The only "new" chip is the X3D one (with 3d v-cache). The others are the same chips as last year, just at a different price point (eg. the 5700X is basically a cheaper version of the 5800X, only being marginally slower).
Kinda' goes to show the resourcefulnes of salvaging for repurposement based on yields right? Sort of reminds me of the now ancient Black Edition releases with unlockable cores with a big fat YMMV sticker on top.
it depends on market situation but usually those kinds of SKUs (if they're released worldwide in large volumes) are not related to yields but rather artificially disabled.
People hear about binning and assume that every product decision has to be related to binning, but usually it's not, it's just market segmentation. AMD had over 80% of Zen2 chiplets coming off the line with 8 fully-functional cores, and clock bins are generally selected such that most units will pass, by design. And that's at launch, on a new node, in 2019. Numbers have only gotten better over the last 3 years.
AMD already has a bin for iGPUs with a defect - it's 5600G/5600U/5600H/etc. And they have 5300G below that allowing even more defects. There's very very few APUs coming off the line with tons of GPU defects but 6 workable cores, or a defective PCIe controller but only a defect in the iGPU part and not the rest, etc.
The problem is that AMD has tons of supply of high-binned parts but the lowest demand for those parts. And they have the highest demand for low-binned parts but the lowest supply of those parts. How do you mesh those two curves? Disable cores on a high-binned part and sell it as a lower SKU. That's why those "black edition with unlockable cores" existed - those unlockable cores were locked off for market segmentation. Nowadays they just don't let you turn it back on.
(Which isn't to say that none of the 5600X/etc are the result of a dead core/etc - but a lot of them aren't, probably most of them aren't, given the likely >90% yields for 8-core at this point. And you pick your 5600X bin such that most 5800X failures can be sold as 5600X, meaning there's very little that falls through the cracks without being just utterly broken. True binning-generated "we have this pile of chips, let's do something with it" style SKUs tend to have extremely limited availability as a result, it's shit like the Ryzen 4700S or the Ryzen 3100, or the NVIDIA 1650 KO.)
Anyway, it's not a coincidence this is coming a few months after the Alder Lake launch. This is market-driven, Alder Lake is not only faster but in many cases it's cheaper as well. AMD coasted a little bit while motherboard supply firmed up for Intel, but they finally have to respond. I'm sure they're selling lots of Milan but consumer marketshare matters too, and AMD is losing steam there with the price increases, with Intel back on top on performance, and with Intel undercutting their pricing heavily.
I concur, as someone who only pays attention to the processor market twice a decade I'm clueless when it comes to navigating the matrix of CPU market segmentation. It makes me yearn for automobile-style naming where you just have a year, a make, and a model.
Largely those style names do exist in the current product names. Since Intel styled the Core i-series 3, 5, and 7 (and the later 9) over 12 years ago, AMD is following their lead with Ryzen 3, 5, 7, and 9.
The first digit is the generation and then the second digit and suffix letter designate additional features. So a Ryzen 7 5700X is a 5th gen Ryzen 7 meaning it has 8 cores. It is a step down in price and performance to a 5800X and also has no Graphics capabilities (those have a G suffix). Intel uses different letter suffixes such as the Core i7 12700KF being a 12th generation Core i7 with overclocking features (K) and no graphics (F).
It is far worst with GPUs. Certainly an industry wide naming convention reset would be nice but everyone is not going to cooperate like that.
It does take a minute to read up on but at least Intel has *mostly* followed this naming convention for 12 product generations so far.
Even within one vendor it's super confusing. i3? i5? i7? Y series? U series? I think 12th gen Intel may even use different naming conventions already... A Y series i7 of olde delivered a significantly different experience than an i3 U series (literally the difference between being able to run some workloads comfortably vs. locking the machine up). And then marketing materials make it harder because they will drop the CPU model name for the nondescript "i5 CPU". I think it's probably somewhat intentional.
if they don't call out mobile/ULV processors with their own series (Y and U series) then people whine that they're trying to "sneak lower-performing processors into the lineup using the same numbering".
similarly, marketing a processor as being a "5700U" or "5700G" despite it being slower than a desktop 5700X could be seen as equally deceptive. That's not really any better than "i7" vs "m7" or whatever.
unfortunately you're just going to have to learn the naming convention, there really isn't a good solution for that, given the wide range of applications that a given series might be applied to. You just happen to think that AMD's naming convention is worth taking the time to learn while throwing your hands up at the Intel naming. Same with people who think Intel is just awful for internally codenaming all their products after Lakes and Coves while eagerly memorizing every single painter and city name that AMD uses in their codenames - says more about your priorities than their naming scheme.
> You just happen to think that AMD's naming convention is worth taking the time to learn while throwing your hands up at the Intel naming. Same with people who think Intel is just awful for internally codenaming all their products after Lakes and Coves while eagerly memorizing every single painter and city name that AMD uses in their codenames - says more about your priorities than their naming scheme.
The person who was saying they didn't understand the Intel naming convention because U and Y SKUs were too complex to understand.
That's normal. Mobile and ULV SKUs are usually coded differently. AMD calls them "U" and "H" skus I believe.
The person in question just doesn't understand the Intel naming convention, which is fine, but AMD and every other company does the exact same thing. It's not that Intel is uniquely confusing, it's that the individual here doesn't feel that Intel's naming convention is worth the brain space. Which is also fine, but it's not a problem with the naming convention.
As a tangential observation (people do make those in discussions!) architecture/product codenames are another place this comes up. There are many enthusiasts who will eagerly memorize that Rembrandt > Renoir but think the idea that Rocket Lake > Coffee Lake is perplexing and confusing. Or at least that's a repeated theme in many of these naming discussions.
I'm doing my best here to say this politely, but a lot of people clearly just don't value those two bits of knowledge equally. And dipping into rhetorical "who are you even responding to!?" doesn't really further the discussion either.
Naming isn't hard and naming discussions aren't interesting.
As someone who understands the intel naming convention quite well and could understand the AMD naming convention if I had a reason to I can still empathize with random-consumer-x who does not need to have that any practical understanding of the naming convention and could be confused as to what they are really purchasing...
Well, sadly, there is more than one use-case for computers so we need multiple power brackets, so U and Y SKUs are going to continue to exist. It's appropriate to call it out with the naming scheme, but that's exactly what AMD and Intel have done here.
There is no solution which is going to be 100% intuitive to someone who specifically doesn't know anything about what they're looking for. If you move those products out to their own separate series, that's what Intel did with the Y series ("m7-xxxY" line - contrast to "i7"). You specifically don't like that. If you mark them within the existing series, that's what Intel and AMD do with the U series. You specifically don't like that either. If you move them into a single series, you end up with something like the Intel Ice Lake/Tiger Lake naming convention, where there is some part of the name that means "cores" and some part of the name that means "power" and part that means "graphics". Other people really didn't like that, because now you have one name that means 5 different things.
(And this is what I mean about the naming discussion being dumb and boring - whatever you think is how it should be done, someone else hates that, and thinks it is too complex and requires too much knowledge on the part of buyers. It's bikeshedding, product naming is low-stakes so everyone has an opinion on it and is very upset that AMD and Intel are ignoring their urgent forums posts. At the end of the day it's just not that interesting, nor are any of these naming schemes that difficult if you bother to learn what they mean.)
Anyway, it's unfortunate that there are features and distinctions which laymen may not understand, but that's a fact of life, there's things car people really care about that a Camry Buyer doesn't know, and that's fine.
Someone who bought an expensive truck with a base-model trim might be upset that a Camry with a top trim is "nicer", because they didn't understand what a "trim" is before they laid down their money, and that's unfortunate but it's not exactly hidden either, nor should we call to get rid of trim levels because one person didn't understand. Someone else might be really upset that their SUV doesn't tow like a truck even though they got the nicest trim level on the SUV.
Again, sorry if this seems frustrated, but this is a topic that has been bikeshedded endlessly. The stakes are low, there's multiple reasonable options available, and there's a whole lot of people who are all really upset that AMD and Intel aren't taking their forums posts on the topic seriously. Naming conventions are fine, they're good enough to not matter.
Intel and AMD use a similar naming scheme, but get a little confusing with suffixes, especially AMD.
Both include a target market, generation, and performance level. With Intel, it's "i{market}-{generation}{performance}", with market being 3 (budget), 5 (mainstream), 7 (enthusiast), 9 (high-end). The numbers used in the "performance" level varies, but higher is always better. For example, i7-3770 is the third-gen chip targeted towards enthusiasts and was the highest performer in its generation. My i9-9900 is the top end model for the 9th gen. Intel will also use suffixes "K" to mean it has an unlocked multiplier (making overclocking easier) and "F" which means it does not have an on-die GPU, so you'll need a discrete GPU card.
AMD is similar, they'll call them "Ryzen {market} {generation}{performance}", ie, Ryzen 5 5600. But where AMD goes crazy is with the damn suffixes that express an additional performance level that is impossible to decipher.
A Ryzen 2700 is a second-generation CPU with 8c/16t.
A 3700 is a third-generation CPU with 8c/16t
A 4700g is a second-generation CPU with 8c/16t and a GPU
A 4500h is a second-gen CPU with 6c/12t and a GPU
There is no 4600G.
An X suffix generally means higher clockrate; an H suffix generally means high efficiency (lower power draw); a U suffix means ultra-efficient (very low power draw).
In the 5000 and 6000 series, possible suffices include X, H, U, HX, HS, but not G.
Some CPUs are OEM-only.
Some CPUs are only sold in packaging for laptop/tinybox manufacturers.
In general, if you get integrated graphics, you lose an entire processor generation, but more recently you only lose top-end features.
for a while, a lot of that could be summarized with "APUs are 1000 higher than their desktop generation" (4000 series APUs = Zen 2 = 3000 series desktop) but then AMD went and made the 5000 laptop series split between generations, a 5700U is a Zen2 part and a 5800U is a Zen3 part.
Intel did split the 10th gen but even then they changed up the naming between the two series.
Heh, I think the best we can do is embarrass them and complain when they do something stupid and confusing like some of the Ryzen 5XXX series being Zen 2 and others are Zen 3 cores. Not that Nvidia and Intel hasn't done the same thing.
Using http://ark.intel.com has been very useful for figuring out the details of Intel CPUs. Not found anything as useful for AMD or Nvidia.
