- Mar 3, 2017
- 1,747
- 6,598
- 136
I know this has already been discussed to death, but i wanted to add a bit of reasoning what irks me about the naming-scheme, and how how this could be improved upon, even when model year is kept as the first number.
The main issue I have has to do how humans tend to read jumbled text (like model numbers) between 5-8 characters.
For short numbers, 2-3 characters long, you grasp all of it (e.g. BMW 754) , but if the word is 5+ characters, the most common subconscious thing to do is to gloss over some of the lesser digits. In most model names the first two or three numbers are important and usually also the last one, or the suffix (14700K, 7800X ...). For AMDs laptop-processors the third digit has usually been the least relevant one in Zen 2 and prior generations, only marking minor clock-speed / GPU changes within the same generation and braket (e.g. 46x0 all being Zen 2 Ryzen 5, 6-cores).
Starting from the 7th gen it suddenly became the most important digit of the bunch.
So how to fix it?
- One solution is to add a letter suffix for "Light" or "Castrated" parts e.g. 9640UL or 9630UL vs 9650U.
- This could also wor the other way, say 8850UX announcing a major upgrade within a model year (vs a rather benign 8840U -> 8850U).
Being the last letter it's easy to grasp, without even looking at the rest of the model number like the H, K, X or X3D suffixes
(could you imagine if X3D would instead be hidden in the model name, say 7850X? What a mess that would be to explain to the illilterate)
I'm sure there are more solutions like this, to keep everyone (mostly) happy.
TL;DR:
So if anyone from AMD marketing reads it, please name the Zen 5 parts 8050UX (or similar) this gen. It would still adhere to your naming scheme and be much more friendlier to customers.
Dense is always the exact same RTL.
That means, yes, 2*512b FMA and 2*512b FADD in dense.
I get what you are saying, but I actually faced this issue (when the naming scheme was relatively new) as laptops with the 7520U were available right next to a near-identical last-gen laptops with the 5625U, priced very closely.in reality something like a 7520 is in a totally different price bracket and type of laptop than a 7540 so the chances that you are cross shopping products with those two CPUs to me seems pretty remote (outside of fire sales perhaps but here bigger number = better).
Dense is always the exact same RTL.
That means, yes, 2*512b FMA and 2*512b FADD in dense.
I wondered about that too, ever since the 512 bit FP datapath option rumor came up. However, there will apparently be a third core design (Zen 5LP), and this one won't have much use for 512b vector FP for its purposes. So that could be what the "option" bit referred to. Perhaps.Didn't AMD tell that 512-bit FPU is optional in Zen5 designs? Using 512bit FPU pipelines and load/store engine and trying to optimize that design to max density is little bit retard way to optimize designs?
I find It surprising that every single model has an Nvidia dGPU and even that limited to 4070 8GB at best. Strix Shouldn't have a problem even with RTX 4090.
View attachment 98468
Lisa - "We're seeing faster ramp up and more designs than Hawk in the same timeframe"
@adroc_thurston "They wouldn’t be getting so many design wins if the product wasn't good."
That's a very strong line-up. Even managed to get some ASUS ProArt wins (Usually Intel dominated). Lenovo, HP, Dell and Acer will probably follow with huge portfolio of design wins for Strix. Computex will be nice.
It's also where Intel hands out their biggest discounts. Operating at near-zero margins is killing them in DCG, and that can only go on for so long. Eventually the box seat tickets won't be enough. Genoa has already made things very uncomfortable, and Turin is only going to make it worse.Enterprise market is mired in corruption. The steady stream of box seats tickets for the local sports team probably has higher weight in brand decision than any technical parameter of the product. Intel thrives in this environment.
In order to offer an increased variety of AMD equipped laptop designs, OEMs first need to develop more designs. Who is there to do this work?I find It surprising that every single model has an Nvidia dGPU and even that limited to 4070 8GB at best. Strix Shouldn't have a problem even with RTX 4090.
There is no RDNA3 option, but also no model without the dGPU. I thought at least those ProArt will be without a dGPU.
BTW, I am not sure pairing It with RTX 4050 makes a lot of sense, It should be ~2x as fast at best compared to the IGP and It will have only 6GB Vram.
I get what you are saying, but I actually faced this issue (when the naming scheme was relatively new) as laptops with the 7520U were available right next to a near-identical last-gen laptops with the 5625U, priced very closely.
The "higher is better" number immediately caught my eye (and got me briefly confused for a few secs as the naming-scheme was new).
I'm pretty sure if I was a student on a strict budget (say 500€) not caring about computers and somebody told me "Lenovo is pretty decent" I'd pick something "new" like this V15 , over "some 2 gen old crap" like this V15 (particularily as the 16GB ram 512GB SSD versions with Ryzen 5625U tend to cost 50-80€ over tha "latest and greatest" 7520 version).
My cousin had a slightly higher budget, but it was in the ballpark, and he also games e-sport titles, and yes laptops with Mendocino popped up in the price bracket then. Would have been a terrible choice for her vs a couple of years older zen 3 midrange CPU.
To those who point out that AMD's part numbering scheme is confusing, to some degree, and suggest how AMD could do "better":
Ask yourself the question, what if it isn't a blunder but careful, purposeful design?
So, desktop Zen 5 is 9000, but laptop Zen 5 is 8000? Eh, AMD manages to screw the naming scheme a little bit for everyone (OEMs, Casuals, Nerds), as a little treat. After 9000, do they switch to 3 digit codes like intel? I know, they should just start from 1000 again, and keep naming it Ryzen. To really put as much mental load on anyone who tries to explain ryzen cpus.
They will progress from arabic numerals to unicode characters from the 1F600…1F64F block.After 9000, do they switch to 3 digit codes like intel?
It is less informative.
@adroc_thurston apparently no, their naming scheme wasn't kept.
And at a glance, this is pretty much what I was asking for in terms of naming.
It even goes beyond into stuff I didn't dare ask for but wanted, like the power draw level letters to be first not last.
Interesting.
"AI gen" 1, tier 7, revision 0.It is less informative.
Sure replace R9 with AI9 if you really want to tout the AI stuff and sure put the HX at the front before the numbers but 170 means nothing.
If the scheme is consistent and a generation gets a new number, then it's fine. For now, no info on whether the first number is Year again, or anything else. But in any case, it's a lot better than calling it "8050" and somehow having a guy understand that his laptop has Zen 4, Zen 4 refresh, or Zen 5 in it.Give it a few gens and you won't have a clue what generation of Zen is inside the chip by the naming convention which is exactly what the current naming scheme gives us.
"AI gen" 1, tier 7, revision 0.
There, all the info in one clean read.
It's literally what I was writing should be done yesterday, vindication feels nice.
If the scheme is consistent and a generation gets a new number, then it's fine. For now, no info on whether the first number is Year again, or anything else. But in any case, it's a lot better than calling it "8050" and somehow having a guy understand that his laptop has Zen 4, Zen 4 refresh, or Zen 5 in it.
I was thinking the same thing, but I was told that no, Zen 5 will be better than anything before including at low power. We'll see about that.Different Zens work best for different use cases. If Zen 5 takes up more power to feed those bigger cores then in a thin and light you might be better off with something older in that specific TDP envelope.
Please do not compare a properly customised Zen 2/RDNA 2 with a generic Z1/PHX implementation done by lazy OEMs that put absolutely no effort in lowering the chip's power draw, and are also using RDNA 3 which is known to be a problem child with power.Look at the Steam Deck vs the Rog Ally or Legion GO. 4c8t Zen 2 + 8CU RDNA 2 is roughly on par at 15W with 8c16t Zen 4 + 12CU RDNA 3.
I will not repeat myself. I've been lengthy enough yesterday.8050 makes it easy to know what Zen is in it. The 3rd digit explicitly tells you, if there is a Zen x+ then it will often be xx35 to distinguish it. If you are confused about being explicitly told something then that is a you issue not a naming issue. As for the rando who walks into a Bestbuy and grabs a laptop, all the names are gobbledygook to them, at best they may have heard of Intel but even then not always so will buy whatever the sales rep suggests.