|
Nick Esasky posted:so, question about how this AMD Starfield promo works: I just got a shiny new PC built for me with a 7700 CPU + a 6750 XT GPU in it, so i'm theoretically eligible to get both the Standard + Premium edition . I have two coupon codes on my recipe, but i have no idea WTH either of them correspond to. If i toss them at AMD's coupon site, will it just check my PC and give me the Premium since i have the video card installed?
|
#
?
Jul 15, 2023 01:14
|
|
|
|
| # ? May 30, 2024 13:54 |
|
|
Hi everyone, I'm living in an apartment with terrible power, and I'd like to reduce my wattage during heavy gaming. I've heard about a voltatge "offset" and I went poking around, but I don't know where these setting might be or what they're called. I've got an MSI B550 tomahawk with a 5800X3D. I'm not looking to push the envelope, but would like to know what a safe-and-sane reduction might be... and where I might find it. Thanks!
|
|
#
?
Jul 15, 2023 20:40
|
|
|
AARP LARPer posted:Hi everyone, You can use Ryzen Master to set power targets, and that's going to be more effective at reducing wattage than messing around with offsets: https://support.punchtechnology.co.uk/hc/en-us/articles/6438162546461-How-to-set-a-TDP-PPT-Power-Limit-on-AMD-Ryzen-CPU-Processors. Is your machine on a UPS or any other device that measures power usage? How low are you looking to get? What power supply are you using, and is it a correctly sized, efficient one?
|
|
#
?
Jul 15, 2023 21:46
|
|
|
AARP LARPer posted:Hi everyone, In the overclocking section, set CPU Core Voltage to Offset Mode. Then set offset mode mark to - (so negative offset) and your offset to a number. Sane numbers are 0.15 (any CPU should be stable with this) to 0.30 (you need to be somewhat lucky). However, this will not really change power use. The CPU will use about the same power, and the lower voltage may allow it to hit higher clockspeed. If you actually want to reduce power you should use PBO Eco Mode. This is the best way to limit power while still getting pretty good performance. That's in Overclocking -> Advanced CPU Configuration -> AMD Overclocking -> Precision Boost Overdrive. You should have 3 choices for Eco Mode in that selection menu, 95W 65W and 45W. The 45W will have some real performance impact, but you should try both 95 and 65 in a couple games and see how they go. Depending on which games you like and what your GPU is, you might not notice the difference. Also generally a GPU is consuming way more power than the CPU in an average gaming PC.
|
|
#
?
Jul 15, 2023 21:56
|
|
|
Thanks so much for this! I apprecitate it and yeah, I've addressed the video card power in another thread. I'm good now.
|
|
#
?
Jul 15, 2023 23:55
|
|
|
So I see that AMD hasn't released a newer version of their 3D V-Cache scheduler thingy. Do they think it works awesomely now, or are they just waiting on Microsoft for a generic implementation, similar to the power stuff back with Zen/Zen+?
|
|
#
?
Jul 17, 2023 21:50
|
|
|
Combat Pretzel posted:So I see that AMD hasn't released a newer version of their 3D V-Cache scheduler thingy. Do they think it works awesomely now, or are they just waiting on Microsoft for a generic implementation, similar to the power stuff back with Zen/Zen+? I'm assuming there is no better version planned. This is just how the 7950X3D will be, and if you want something better then you will need to wait for the Zen 5 version. They're trying to bolt a heterogeneous design onto an architecture that was not designed for it, and the hardware scheduler just doesn't have the features to make it work well. That's why AMD is relying on software to detect games and use the brute-force process affinity approach. If they want to do something smarter, they'll need something like the Thread Director intel developed for their big little chips.
|
|
#
?
Jul 17, 2023 23:30
|
|
|
AMD is supposedly going to have big/little chips pretty soon, rumored to be next year on mobile with Strix Point (a mix of 4 Zen5 and 8 Zen5c cores). Hope they are already working on some sort of solution that doesn't rely on the OS scheduler.
|
|
#
?
Jul 18, 2023 00:01
|
|
|
It’s still the OS scheduler, thread director is “hints” on what to run. The kernel’s role is still to execute threads on specific cores.
|
|
#
?
Jul 18, 2023 00:09
|
|
|
The worst of it should have already been shaken out by Intel by this point, then it's just specific architecture peculiarities.
|
|
#
?
Jul 18, 2023 00:11
|
|
|
hobbesmaster posted:It’s still the OS scheduler, thread director is “hints” on what to run. The kernel’s role is still to execute threads on specific cores. Right, Intel still relies heavily on kernel-level scheduling, but I doubt it would perform anywhere near as well without the thread director. AMD lacks anything like that, and it's not something they can just patch in either.
|
|
#
?
Jul 18, 2023 00:12
|
|
|
hobbesmaster posted:It’s still the OS scheduler, thread director is “hints” on what to run. The kernel’s role is still to execute threads on specific cores. sure but we know from recent experience that the scheduler in many OSes is extremely basic, which is what thread director is really all about efb
|
|
#
?
Jul 18, 2023 00:24
|
|
|
Cygni posted:sure but we know from recent experience that the scheduler in many OSes is extremely basic, which is what thread director is really all about Windows actually does an amazing job considering what is demanded of it as a general purpose desktop OS.
|
|
#
?
Jul 18, 2023 00:43
|
|
|
https://twitter.com/Buildzoid1/status/1681454391252094976?t=uoX3MprbhCFKj9YhdRGpxA&s=19 Love to vicariously live out OC tinkering through Buildzoid
|
|
#
?
Jul 19, 2023 05:10
|
|
|
I should probably update my BIOS and see if I can get this poo poo board POSTing on DDR5-6000 now that I've fixed the random lockup issue.
|
|
#
?
Jul 19, 2023 05:33
|
|
|
Yeah, ASRock's new bios allows for really high memory speeds, but it may not help performance much in most things because the uclock goes into 2:1 mode with the memclock. As seen in that screenshot, 4000 memclock with 2000 uclock, while on a DDR5-6000 EXPO kit, you'll get 3000 memclock and uclock.
|
|
#
?
Jul 19, 2023 08:05
|
|
|
AMD have repeatedly stated that DDR5 at 6000MT/s is the sweetspot for Ryzen 7000-series CPUs, and everything we've seen so far seems to bear this out. If we can get 4x 48GB UDIMM DIMM-wide-ECC DDR5 at 6000MT/s at some point between now and when the Zen 5 X3D chips have come out and dropped in price, I'll be ecstatic.
|
|
#
?
Jul 19, 2023 10:03
|
|
|
Dr. Video Games 0031 posted:Yeah, ASRock's new bios allows for really high memory speeds, but it may not help performance much in most things because the uclock goes into 2:1 mode with the memclock. As seen in that screenshot, 4000 memclock with 2000 uclock, while on a DDR5-6000 EXPO kit, you'll get 3000 memclock and uclock. I've got an ASUS board that, at launch, wasn't able to consistently POST above 4800, even with kits that were rated by both ASUS and AMD as being valid at 6000. I'm hoping that nine months and several BIOS updates later it'll be able to do so as advertised. Dozens of claims on reddit of this have been posted and it's just beyond hosed. ASUS really hosed the dog on this board, but hopefully after almost a year of incremental improvements maybe it'll do what they said it would when it came on the market.
|
|
#
?
Jul 19, 2023 11:14
|
|
|
Dr. Video Games 0031 posted:Yeah, ASRock's new bios allows for really high memory speeds, but it may not help performance much in most things because the uclock goes into 2:1 mode with the memclock. As seen in that screenshot, 4000 memclock with 2000 uclock, while on a DDR5-6000 EXPO kit, you'll get 3000 memclock and uclock. It's not an ASRock thing, it's the newest AGESA 1.0.0.7b. An AMD employee spoke about it here. The main change is the new "DDR5 Nitro Mode" (very AMD name), which is the frequency-at-all-costs 2:1 mode. I fiddled around with it on my X670E Hero, and it's not just a binary setting, there is a whole page of settings that you can tweak. The descriptions mention latency costs for Nitro mode settings even beyond the cost of dropping the UCLK, so it remains to see how viable it is for gaming. Plus the training times as soon as you enable nitro mode are absurd, like 5 minutes to train 7200 on my board. The other part that the employee mentions is tweaks to enable 6400 in 1:1 on more CPUs, and sure enough, I can now run 6400C28 1:1 with my 2x32GB M-Die kit. Was impossible to stabilise above 6200 on 1.0.0.7A Patch A.
|
|
#
?
Jul 19, 2023 13:14
|
|
|
BurritoJustice posted:It's not an ASRock thing, it's the newest AGESA 1.0.0.7b. An AMD employee spoke about it here. The main change is the new "DDR5 Nitro Mode" (very AMD name), which is the frequency-at-all-costs 2:1 mode. I fiddled around with it on my X670E Hero, and it's not just a binary setting, there is a whole page of settings that you can tweak. The descriptions mention latency costs for Nitro mode settings even beyond the cost of dropping the UCLK, so it remains to see how viable it is for gaming. Plus the training times as soon as you enable nitro mode are absurd, like 5 minutes to train 7200 on my board. compared to some of my 10 min training times with gigabyte and then telling me nah gently caress you go to stock settings, I will take 5 min training times.
|
|
#
?
Jul 19, 2023 17:21
|
|
|
wargames posted:compared to some of my 10 min training times with gigabyte and then telling me nah gently caress you go to stock settings, I will take 5 min training times. Zen/Zen+ was crazy.
|
|
#
?
Jul 19, 2023 17:23
|
|
|
hobbesmaster posted:Zen/Zen+ was crazy. this is a b650 and 7700x.
|
|
#
?
Jul 19, 2023 17:24
|
|
|
Cygni posted:AMD is supposedly going to have big/little chips pretty soon, rumored to be next year on mobile with Strix Point (a mix of 4 Zen5 and 8 Zen5c cores). Hope they are already working on some sort of solution that doesn't rely on the OS scheduler. For one thing, the ISA differentiation has caused no end of trouble for production systems where people would experience software crashes, but having to rely on Yet Another Firmware in the form of Intels Thread Scheduler really doesn't seem like the way forward, since it just introduces even more vendor lock-in.
|
|
#
?
Jul 19, 2023 17:26
|
|
|
wargames posted:this is a b650 and 7700x. Everything old is new again
|
|
#
?
Jul 19, 2023 17:29
|
|
|
BlankSystemDaemon posted:Turns out, it won't really be a traditionally-heterogenous multi-processing setup like ARM and Intel have done with their HMPs. So yeah, I assume AMD has a plan here, and I hope they do... because the workaround for the similar but not identical dual-die X3D part issue was not super confidence inspiring. e: apparently the 5c cores will have SMT enabled, so there are actually 4 performance states to calculate. from today's news: https://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=996435 Cygni fucked around with this message at 21:44 on Jul 19, 2023 |
|
#
?
Jul 19, 2023 20:06
|
|
|
Double-post because its unrelated: For server land, STH has both Bergamo and Genoa-X in hand and has benchmark numbers: https://www.youtube.com/watch?v=BxLBLEeq6yg
|
|
#
?
Jul 19, 2023 20:19
|
|
|
Cygni posted:Underneath the marketing speak in that interview, they still have some of the exact same problems for mobile/desktop as Alder/Rocket (and Lakefield before it) had. Regardless of ISAs or theoretical max IPC of the cores themselves, they will have fast power hungry cores with lots of cache, moderate efficiently clocked cores with less cache, and slow SMT threads on the big cores... all of which have dynamic performance based on power and heat targets, and will be working with OSes that are making decisions of what to put where based on high level and slow to implement assumptions. It might be simpler for AMD to address than on ARM or Intel, but they still have to address the issue to maximize the design. Relying on the OS scheduler making adjustments in the 10-100s of millisecond range without knowledge of the underlying architecture efficiency tables isn't gonna be ideal. I didn't see that as marketing speak, but rather intentional ambiguity because he wants to basically walk back some of what Papermaster said, without outright just saying that or making it look like what Papermaster said has been wrong. My personal interpretation of that interview is that he indicated that AMD still has relatively little interest in going to LITTLE.big/etc., and instead would rather optimize the cores for their specific uses, i.e. Zen 4c/5c could be ideal for power-efficient laptops, servers, etc., while standard Zen 4/5 could go into gaming laptops, desktops, etc., where power efficiency is less of a concern. So it isn't the Alder/Rocket Lake problem at all. And to me it makes sense, because unless you're applying it to a closed hardware system like a mobile device or Apple's closed ecosystem where you can really optimize for heterogeneous cores, then you'll end up like the Alder Lake/etc. circumstances. Intel could have just thrown more efficiency cores in lieu of performance cores and called it a day, and Alder Lake would probably have been even better? Edit: Ironically if the performance ends up being accurate to what we saw, then Zen 4c could be AMD's "Pentium M-> Core" moment of sorts. Canned Sunshine fucked around with this message at 15:39 on Jul 21, 2023 |
|
#
?
Jul 21, 2023 15:37
|
|
|
What is an OK temp for a 5600X3D if I run something like Prime95 for 5-10 minutes? Just want to see if the cooler I have is sufficient, or if I need a better one. I would be using the temps reported in HWMonitor to verify.
|
|
#
?
Jul 21, 2023 15:50
|
|
|
low 80s for cinebench multicore on 5800X3D is pretty normal
|
|
#
?
Jul 21, 2023 15:51
|
|
|
chocolateTHUNDER posted:What is an OK temp for a 5600X3D if I run something like Prime95 for 5-10 minutes? Just want to see if the cooler I have is sufficient, or if I need a better one. I would be using the temps reported in HWMonitor to verify. I think prime95 small FFTs will usually hit 90 on the zen 3 3Ds regardless. For the more memory focused tests or something like cinebench what temperature are you seeing and what frequency? 4.35GHz all core stable should be possible but would by no means be necessary to get good gaming performance because that’ll be mainly be heavily loading just a couple cores. hobbesmaster fucked around with this message at 18:09 on Jul 21, 2023 |
|
#
?
Jul 21, 2023 17:23
|
|
|
My 5800X3D sees 4.42 all core stable at ~70C under "real" all-core workloads and still hits 90C with prime95 small FFT. That test is just brutal on these things.
|
|
#
?
Jul 21, 2023 17:43
|
|
|
also just happened to read an aio review, and there's a decent tpu overview of cooler performance, regular and noise-normalized
|
|
#
?
Jul 21, 2023 22:55
|
|
|
Looks like Storm Peak is getting close to launch (the PRO WX parts are gonna be server Genoa with a chipset hung off of it for HEDT)
|
|
#
?
Jul 22, 2023 01:45
|
|
|
So I guess 4x32 works better now on AM5 than at launch?
|
|
#
?
Jul 22, 2023 01:59
|
|
|
https://twitter.com/harukaze5719/status/1682927434314948609?t=xtX0fyPeZ6LQIQpIBmHf_g&s=19 Supposed to be launching this month?
|
|
#
?
Jul 23, 2023 06:47
|
|
|
people generally need to suck it up and dwi, 7000 series is getting cheap enough it's pretty much just better now. Getting 7600X or whatever is really fine too. Especially now that 4x32GB+ appears to be getting stable at reasonable clocks on gear 2. that's fine, especially with the 7800X3D. with some of those skus being APU cutdowns probably, people should carefully consider the implications of the particular SKU they're buying in terms of pcie lanes (some may be x8) and cache configurations, reduced cache on APUs will hurt games somewhat. I, again, think 7600X bundle (often less than 7600) and 7800X3D are the ones that make sense for most people. 7800X3D is a value winner that helps drop your VRM power while boosting performance, get that Asrock Lite whatever. I used to still think AM4 would be better for UDIMM server builds because of 4x32 easy compatibility and I think AM5 is probably good enough to not matter now, (for gamers) especially with the ram bundle deals. otoh, zen5 is going to be a big deal when it comes out in probably end of next year or w/e. legit big improvement, zen4 was conservative to port onto DDR5 and zen5 is the big rewrite. it should be fairly significantly better, although I think they're taking the same big/little dive a little bit and gaming enthusiasts still want 8C X3D all-p-core type skus. Paul MaudDib fucked around with this message at 08:00 on Jul 23, 2023 |
|
#
?
Jul 23, 2023 07:55
|
|
|
The 7800X3D is not really a good value when the 7700X is almost as good in most games and better in non-gaming tasks while costing $100 less. If you're blessed with a Micro Center, you can get a 7700X, a decent motherboard, and 32GB of RAM for the same price as the 7800X3D on its own. That's an impossible to beat value, to be honest. The 7800X3D basically only makes sense for top-of-the-line builds or people obsessed with sims. The 7500F is not an APU, and it doesn't have any less cache per core than the rest of the Zen 4 non-3D lineup. It's a 7600X with no iGPU, 1GHz lower base clock, and 300 MHz lower boost clock. Base clock is largely irrelevant these days, so it will only be marginally slower than the 7600X in practice. It could be a solid mainstream choice if the pricing is good, though right now it'll only be $20 less than the 7600X, making it a bit more questionable. Dr. Video Games 0031 fucked around with this message at 08:39 on Jul 23, 2023 |
|
#
?
Jul 23, 2023 08:35
|
|
|
I don't think the CPU prices themselves have ever been the problem with AM5 adoption. DDR5 is in a much better spot than it was around the 7000 series launch but motherboards still look about the same. 235-250 eur for a barebones PCIE 4.0 itx board feels rough.
|
|
#
?
Jul 23, 2023 10:22
|
|
|
You can't divorce the CPU prices from the motherboard prices. Since AM5 is a new platform, when you're buying one, you're buying the other. A discount on CPUs is as good as a discount on motherboards, and these lower CPU prices makes the AM5 platform much more affordable. But also, there have been some modest discounts on motherboards too, as well as some fairly good bundle prices on Newegg (and some killer bundle prices at Micro Center—see above). For shoppers on a budget, there are even some acceptable B650 motherboards in the $110 - $150 range now too.
Dr. Video Games 0031 fucked around with this message at 10:28 on Jul 23, 2023 |
|
#
?
Jul 23, 2023 10:25
|
|
|
|
| # ? May 30, 2024 13:54 |
|
|
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.
|
|
#
?
Jul 23, 2023 13:16
|
|