|
Cygni posted:I bought some 3600CL16 2x16 B-Die when I upgraded, and i absolutely don't think its worth it from what I've seen. Here's my testing. IF/UCLK always in sync, everything is 2 ranks per channel, dual channel on, no PBO or overclocking, 3090 FT3 Ultra with stock settings, same boot drive, win/driver version etc: Awesome, thanks!
|
# ? May 6, 2021 05:10 |
|
|
# ? Apr 23, 2024 10:37 |
|
This isn't surprising. Most 2x16 kits of B-Die are ok but not amazing. Most overclocking was done on 2x8 kits where they can single rank on the DIMMs IIRC. I have a mediocre set and it takes at least 1.42V just to hold 3600 16-17-16 which the ICs are known to handle just fine. Perfectly satisfied with that but not really delighted.
|
# ? May 6, 2021 06:16 |
|
Yeah I tried both 3800CL16 with 1:1 IF and 3600CL14 and neither really gave any measurable gaming performance increase. That’s some of the data that got messed up by windows update and isn’t comparable, so I didn’t include it on that chart. I imagine there might be better gains in games that are CPU limited and crank 300+ FPS but those aren’t really the kinda games I care about, personally, so I didn’t test any.
|
# ? May 6, 2021 06:35 |
|
Grain of salt, but going up a node for a console revision is interesting: https://twitter.com/serkantoto/status/1390244523712794630
|
# ? May 6, 2021 11:25 |
|
kliras posted:Grain of salt, but going up a node for a console revision is interesting: I suspect 6 nm should just be an updated version of the current 7 nm, possibly a half-node, so it's not going up a node. I guess it's just going to be too expensive to get a full shrink to 5 nm in the foreseeable future.
|
# ? May 6, 2021 11:40 |
|
Oh I read it as the consoles being 5nm now and going up to 6.
|
# ? May 6, 2021 11:52 |
|
I'm not entirely sure, but I think it's still mostly just Apple using the 5 nm process at TSMC. Consoles should be on 7 nm with the rest of AMD's stuff.
|
# ? May 6, 2021 12:07 |
|
Yeah that's just me not having had my coffee yet and wishing for some minor Slim revision to fix their dumb coil whine and fan noise issues.
|
# ? May 6, 2021 12:53 |
|
You have coil whine on your console? That sucks! I have that on my old Wii console when it's in standby mode (red light), though not when it's on or in that semi-standby mode (yellow light) where it could receive updates/messages over the Internet back in the day.
|
# ? May 6, 2021 13:22 |
|
Nah, I just haven't tried buying one yet because of the litany of weird hardware issues. The controllers also seem to need some rework. Coil whine is one of the worst sounds in the world after I first heard it on my GPU during stress tests. And it's one of those things that reviews struggle to cover.
|
# ? May 6, 2021 13:27 |
|
Just setup my Plex server with my old memory and 2600x and it had a super annoying three tone coil wine as activity happened. I swear it was an arpeggio. I turned it off, turned it on its side and booted it back up and the noise was gone. Thank God, hope it doesn't come back.
|
# ? May 6, 2021 13:34 |
|
Bofast posted:I'm not entirely sure, but I think it's still mostly just Apple using the 5 nm process at TSMC. Consoles should be on 7 nm with the rest of AMD's stuff. Khorne fucked around with this message at 15:50 on May 6, 2021 |
# ? May 6, 2021 15:38 |
|
I havent messed with my G.Skill B-Die much but I do know it can run DDR 3200 @ 1.25v with rank of 1 on a Ryzen board.
|
# ? May 6, 2021 15:56 |
|
when did lithography become a marketing bullet point, anyway? Was Intel bragging about its 180nm process for Netburst processors? Did AMD brag about finally moving to 32nm after K10?
|
# ? May 6, 2021 15:58 |
|
gradenko_2000 posted:when did lithography become a marketing bullet point, anyway? Was Intel bragging about its 180nm process for Netburst processors? Did AMD brag about finally moving to 32nm after K10? Yeah, they were bragging about that
|
# ? May 6, 2021 16:14 |
|
Shemp the Stooge posted:Yeah, they were bragging about that Yep. I can't tell you exactly when it became a thing, but sometime between me being starting highschool (1989) and me starting to work professionally (1996) the narrative shifted from "new this year: the 80486!" to "the [new CPU] will be fabbed on the new 2 micron process".
|
# ? May 6, 2021 18:25 |
|
gradenko_2000 posted:when did lithography become a marketing bullet point, anyway? Was Intel bragging about its 180nm process for Netburst processors? Did AMD brag about finally moving to 32nm after K10? I think sometime around 28 nm both Samsung and TSMC (and probably GlobalFoundries?) started drifting away from fairly standard naming schemes on their processes and adjusted the numbers a bit for marketing, which made them a lot less comparable to Intel who generally stuck to reasonably process numbers (except for the 14+++ nonsense).
|
# ? May 6, 2021 20:17 |
|
Is Warhol/Zen 3+ still on track for holiday season 2021, or is it scrapped? I'm currently targeting fall/winter of this year to make my architecture leap for a new desktop platform.
|
# ? May 7, 2021 16:12 |
|
Nomyth posted:Is Warhol/Zen 3+ still on track for holiday season 2021, or is it scrapped? I'm currently targeting fall/winter of this year to make my architecture leap for a new desktop platform. I don't think there have been any announcements at all and Zen3+ is more an extrapolation from previous practice than anything else. There's not even any hint whether 3+ would be on AM4 or whether AMD will use it for the first run on the new socket.
|
# ? May 7, 2021 16:42 |
|
gradenko_2000 posted:when did lithography become a marketing bullet point, anyway? Was Intel bragging about its 180nm process for Netburst processors? Did AMD brag about finally moving to 32nm after K10? It only STOPPED being a marketing point by Intel when it became abundantly clear that Intel wasn't just having "delays" or "problems" with 10nm, but that something was PROFOUNDLY busted. Because why would you point out that you're still stuck on 14nm+++++++++++++++++++++?
|
# ? May 7, 2021 16:57 |
|
Nomyth posted:Is Warhol/Zen 3+ still on track for holiday season 2021, or is it scrapped? I'm currently targeting fall/winter of this year to make my architecture leap for a new desktop platform. Rumors are that it’s dead, replaced by an “XT” refresh later this year. One of the last rumors was that it was just going to be a revision on the IO die anyway. With all of the drama surrounding adding Zen3 support to older boards and AMD bending on B550/X570 exclusivity, I wonder if adding another range of products to support was more trouble than it was worth, especially when they can barely keep em in stock anyway. GN’s sources did imply that Warhol is/was real, but was not “Ryzen 6000”, for whatever that is worth.
|
# ? May 7, 2021 17:05 |
|
Ahahaha we're gonna get a 5600XT before we even get a 5600 non-X
|
# ? May 7, 2021 17:07 |
|
Fine by me, now if they'd also shitcan Rembrandt, because nobody wants to see another DDR4 APU with Vega cores, they can completely do away with the entire Zen3+ step.
|
# ? May 7, 2021 17:16 |
|
the rumors I've read are that Zen3+ is Zen3 but on AM5, and that probably makes sense as something to bridge onto the new socket so they don't have to delay that launch if the real Zen4 gets delayed into 2022 due to 5nm capacity problems. (or partially delayed, might be another split launch like Zen3 where either server or consumer launch but not both at the same time). Just swap in a new IO die and they can keep the rest of the process moving. As far as chipset support... I highly suspect AMD is going to avoid making the big promise of "4 years of socket lifetime with no motherboard swaps" this time around. They seem to be pushing for a rolling window of compatibility, a chipset supports 1 generation behind and one generation ahead of the chips it's released alongside or something like that. That's obviously what they wanted to do with Zen2/Zen3 support. Which is fine but kind of removes the possibility of dropping a big upgrade into your PC in a couple years, since the only chips a motherboard will support will be next year's, you can't skip 2-3 generations anymore. Paul MaudDib fucked around with this message at 17:38 on May 7, 2021 |
# ? May 7, 2021 17:31 |
|
I’m really curious what FCLK a zen 3 chiplet with a new ddr5 io die could run.
|
# ? May 7, 2021 17:32 |
|
AMD previously did a model where AM3 and FM2 would support the first two generations of a CPU, and then you'd need a new board with AM3+ or FM2+ for the newer CPUs, but those sockets/boards would still support the older CPUs It was sort of a halfway compromise from Intel doing hard changeovers every two generations I feel like AMD ditched that at first because they were on the back foot after Bulldozer and wanted to offer more value, but by the time they were doing Zen 3, they knew they were successful and competitive enough that they could force a socket change and still sell through their product. They even tried with the BIOS fuckery until their hand got forced by backlash, but I'd bet that if they had a physical socket ready to go they would have done it, completely bisecting the stack after Zen 2. A320, B350, and X370 would support Bristol Ridge, Zen 1, Zen+, and Zen 2, and A520, B550, and X570 would support Zen 3 on an AM4+ socket gradenko_2000 fucked around with this message at 18:01 on May 7, 2021 |
# ? May 7, 2021 17:59 |
|
I also wouldn't mind if x86 cooler mounting points were unified between AMD and Intel, AM5 would be a great time to do it.
|
# ? May 7, 2021 18:03 |
|
Intel is moving to new (oblong) mounting points of their own later this year, so prolly won’t happen. Still haven’t heard a peep about AM5 being LGA or PGA, either. I imagine it’s LGA for pin density.
|
# ? May 7, 2021 18:11 |
|
Cygni posted:Intel is moving to new (oblong) mounting points of their own later this year, so prolly won’t happen. Still haven’t heard a peep about AM5 being LGA or PGA, either. I imagine it’s LGA for pin density. Bring back cartridge cpus like pentium II! (No don’t!!)
|
# ? May 7, 2021 18:15 |
|
priznat posted:Bring back cartridge cpus like pentium II! (No don’t!!) They're great as long as you pop it out and blow into the contacts once in a while
|
# ? May 7, 2021 21:16 |
|
Denver Microcenter has 5950s, 5800s, and 5600s in stock today, but no 5900s. This is the first time I've seen a 5950 in stock since release week, I think. I know the global semiconductor shortage is far from over, but maybe supply on Zen 3 is -- after six months -- beginning to normalize?
|
# ? May 7, 2021 21:57 |
|
mdxi posted:Denver Microcenter has 5950s, 5800s, and 5600s in stock today, but no 5900s. This is the first time I've seen a 5950 in stock since release week, I think. You can't mine shitcoins with it, so without the insane demand GPU coin miners are putting on GPU inventory, inventory is slowly beginning to transition from 'paper launch' to 'in stock, on shelves'.
|
# ? May 7, 2021 22:07 |
|
mdxi posted:Denver Microcenter has 5950s, 5800s, and 5600s in stock today, but no 5900s. This is the first time I've seen a 5950 in stock since release week, I think. Yeah today I noticed that my local Microcenter has the full lineup at MSRP for the first time I have seen.
|
# ? May 7, 2021 23:59 |
|
Cygni posted:Rumors are that it’s dead, replaced by an “XT” refresh later this year. One of the last rumors was that it was just going to be a revision on the IO die anyway. With all of the drama surrounding adding Zen3 support to older boards and AMD bending on B550/X570 exclusivity, I wonder if adding another range of products to support was more trouble than it was worth, especially when they can barely keep em in stock anyway. Ah, I didn't catch the fact that AMD was going to let Intel make the first move this winter. I'm late getting off of the Skylake train and having some doubts about the ability of Alder Lake (+ Microsoft OS scheduling) to cater to the DIY crowd
|
# ? May 8, 2021 01:27 |
|
Any rumors on Van Gogh? Jonesin' for a 20W NUC with Windows, ODROID H2+ with Linux, and a Mac Mini M2 all hanging off the back of the same monitor
|
# ? May 8, 2021 01:37 |
|
Nomyth posted:Ah, I didn't catch the fact that AMD was going to let Intel make the first move this winter. I'm late getting off of the Skylake train and having some doubts about the ability of Alder Lake (+ Microsoft OS scheduling) to cater to the DIY crowd There are also rumors that Zen5 is, itself, a big-little design lol. But thats probably talkin 2024-2025 or somethin. Far enough away that the rumors are probably not super accurate.
|
# ? May 8, 2021 02:10 |
|
Cygni posted:There are also rumors that Zen5 is, itself, a big-little design lol. But thats probably talkin 2024-2025 or somethin. Far enough away that the rumors are probably not super accurate. article for reference: https://videocardz.com/newz/amd-3nm-zen5-apus-codenamed-strix-point-rumored-to-feature-big-little-cores it is probably important to distinguish between Zen the architecture and Ryzen the product here. Zen5 will be the "big" core in this hypothetical configuration, combined with an unknown "little" core. It is probable that AMD will keep offering Zen-only designs for enthusiasts with no little cores, because it costs them basically nothing to put 2-4 Zen chiplets on a package instead of 2 Zen + 2 little chiplets (f.ex). Although I suppose an assumption I'm making here is that AMD will have Zen-only chiplets and Little-only chiplets. If the chiplets are actually mixed in and of themselves, and they don't have a big-only chiplet, that assumption falls apart. But I think the server market probably will want "big-only' chiplets and if they exist, then it costs AMD basically nothing to offer a product using only big chiplets. APUs will likely be mixed since that's probably an attractive approach for power optimization (off-chiplet infinity fabric is always relatively power-hungry compared to a monolithic die).
|
# ? May 8, 2021 02:39 |
|
Paul MaudDib posted:article for reference: https://videocardz.com/newz/amd-3nm-zen5-apus-codenamed-strix-point-rumored-to-feature-big-little-cores Paul, you're more familiar with Intel than I am, are you aware of any functional benefits to die-stacking a la Foveros with regards to power consumption? Because I think there's still an AMD patent from a handful of years back when we were still speculating that the IO die was going to be an interposer that Zen chiplets and an HBM bump were stacked onto.
|
# ? May 8, 2021 03:08 |
|
Paul MaudDib posted:Although I suppose an assumption I'm making here is that AMD will have Zen-only chiplets and Little-only chiplets. If the chiplets are actually mixed in and of themselves, and they don't have a big-only chiplet, that assumption falls apart. But I think the server market probably will want "big-only' chiplets and if they exist, then it costs AMD basically nothing to offer a product using only big chiplets. Thats a good point, I hadnt considered the little cores being off on their own ~Shame chiplet~. There is gonna be a lot of complexity in this next wave of chiplet/tile, big-little, fighting ARM, fully SoC tom foolery were gonna be entering.
|
# ? May 8, 2021 03:13 |
|
|
# ? Apr 23, 2024 10:37 |
|
And just think, if AMD hadn't come back, we'd still be staring at 4c/8t on 14nm.
|
# ? May 8, 2021 03:22 |