|
https://twitter.com/g01d3nm4ng0/status/1724397264477561005?t=YnLrs1mZwipvwtCn4fIBuQ&s=19
|
# ? Nov 14, 2023 13:22 |
|
|
# ? Jun 6, 2024 07:02 |
|
not sure how those make sense but I'd take a 5700X3D for my B450 system, maybe
|
# ? Nov 14, 2023 13:37 |
|
I'd be interested in the power draw. If they can get them at 5600x levels that would be very attractive It would mean less power consumption, less heat generated, smaller cooling factors for smaller systems etc while being fast as heck
|
# ? Nov 14, 2023 14:46 |
|
gradenko_2000 posted:https://twitter.com/g01d3nm4ng0/status/1724397264477561005?t=YnLrs1mZwipvwtCn4fIBuQ&s=19 So I guess AMD has a lot of leftover Zen 3 inventory that they're struggling to sell unless they throw v-cache at it? This may also be a sign that the packaging has gotten much cheaper. The 5500's biggest weakness before was that it had reduced cache vs the 5600(X), which significantly hurt gaming performance. Slightly lower clock speeds compared to the 5600X3D does not seem as big of a downgrade.
|
# ? Nov 14, 2023 14:51 |
Dr. Video Games 0031 posted:So I guess AMD has a lot of leftover Zen 3 inventory that they're struggling to sell unless they throw v-cache at it? I was trying to figure out how this made sense, and I think this is it. They have some dies that can't hit 4.5GHz under the vcache blanket, but they also don't think they can sell them any other way.
|
|
# ? Nov 14, 2023 16:54 |
|
Zedsdeadbaby posted:I'd be interested in the power draw. If they can get them at 5600x levels that would be very attractive 5800X3D has less max power draw than normal 5800X, so the same will probably apply to the others. But TBQH if you want to make a low power draw system for like SFF or whatever, you can take a 5800X3D or any other Ryzen and just set a lower PPT. You don't need a special CPU. They are very efficient at this. Theris posted:I was trying to figure out how this made sense, and I think this is it. They have some dies that can't hit 4.5GHz under the vcache blanket, but they also don't think they can sell them any other way. I'd be very skeptical that you could take "leftover" dies that are already cut out from the wafers and turn them into X3D chips. The CPU die needs to be thinned down, which I would think is something you'd have to do during the fab process while they're all on the big wafer? Also at this point there's no way that zen 3 process is producing scads of flawed chips. I'd go with: 1. X3D packaging tech is mature / cheaper as Dr VG said. 2. AMD bought fab capacity from TSMC that they're having trouble filling. The market is down. Producing a bunch of cheaper X3D chips for the old platform might entice people on AM4 who aren't buying a shiny new AM5 system to throw some money on a CPU upgrade. I'd certainly be tempted, if this is true. I have a 3700X, which I bought back when AMD was saying that X370 would never get support for 5000 CPUs. I loving hate rewarding those shenanigans by them giving them more money for yet another AM4 CPU... But if the price is good I can forgive.
|
# ? Nov 14, 2023 17:17 |
|
The 5500 is a $100 CPU now so I want to see what they price a 5500X3D at
|
# ? Nov 14, 2023 17:41 |
|
Dr. Video Games 0031 posted:So I guess AMD has a lot of leftover Zen 3 inventory that they're struggling to sell unless they throw v-cache at it? Note that these are dies they put v-cache on, and then decided to sell as lower SKUs, not dies that were manufactured to be normal Zen3 that they later decided to put v-cache on. The v-cache dies undergo manufacturing steps fairly early on that the non-v-cache ones don't. They use the same floorplan, but they don't etch out the through-silicon vias for the non-v-cache ones. (That would be a waste of money and time). Presumably they have some that don't meet targets to be the better SKU.
|
# ? Nov 14, 2023 17:54 |
|
I think there's a little confusion here because of the 5600X3D. Those were chips that were explicitly manufactured as 5800X3Ds, but not all their cores met specification while their L3 cache did. It was therefore not too difficult to disable those cores and add software support to read those chips as "5600X3Ds" and have them operate as hexa-core CPUs. The 96MB cache is integrated to the die and isn't something that can be added post-manufacturing. Assuming the rumor is valid, I think what's going on is: - They've noticed by now that the 7800X3D has been the absolute top seller across multiple retailers for months, badly crushing out even their own SKUs - They noticed the 5600X3D, despite being exclusive to an American store, is also selling like crazy - They, for whatever reason, don't want to do a 7600X3D similarly to the 5600 - The old, proven Zen3 designs are a bit easier to re-engineer for v-cache - Despite sales, they've heard some complaints about the price of the 7800X3D, and have noticed how the 5600X and 5800X are generally best-sellers #2 and 3 and they want to put out X3D parts that run a little further up and down the price scale (like, $125-150 for the 5500X3D and $200-250 for the 5700X3D) - Hey presto, we can get some fab time - Hey hey hey, suddenly we have enough units to test the waters with a budget and midline X3D release, to see how it goes and if we should give this a go in the Ryzen Zen4+ series I think it's a little silly to not have them out before the Christmas season, but. We'll have to see if it happens and if it does, how widely available the chips are. If they perform like other X3Ds, I have no doubt they'd sell. From all indications, the 5600X and 5800X are still wildly popular despite being on the "dead" AM4 platform.
|
# ? Nov 14, 2023 18:49 |
|
SpaceDrake posted:The 96MB cache is integrated to the die and isn't something that can be added post-manufacturing. No, the whole point of X3D is that the cache isn't integrated into the CPU die itself. It's on additional pieces of silicon stacked on top. And that is added post-fabrication. TSMC makes a wafer full of CPU dies, AMD takes them elsewhere* and the cache silicon is stacked on during the packaging process, not the fab process. However, the chiplets that become X3D CPUs have to be thinned down such that CPU chiplet + cache silicon has the same height as normal not-X3D CPU chiplets. I don't know how that's done, but I can't imagine anything other than chemical etching? And if that's the case I'm absolutely sure you'd want to do it on a whole wafer, not individual dies. * "malaysia" is popping into my head?
|
# ? Nov 14, 2023 20:08 |
|
Klyith posted:However, the chiplets that become X3D CPUs have to be thinned down such that CPU chiplet + cache silicon has the same height as normal not-X3D CPU chiplets. I don't know how that's done, but I can't imagine anything other than chemical etching? No, that's mechanical. They are basically lapped until they are paper-thin. But that's not the bottleneck, the through-silicon-vias are. The cache chip goes on the backside of the CPU. In order to connect to it, they have to drill holes through the entire die and fill them with metal. This is done by etching, after manufacturing transistors but before building the metal stack. This is a fairly long and expensive process, and AMD does not do it for the dies they don't plan to put vCache on. They use the same floorplan, but people have checked with electron microscopes that normal CPUs just have the spots where the vias would start, but don't have holes drilled through the die. Because of this, they need to decide whether a CPU is going to get vCache fairly early. Some of those chips are not going to be fully intact, and some of them are not going to meet the clock speed requirements for the top-of-the-line stuff. It's presumably these rejects that are now being sold as SKUs lower down the product stack.
|
# ? Nov 14, 2023 20:52 |
|
Is it possible they made too many 5800X3Ds and some of them are getting rebadged?
|
# ? Nov 14, 2023 21:00 |
|
Inept posted:Is it possible they made too many 5800X3Ds and some of them are getting rebadged?
|
# ? Nov 14, 2023 21:45 |
|
Tuna-Fish posted:No, that's mechanical. They are basically lapped until they are paper-thin. Wild! Do you know if that's done on whole wafers, or individual dies after they break them up? Tuna-Fish posted:But that's not the bottleneck, the through-silicon-vias are. [...] Because of this, they need to decide whether a CPU is going to get vCache fairly early. Aha. So these can't be leftover general-purpose Zen 3 dies, but they could be flawed 5800X3D -- or Epyc X3D -- chips. I'm still doubtful they have tons of semi-flawed chips to sell. Microcenter-only 5600X3D? Sure. General availability rather than another limed release thing, no. If their yields had been that terrible this whole time they wouldn't be selling X3Ds for $300. The vias being a big time-sink could jive with "we pre-bought too much TSMC capacity" though. If they can't sell all the regular chips they could be making, make more of the PITA ones that take twice as long, and sell them for slightly more money.
|
# ? Nov 14, 2023 22:49 |
|
The default assumption for me is that it is the regular value engineering reasons and mechanisms leading to new parts. Harvesting may play a role, but generally its not relevant on a mature process. Yields for Zen3 are probably well north of 90%, you arent gonna get many dies with spot errors that only occur specifically in one core on the die. You also are unlikely to go through all the vcaching steps before realizing you have an issue. They might expect Intel to be aggressive with low end pricing soon (especially overseas), and feel the need for low priced parts that can compete in that space without cannibalizing sales of higher priced parts in higher priced markets. Cost sensitive markets have not been super receptive of the high platform costs on AM5, so some new AM4 parts late in the life cycle makes sense. But how do you juice the performance of those AM4 parts to be competitive with the new cheap intel options? By slapping some vcache on em, which also likely has a high yield and deflated costs by now. Seems like a logical conclusion to come to for AMD, honestly.
|
# ? Nov 14, 2023 22:54 |
|
Parametric yields are a thing. There is no way 90%+ of vCache Zen3 dies can do the 5800X3D spec, if they could, you would assume you could find a lot of golden samples that could do much more, there's just too much variability for that. That's why you have lower-tier SKUs that have lower clocks, so you can sell everything you can make.
|
# ? Nov 14, 2023 23:12 |
|
Klyith posted:Wild! Do you know if that's done on whole wafers, or individual dies after they break them up? Lapping/polishing is used a lot in chip making, it's a lot more precise than people think and there are a lot of steps where removing material to an uniform depth is useful.
|
# ? Nov 14, 2023 23:14 |
|
Tuna-Fish posted:Parametric yields are a thing. There is no way 90%+ of vCache Zen3 dies can do the 5800X3D spec, if they could, you would assume you could find a lot of golden samples that could do much more, there's just too much variability for that. That's why you have lower-tier SKUs that have lower clocks, so you can sell everything you can make. The interesting thing here is that voltage and frequency are very conservative for desktop zen 3 on the 5800x3d and the 5600x3d knocks off 100MHz. It would not surprise me if 90%+ can meet 5800x3D specs so long as all 8 cores are ok. On the other hand, am4 x3d has some strict voltage limits so maybe a 65W 5800x3d is indeed much easier to make.
|
# ? Nov 14, 2023 23:59 |
|
Tuna-Fish posted:Parametric yields are a thing. There is no way 90%+ of vCache Zen3 dies can do the 5800X3D spec, if they could, you would assume you could find a lot of golden samples that could do much more, there's just too much variability for that. That's why you have lower-tier SKUs that have lower clocks, so you can sell everything you can make. Much more important than harvesting to save pennies on manufacturing costs is SKU segmentation to support drastically higher priced products. The vast majority of lower-tier SKUs run lower clocks not because they have to for some technical or silicon quality reason, but because they need to in order to justify the existence of the higher priced SKUs. See: the 7800X3D's enforced 5050mhz clock limit. That is how silicon is priced in 2023, and there are lots of industry articles out there that talk about it. Separately, the frequency tables for the 5800X3D were set early in the development of the 3D packaging process over 2 years ago, and is mostly limited by the VID/heat generation. It would not surprise me at all if they are getting extremely high yields of parts able to hit those specs, especially as the the 5800X3D has a hard, lower frequency cap. "Golden samples" may exist and a higher SKU could be created, but that would eat into AM5 sales so there really is no reason to ever launch that kind of SKU. Lower SKUs, however...
|
# ? Nov 15, 2023 00:19 |
|
gradenko_2000 posted:https://twitter.com/g01d3nm4ng0/status/1724397264477561005?t=YnLrs1mZwipvwtCn4fIBuQ&s=19
|
# ? Nov 15, 2023 02:28 |
|
At the very least the 5500x3d may lead to some hilarity in benchmark results.
|
# ? Nov 15, 2023 03:53 |
|
Chirping as a retail grunt, the first mention I saw of the 5600x3d was on June 10th from Tom's Hardware https://www.tomshardware.com/news/amd-ryzen-5-5600x3d-rumor-foretells-of-a-budget-am4-gaming-champ, and we got them the week of June 25th, so about a two-three week lead I guess. so they could be in stores after black friday, or could be in there before. I would assume that we'll get the thread rippers in the run up to black friday (this week), which is gonna be great!
|
# ? Nov 15, 2023 03:54 |
|
Cygni posted:The default assumption for me is that it is the regular value engineering reasons and mechanisms leading to new parts. Harvesting may play a role, but generally its not relevant on a mature process. Yields for Zen3 are probably well north of 90%, you arent gonna get many dies with spot errors that only occur specifically in one core on the die. You also are unlikely to go through all the vcaching steps before realizing you have an issue. yeah, I tend to agree with this. There's a big price/performance gap between an i3-13100 and a non-X3D AM4 part, and while a 7600 can beat an i3, it's also a steep step up. they'd be filling in that segment
|
# ? Nov 15, 2023 04:41 |
|
SpaceDrake posted:I think there's a little confusion here because of the 5600X3D. Those were chips that were explicitly manufactured as 5800X3Ds, but not all their cores met specification while their L3 cache did. It was therefore not too difficult to disable those cores and add software support to read those chips as "5600X3Ds" and have them operate as hexa-core CPUs. VCache is added after the dies are fabbed and tested for defects, 5600X3D isn't physically defective 5800X3D at all. It's a relatively expensive (compared to the cost of a die) process so they don't waste time stacking onto faulty dies. It's just artificially cut to hit a lower end of the market, like the vast majority of CPUs.
|
# ? Nov 15, 2023 08:36 |
|
https://twitter.com/AnhPhuH/status/1724853349013692474 it's still going
|
# ? Nov 16, 2023 11:21 |
|
BurritoJustice posted:VCache is added after the dies are fabbed and tested for defects, 5600X3D isn't physically defective 5800X3D at all. It's a relatively expensive (compared to the cost of a die) process so they don't waste time stacking onto faulty dies. The 5600x3d was supposedly CPUs that failed the VCache addition step. Whether something caused some cores to have physical damage or perhaps their temperature/voltage curves are just terrible.
|
# ? Nov 17, 2023 04:21 |
|
hobbesmaster posted:The 5600x3d was supposedly CPUs that failed the VCache addition step. Whether something caused some cores to have physical damage or perhaps their temperature/voltage curves are just terrible. Isn't the only quote we have to this effect from a Microcentre rep to GamersNexus? I still think it only makes sense as an artificial part, I have no idea why adding the stacked cache fails in a way that kills only two of the cores, whereas it makes a lot of sense to just have a cut down 6 core with good performance to challenge the lower end of the market. It's not a unique thing, almost all CPUs are cut down due to segmentation instead of binning. Especially tiny chiplet CPUs on an extremely mature node.
|
# ? Nov 17, 2023 04:37 |
|
BurritoJustice posted:Isn't the only quote we have to this effect from a Microcentre rep to GamersNexus? Useless anecdata: We have both a 5800x3d and 5600x3d desktop and the 5800x3d can take maximum negative voltage offsets happily whereas the 5600x3d fails core cycler if you touch anything. The zen 3 x3d does have a different voltage/frequency precision boost table from normal zen3 and it’s lower and capped for the same frequencies because of the VCache. It doesn’t seem like too much of a stretch that taking away the “throw voltage at it” lever will cause some marginal parts to fail when you add more stuff to it.
|
# ? Nov 17, 2023 04:49 |
|
BurritoJustice posted:Isn't the only quote we have to this effect from a Microcentre rep to GamersNexus? If it makes sense to sell it as a product segment, why sell it only at one specific US-only retailer where it is available only in-store? That's not a market, it's a rounding error. Maybe AMD later decided that the 5600X3D was a good idea and that's why the supply, which microcenter described to GN as "limited time", has apparently never run out.
|
# ? Nov 17, 2023 14:10 |
|
there's also the possibility that the zen 3 incarnation of v-cache is now cheap and plentiful enough (e: or milan-x sales have just dipped due to zen 4 epyc x3d taking over) to just slap on common zen 3 dies which aren't selling quite as well these days due to actual competition in the segment they're currently in
Anime Schoolgirl fucked around with this message at 19:31 on Nov 17, 2023 |
# ? Nov 17, 2023 19:25 |
|
I guess the tweet rumour about upcoming AM4 CPUs might change what I should do but: My current system is a 5600x + 3060ti with 16 gb of 3200 mhz ram. I've been wondering what would be the last viable moment to purchase a 5800x3d, in hopes of staying on this PC for quite a while? When are those chips expected to hit their rock bottom prices? Not too keen on the used market, even though I could probably score one from a decently trusted forum community domestically in Finland at some point. I know I'm more GPU limited in absolutely most use cases, but I mainly play FFXIV which is notoriously heavy on the CPU as most MMOs are and specifically core frequency on just a few cores. (graphics update may change this with the next expansion but there's absolutely no way to know yet). I do plan to get back to Street Fighter 6 and play Control and Alan Wake eventually, and I'm a dumbass who runs on an UW 1440p 144hz (It's mostly been worth it for FFXIV) but I do have 3 quality monitor arms which makes it decently easy to move my 27" IPS 165hz for competitive gaming if I want to. But I'm genuinely not very bothered about most triple A releases. In fact I believe Baldur's Gate 3 is the only game in aaaages that I've bothered with which has been on the limelight. (And I mostly play it on my 4k TV with a controller). I do plan to upgrade the GPU at some point but I'm not even sure it'll be the next generation, GPU prices are stupid and probably won't go down, I'm almost predicting I won't bother with the 5000 series which is slated for.... 2025 maybe. I know the VRAM on 3060 Ti is crippling but it's no issue for the stuff I care about now. edit: There are some actual posts about comparing the 5600x/5800x to the 5800x3d in FFXIV around and the uplift is nothing to sneeze at.
|
# ? Nov 20, 2023 02:14 |
|
If you have a friend with a microcenter nearby 5600x3d is probably your best bet for staying on AM4 if you are that invested in looking for 5800x3d sales.
hobbesmaster fucked around with this message at 02:50 on Nov 20, 2023 |
# ? Nov 20, 2023 02:23 |
|
Nah, as I said I'm in Finland and the 5600x3d is US/NA exclusive.
|
# ? Nov 20, 2023 02:26 |
|
As an aside question: the whole thing with 7800X3Ds cooking themselves was largely sorted out over the spring and early summer, right? Modern motherboards should ship with BIOS that will ensure the processors don't cook themselves, even when using EXPO? And the main casualty is Asus's reputation? (Also, I appreciate the lengths GN went to investigate that issue, but lol those videos are an absolute mess organizationally.)
|
# ? Nov 20, 2023 04:36 |
|
Yes, new agesa updates have come out that have fixed the issue. They come as part of bios updates. Some mobos that have sat on shelves might not have it, but the issue isn’t instantly destructive so you have more than enough time to update the bios. If you are paranoid, turn off all memory overclocking and you’ll be fine.
|
# ? Nov 20, 2023 05:52 |
|
Kibner posted:Yes, new agesa updates have come out that have fixed the issue. They come as part of bios updates. Some mobos that have sat on shelves might not have it, but the issue isn’t instantly destructive so you have more than enough time to update the bios. If you are paranoid, turn off all memory overclocking and you’ll be fine. That's about what I figured. If I do ultimately go for a 7800, I'll be updating the BIOS first thing just to be sure. Thank you!
|
# ? Nov 20, 2023 05:55 |
|
Tuna-Fish posted:Parametric yields are a thing. There is no way 90%+ of vCache Zen3 dies can do the 5800X3D spec, if they could, you would assume you could find a lot of golden samples that could do much more, there's just too much variability for that. That's why you have lower-tier SKUs that have lower clocks, so you can sell everything you can make. CoWoS (oWo) stacking comes after binning. you know how many chips can do it before you stack them. there are also very few stacking problems that leave you with a fully functional die other than a single core, tbh AMD maintains binning streams for 2C, 3C, 4C, and 8C stacks. They have never released a 6C Epyc X3D stack - that would be a 48C Milan-X. Doesn't exist.
|
# ? Nov 20, 2023 07:48 |
Dessel posted:edit: There are some actual posts about comparing the 5600x/5800x to the 5800x3d in FFXIV around and the uplift is nothing to sneeze at. The first "holy poo poo this thing is amazing" moment I had with the 5800x3d was FFXIV staying perfectly glassy smooth in the middle of a 100+ player hunt train.
|
|
# ? Nov 20, 2023 13:45 |
|
maybe use threadripper on a linux install instead of windows if you want to squeeze all performance out of it: https://twitter.com/phoronix/status/1726973842025803941
|
# ? Nov 22, 2023 14:26 |
|
|
# ? Jun 6, 2024 07:02 |
|
kliras posted:maybe use threadripper on a linux install instead of windows if you want to squeeze all performance out of it: The windows scheduler kind of shits the bed when there are more than 64 threads. This is why it was common knowledge for people who ran 3995WX on windows to turn off SMT in bios, because the extra threads would never help more than hurt. (edit:) As in, just 20% difference in computational loads is actually much better than I expected for a 96-core, 192-thread CPU.
|
# ? Nov 22, 2023 17:44 |