|
Stickman posted:I still don’t understand who would want one of these over a 1600 AF, but I assume they’ll be marketed as “BETTER AT GAMES” (to budget gamers on a 60Hz monitor). I don't want to have to sort through having to get an AF specifically, and a 3600 is outside of my price range, and I *am* on a 60Hz monitor, so... me I guess? 
|
#
?
Apr 24, 2020 18:07
|
|
|
|
| # ? May 15, 2024 19:03 |
|
|
To be fair, they probably *will* be better at games at least for the immediate future, which is why I thought one of the potential use cases is someone building a gaming pc now but who knows they'll want to upgrade to zen3 which AMD says is coming later this year. 3100 now, then drop in a hypothetical 4600/4700(x) in November doesn't sound so bad. Edit: although I guess if you were willing to pay 9900k money for probably 9900k gaming performance you could just... buy a 9900k now so I guess it's a dumb idea after all. VorpalFish fucked around with this message at 18:50 on Apr 24, 2020 |
|
#
?
Apr 24, 2020 18:44
|
|
|
Placeholder makes sense, but it's not likely to be a noticeable difference for gaming unless you have a high-refresh monitor, at which point you'll probably be getting a better processor anyway.gradenko_2000 posted:I don't want to have to sort through having to get an AF specifically, and a 3600 is outside of my price range, and I *am* on a 60Hz monitor, so... me I guess? That's fair enough - the AF isn't exactly widely advertised or distributed. It is super easy to get when it's in stock, though.
|
|
#
?
Apr 24, 2020 19:44
|
|
|
VorpalFish posted:Edit: although I guess if you were willing to pay 9900k money for probably 9900k gaming performance you could just... buy a 9900k now so I guess it's a dumb idea after all. it'll probably be closer to 8700K or 10700K money for 9900K/10700K performance. Also, it'll be immeasurably easier to cool, there's no denying that the 9900K/10700K is a toasty boi and needs some pretty serious cooling or even delidding+lapping or direct die cooling (which you should farm out to siliconlottery or rockit cool for $60, not worth the risk to DIY a soldered processor at home). at this point I wouldn't buy a 9900K, the point of the 9900K was that you could have 4700X performance two years earlier, even with the 9900K occasionally dipping to $400-420 I'd just wait it out at this point. Get a 1600AF or a 3100/3300X if you are dying to build the system right this second, but if your current system can still hang in there like at all, just hold on and wait, the "placeholder" thing is a bit of a waste of money. Paul MaudDib fucked around with this message at 20:14 on Apr 24, 2020 |
|
#
?
Apr 24, 2020 19:53
|
|
|
yeah i wouldnt buy anything right now unless you absolutely have to, honestly. i know people always say that, but this really is one of those times. Intel's new stuff is getting announced in like 6 days with big price cuts, which will likely trigger some AMD price cuts too. And then Zen 3 a bit further out. Cygni fucked around with this message at 20:02 on Apr 24, 2020 |
|
#
?
Apr 24, 2020 20:00
|
|
|
Cygni posted:Yup. Especially when you consider that the boxed cooler will absolutely cost more on the BOM than the compute die in the chip will. The silicon is the cheap part. The total cost to make a Athlon 3300G and a 3800X are roughly identical, probably in the neighborhood of $40. So these parts mostly exist for market segmentation reasons (trying to scoop up all the customers, including OEM, that couldn't be convinced to buy a $160 3600), which pisses a lot of people off but thats how the market survives I guess. I'm guessing this is why AMD is actually happy to constantly drop prices when a new gen gets out. I've observed the low end SKUs for a gen only exist as long as a new gen hasn't been released, then they disappear into obscurity almost immediately after. But dropping a 3600 to $120 probably makes them more money than selling a 3300X at $120 and applies considerable pressure to Intel. Actually I think the 3100 and 3300X are a telegraph in how low they're willing to sell any Zen2 processor for, so ~$120 3600s a couple months after the 4600 releases at $200 is probably really likely.
|
|
#
?
Apr 24, 2020 20:10
|
|
|
Think they'll EOL 14nm Zen when 4000 parts get announced?
|
|
#
?
Apr 24, 2020 23:41
|
|
|
Since my Ryzen 3 2200G + B450 mainboard has issues getting to boot with 4 sticks of RAM at speeds above 2133MHz—have tried different timing combinations with limited to zero success—would upgrading to a Ryzen 4xxx series CPU potentially remedy that? Or is running 4 sticks of RAM in daisy chain simply that limiting? [edit] As an aside, when I looked at the RAM modules in CPU-Z, two of the sticks' DRAM were manufactured by Samsung, and the other two by SK Hynix. Would that also potentially cause issues when trying to run all 4 sticks at higher clocks? Or does DRAM manufacturer not typically matter? All 4 sticks are Corsair Vengeance with matching part numbers. teagone fucked around with this message at 00:21 on Apr 25, 2020 |
|
#
?
Apr 25, 2020 00:18
|
|
|
teagone posted:Since my Ryzen 3 2200G + B450 mainboard has issues getting to boot with 4 sticks of RAM at speeds above 2133MHz—have tried different timing combinations with limited to zero success—would upgrading to a Ryzen 4xxx series CPU potentially remedy that? Or is running 4 sticks of RAM in daisy chain simply that limiting? Raven Ridge (the 2xxxG ryzens) is a first-gen Zen core and memory controller which were bad, but if it can't even do basic DDR 2666 then something is wrong. I'd try going in and re-seating the ram to make sure there isn't some connection that's barely working. (Note that the true MHz is half the DDR speed -- ex 3200 ram runs at 1600 actual MHz. Hopefully you're not trying to OC ram using MHz instead of DDR speed.) Ram modules from different manufactures shouldn't be a problem if they've got the same specs. Corsair's habit of changing them with semi-hidden version numbers on the same model is kinda sucky for OCers who want specific stuff that OCs the best, but any of them should do the rated speed.
|
|
#
?
Apr 25, 2020 01:42
|
|
|
Klyith posted:Raven Ridge (the 2xxxG ryzens) is a first-gen Zen core and memory controller which were bad, but if it can't even do basic DDR 2666 then something is wrong. I'd try going in and re-seating the ram to make sure there isn't some connection that's barely working. Oh yeah, I know how the double data rate frequencies work, haha. My motherboard conveniently has preset OC timings for me to set in the BIOS (under a setting called MSI Memory Try It!), though I never tried any of the preset timings slower than 2933 MHz. Entering the timings manually didn't make a difference either. I'm also sure I could boot the 4 sticks at 2666MHz just fine, but I was more so trying to go after RAM speeds closer to what I paid for (3200). Running 4 sticks at CAS 16 2933 MHz would boot sometimes, but 3066/3200 would always fail, no matter the timings. In the end, for now, I'm just running 2 sticks at the 3200 MHz XMP profile, which works no problem. Was thinking that maybe if I just upgrade the CPU to a desktop Renoir chip, I could slot the extra 2x8GB sticks back in and maybe then I'd be able to run all 4 sticks at 3200 MHz. Otherwise I'd likely just sell the extra RAM or something, heh. [edit] That said, I'll re-slot the RAM in again this weekend and see if maybe the inconsistent booting when running 4 sticks at 2933 MHz was maybe due to one of the sticks not being properly seated.
|
|
#
?
Apr 25, 2020 01:56
|
|
|
Seamonster posted:Think they'll EOL 14nm Zen when 4000 parts get announced? Unless if you mean the entire process itself, then it's not likely because of the WSA with Global Foundries requiring wafers sold till i think 2023 or 2025?
|
|
#
?
Apr 25, 2020 02:47
|
|
|
teagone posted:Oh yeah, I know how the double data rate frequencies work, haha. My motherboard conveniently has preset OC timings for me to set in the BIOS (under a setting called MSI Memory Try It!), though I never tried any of the preset timings slower than 2933 MHz. Entering the timings manually didn't make a difference either. I'm also sure I could boot the 4 sticks at 2666MHz just fine, but I was more so trying to go after RAM speeds closer to what I paid for (3200). Running 4 sticks at CAS 16 2933 MHz would boot sometimes, but 3066/3200 would always fail, no matter the timings. Oh I see, you meant to write 3133 in your previous post. Yeah that's not crazy or broken then. Worse than I'd expect but 4 sticks of not B-die or E-die in daisy chain on a Zen 1 CPU is not gonna be great. Surprised that you can't get 3200 no matter what the timings, but it could be the G types are even worse than the normal ones. They've got a lot more going on with the SOC side of the chip after all. (Which is another thing you can try, more SOC volts, but that's a can of worms that I can't help with. All I know is there's not a lot of headroom on the SOC before danger high voltage.) So yes a Zen 2 with the much improved memory controller or upcoming 3 series might solve that. Or it might not. 
|
|
#
?
Apr 25, 2020 04:00
|
|
|
Klyith posted:Oh I see, you meant to write 3133 in your previous post. Yeah that's not crazy or broken then. Worse than I'd expect but 4 sticks of not B-die or E-die in daisy chain on a Zen 1 CPU is not gonna be great. Surprised that you can't get 3200 no matter what the timings, but it could be the G types are even worse than the normal ones. They've got a lot more going on with the SOC side of the chip after all. Oh, no I did actually mean 2133 lol. That's what the BIOS defaults to whenever I set a timing that causes a failed post. I've just never tried setting the RAM speed to anything slower than 2933 MHz. The PC runs a-ok with 4 sticks at 2133; works no problem. I'll try booting with all 4 sticks at 2666 sometime this weekend to rule out any other wonkiness. Another thing I forgot to mention is I also tried bringing up the DRAM voltage when trying for 2933+ speeds on all 4 sticks, but that also didn't make a difference. And I've actually tried increasing SoC voltage in the past to OC the Vega graphics on this 2200G, but I got flustered when I couldn't shake all the lockups/crashes after a handful of attempts lol. So I said screw it and haven't touched the SoC voltage since. If a Zen 2 APU doesn't fix the issue, then I'll just end up selling the RAM or maybe use it for another build or something, heh.
|
|
#
?
Apr 25, 2020 04:24
|
|
|
Seamonster posted:Margins aside, what kind of volume do they really think is going to move? OEMs office boxes? No iGPU... Maybe in lower average income countries? Half the world lives on just five dollars a day
|
|
#
?
Apr 25, 2020 09:45
|
|
|
teagone posted:Oh, no I did actually mean 2133 lol. That's what the BIOS defaults to whenever I set a timing that causes a failed post. I've just never tried setting the RAM speed to anything slower than 2933 MHz. The PC runs a-ok with 4 sticks at 2133; works no problem. I'll try booting with all 4 sticks at 2666 sometime this weekend to rule out any other wonkiness. Another thing I forgot to mention is I also tried bringing up the DRAM voltage when trying for 2933+ speeds on all 4 sticks, but that also didn't make a difference. 2133 is just the default "fallback" JEDEC speed, if your RAM wouldn't even run that something would be seriously wrong The APUs should have the Zen+ memory controller already which is why you can run DDR4-3200 with random memory in the first place. Zen 1 could barely hit more than 2933 without special RAM (eg. B-die stuff). However, I could see the two different sets not running at "rated" speeds together since you're still approaching the limit of what the Zen+ IMC can do (with 4 DIMMs specifically), and with different memory chips that may like different timings/voltage just enough so it breaks near maximum speed (especially with Auto-OC options).
|
|
#
?
Apr 25, 2020 09:56
|
|
|
Anime Schoolgirl posted:It's already EOL, hence the new 1600 and 1200 parts on 12nm. Still surprised there are no 1700 AF but I suppose they probably get snatched up for any Naples orders or replacements. Actually like besides contractual obligations, who the gently caress is even going to order Naples anymore? Is there a use case for it, cause like 24C Rome will beat the snot out of 32C Naples.
|
|
#
?
Apr 25, 2020 10:59
|
|
|
Anime Schoolgirl posted:It's already EOL, hence the new 1600 and 1200 parts on 12nm. Yeah if anything, it's Zen2 that will be EOL since it uses basically the same process as Zen3.
|
|
#
?
Apr 25, 2020 13:14
|
|
|
This is probably a dumb question; but is it the case that generally Ryzen 3000 CPUs run hot? At stock on a Ryzen 3900X on a Asrock B450M Pro4 with a NH-U12S cooler, F@H peaks at around 90C. With the stock cooler it would peak at 95C.
|
|
#
?
Apr 25, 2020 19:52
|
|
|
Evil Robot posted:This is probably a dumb question; but is it the case that generally Ryzen 3000 CPUs run hot? At stock on a Ryzen 3900X on a Asrock B450M Pro4 with a NH-U12S cooler, F@H peaks at around 90C. With the stock cooler it would peak at 95C. That sounds way high. Do you have pbo turned on / no airflow / 40c ambient temps.
|
|
#
?
Apr 25, 2020 19:57
|
|
|
orcane posted:2133 is just the default "fallback" JEDEC speed, if your RAM wouldn't even run that something would be seriously wrong Lol, good to know! And yeah, I guess I'll just wait until a Zen 2 desktop APU hits the market, or maybe a Zen 3 APU if it's still on AM4. We'll see. I'll hold onto the RAM for now.
|
|
#
?
Apr 25, 2020 21:06
|
|
|
Evil Robot posted:This is probably a dumb question; but is it the case that generally Ryzen 3000 CPUs run hot? At stock on a Ryzen 3900X on a Asrock B450M Pro4 with a NH-U12S cooler, F@H peaks at around 90C. With the stock cooler it would peak at 95C. I've got a 3900X, MSI X470, NH-U14S in a Fractal Meshify C case with the fans tuned down to silence.
|
|
#
?
Apr 25, 2020 22:47
|
|
|
Spatial posted:I see around 10C less than you. I have a Fractal Define Mini C which is the "silent" version of your case and runs 3-5 degrees hotter due to less airflow. I have 2x Fractal GP12 x2 fans intake and 1 outtake. Ambient temps are a little high - 25-27C. Just manually turned off PBO in BIOS but it didn't seem to help. Evil Robot fucked around with this message at 23:54 on Apr 25, 2020 |
|
#
?
Apr 25, 2020 23:50
|
|
|
Evil Robot posted:I have a Fractal Define Mini C which is the "silent" version of your case and runs 3-5 degrees hotter due to less airflow. I have 2x Fractal GP12 x2 fans intake and 1 outtake. Ambient temps are a little high - 25-27C. Just manually turned off PBO in BIOS but it didn't seem to help. Just to troubleshoot you could remove the side from your case. If temps drop a lot, it's a case issue. If they only drop a bit, it's a heatsink/mounting issue. Or maybe it's getting way too much voltage for some reason.
|
|
#
?
Apr 26, 2020 00:08
|
|
|
Thansk for the help everyone. I'm still a little mystified. Couple observations:
Evil Robot fucked around with this message at 00:59 on Apr 26, 2020 |
|
#
?
Apr 26, 2020 00:40
|
|
|
SwissArmyDruid posted:IO die is still 14nm, right? Whatever lets them get through their GloFo WSA.
|
|
#
?
Apr 26, 2020 01:35
|
|
|
I guess you could try remounting and repasting the heatsink, but 50 over ambient in prime 95 with that CPU and cooler doesn't sound like you're making bad contact. Are holding pretty steady at the 140w power limit in prime?
|
|
#
?
Apr 26, 2020 01:44
|
|
|
Does F@H use AVX? Are you running an old AVX-less version of Prime (or turned it off in a config file)?
|
|
#
?
Apr 26, 2020 01:51
|
|
|
What are you using to monitor temperatures? My Asus B450 was reporting a 15C offset from ryzen master. My Gigabyte X570 is almost the same as ryzen master.
|
|
#
?
Apr 26, 2020 04:18
|
|
|
Combat Pretzel posted:That WSA still active? I thought they dissolved it, since GloFo cancelled 7nm? https://www.anandtech.com/show/13915/amd-amends-agreement-with-globalfoudries-set-to-buy-wafers-till-2021 They renegotiated it to allow AMD to go anywhere for their 7nm needs without having to pay for the privilege, and set wafer targets through 2021. The WSA is still in effect for anything 12nm+ (that's a 12nm enhanced node, not 12nm and larger) until 2024. Which still leaves their IO dies, their non-bleeding-edge silicon, and their chipsets to be made at GloFo. Who knows if this will incite GloFo to get back into the game, since AMD and Nvidia are snapping up every freed-up wafer from other companies cutting back production because of COVID. When this is all over, AMD, Apple, Nvidia, Qualcomm, and Broadcom will just monopolize every last bit of TSMC's production while leaves those smaller customers that are rolling back now out in the rain in the future. SwissArmyDruid fucked around with this message at 05:52 on Apr 26, 2020 |
|
#
?
Apr 26, 2020 05:46
|
|
|
Evil Robot posted:Strangely prime95 small fft with 24 threads doesn't get anywhere near the F@H temps (maxes out at 78C or so). Is F@H just more demanding than most other benchmarks?
|
|
#
?
Apr 26, 2020 11:15
|
|
|
In that case you can run Furmark/Kombustor together with Prime95
|
|
#
?
Apr 26, 2020 11:25
|
|
|
I'm using the latest version of prime95 and even when I pause GPU folding in F@H I still see a significant temperature differential on my CPU between the two programs.
Evil Robot fucked around with this message at 19:55 on Apr 26, 2020 |
|
#
?
Apr 26, 2020 19:51
|
|
|
Tried one last thing - cleaning / remounting cooler. I actually used the wrong Noctua spacers (supposed to use gray not white with AM4) hoping it would make a difference but nope. Still at 90C in F@H. Just going to deal with it; every day use is fine. PS make fun of my old / new thermal paste job:
|
|
#
?
Apr 26, 2020 21:32
|
|
|
Looks fine to me. You could put even more, you had zero percent spillage!
|
|
#
?
Apr 26, 2020 21:38
|
|
|
Evil Robot posted:Tried one last thing - cleaning / remounting cooler. I actually used the wrong Noctua spacers (supposed to use gray not white with AM4) hoping it would make a difference but nope. Still at 90C in F@H. Just going to deal with it; every day use is fine.  you should probably be able to infer where the thermal paste should be for a dual chiplet
|
|
#
?
Apr 26, 2020 23:27
|
|
|
I'm not convinced that putting more thermal paste over the 3900x chiplets does anything compared to the even coat over the whole IHS you get when you do any reasonable thermal paste application method.
|
|
#
?
Apr 27, 2020 00:27
|
|
|
Anime Schoolgirl posted:the answer key for the 3600 is I'd probably just diagonal line that
|
|
#
?
Apr 27, 2020 00:53
|
|
|
Let's not get into paste nuance. As long as you cover the heat transfer plate you're not going to be able to make a degree of difference.
|
|
#
?
Apr 27, 2020 01:23
|
|
|
Is the typo in the thread title some insider joke I'm not getting or has no one ever asked a mod to fix it?
|
|
#
?
Apr 27, 2020 02:24
|
|
|
|
| # ? May 15, 2024 19:03 |
|
|
mcbexx posted:Is the typo in the thread title some insider joke I'm not getting or has no one ever asked a mod to fix it? AMD made that typo on one of the keynote slides and we're never going to let them forget 
|
|
#
?
Apr 27, 2020 02:27
|
|
