or even the 4/8). (E: Possibly less of a core issue and more overall performance?) Hitman 2 is a little less clear, but the 8100 looks like it has similar 3% frame times to the 2400G: ![]() Maybe it's close enough to the edge that the improved single-core performance of the 7600k can eliminate hitching. I've seen several people complaining about stutter/hitching with those processors, but that could be a number of other things, too. So far as "which wins", though, if you care about playing the games with frame pacing issues, that's a much bigger deal that getting a small amount of extra performance for a game that's already playing smoothly! I'd still like to know what's going on with New Dawn, though - is that a Ryzen issue or a single-core performance issue? Stickman fucked around with this message at 20:16 on Jun 13, 2019 |
![]() |
|
I'll just add anecdotally that Hitman 2 had places in missions where it runs at like 30 fps on my 1080ti with an i5-3550 at 2560x1440 so like that game has some optimization issues on some maps. Mainly thinking of Mumbai and Sapienza though parts of Miami and a few others had drops too. Kinda wish those graphs had 1% frame rates and said which map because something like hawke bay averaged 120+ whule the ICA tutorial was more like 60-80. MagusDraco fucked around with this message at 20:19 on Jun 13, 2019 |
![]() |
|
MagusDraco posted:I'll just add anecdotally that Hitman 2 had places in missions where it runs at like 30 fps on my 1080ti with an i5-3550 at 2560x1440 so like that game has some optimization issues on some maps. If you increased the Simulation Quality beyond the lowest setting then that's to be expected, the higher settings are specifically meant for >4 core systems. quote:Simulation Quality
|
![]() |
|
Pretty sure I was on base the entire time so dunno what my dumb computer does. It's been awhile since I played the game though. Fell off sometime after the first elusive target Edit: unless that wasn't an option at all when the game launched and defaulted to higher than base. That could explain some of it. Sapienza still ran poorly in the town square/circle near the beach but whatever
|
![]() |
|
Paul MaudDib posted:There is a place for 4/4 and 4/8 especially at 5 GHz, and I'm positive that AMD will get around to introducing one once they've filled demand for the enthusiast parts. Again, it's just that the place is more like $75-100, and not $200 like Intel want to charge. Maybe not though? With 8 cores on their chips I doubt there will be a ton of parts that are 50% damaged but still work, so there probably won't be a large stock of salvage chips to sell as 4c parts. And 7nm is expensive, so you'd rather not sell those for $100 if you didn't have to. They might just keep the GloFlo 14+ active for the APUs and cheap low-end 4 core parts. As much as I agree that a 4c/8t cpu that can get close to 5ghz would be a pretty good budget gaming option, AMD might not want to sell them. They've had problems before with undercutting themselves. Intel isn't competing in that market at all, so AMD doesn't have to.
|
![]() |
|
Especially given the cost of the IO die and packaging, I guess. First-gen Ryzen was a monolithic package too. No budget parts on 7nm until next year+ would be kind of disappointing though.
|
![]() |
|
Klyith posted:Maybe not though? With 8 cores on their chips I doubt there will be a ton of parts that are 50% damaged but still work, so there probably won't be a large stock of salvage chips to sell as 4c parts. And 7nm is expensive, so you'd rather not sell those for $100 if you didn't have to. Salvaged 4 core chiplets could just be used as the other chiplet in a 12 core 3900X.
|
![]() |
|
BeastOfExmoor posted:Salvaged 4 core chiplets could just be used as the other chiplet in a 12 core 3900X. Ryzen parts have a fixed config, they don't do die harvesting like that. A 12C will always be 6+6 core dies. edit: and what's more, the CCX config per die will be fixed too - 12C will always be (3+3) + (3+3). Paul MaudDib fucked around with this message at 23:25 on Jun 13, 2019 |
![]() |
|
Paul MaudDib posted:Ryzen parts have a fixed config, they don't do die harvesting like that. A 12C will always be 6+6 core dies. Weird, I would've sworn I'd seen it reported as 8+4 at Computex, but Anandtech seems to agree with you.
|
![]() |
|
Would it be worth shelling out the extra 35 or so bucks for a 2200G over an Athlon 220G?
|
![]() |
|
The parts in the stack are all price engineered anyway. You aren't actually paying that price difference from top to bottom of the stack because of any real differences in production cost (with the exception of the 2 die parts obvi). You are paying that price difference because AMD thinks it can get you to pay it. The bill of materials for a 3800X will be something like $40 with the HSF and box, if that. But AMD needs to recoup those development costs, and you do that by charging a premium on the high end. If they think there is demand for a 4/4 part, they will sell it. Sure, they will probably reuse some salvaged dies if they can, but the majority will be fully functional dies they disable. The Deneb 2 and 3 core parts are a great example of that. That said, by spinning off the IO, they have increased the likelihood that defects actually impact a CPU trace and nothing else, so it is possible the salvage rate is higher with Zen2 than prior parts. Only TSMC/AMD know those numbers now, i bet. Worth saying that AMD did release a Zen+ 4/4 part (the 2300X) but from what I can tell, the vast majority went into low end gaming prebuilts for like ibuypower and stuff.
|
![]() |
|
![]()
|
# ? May 18, 2024 18:46 |
|
https://twitter.com/1usmus/status/1139210323976773633?s=19
|
![]() |