Congratulations to AMD for the horrible naming so consumers can get tricked into buying a rebadged ryzen 4000 cpu as ryzen 7000. Also, why isn't there a flagship U series part with RDNA3 GPU on 4nm? Who cares about the IGPU on H series parts which will probably end up in a laptop with a dedicated GPU?
Probably because they want to charge as much as possible for that node plus the additional AI tech. The nearest you'll get is the Ryzen 7 7736U which, uniquely in the range, breaks their new naming system. It's actually the much sought after, rarely spotted R7 6800U released in Apr 2022 but with a new moniker. Glad to see this sorely under-supplied ultra premium CPU might finally get the distribution it deserves.
I agree. Perhaps the saving-grace could be from the likes of the r7-7840hs potentially underclocking from it's 35W TDP, all the way down to 15W TDP. I'd like to see something like this in the Valve Steam Deck.
It seems perfectly fine to me and possibly the best they could do while still remaining honest.
The top chips get the latest architecture. The devices that will have dGPUs for certain get slightly older iGPUs because they don't need more as much as the premium thin and light do.
The cheaper and lower performance the part, the older the architecture. Seems fair to me when the lowest is Zen 2, and that runs very well.
Wow I couldn't disagree more with everything you said.
First of all, actually the naming is super simple. Look to the 3rd digit, if it is a 4, it is the current core. Way easier than anything Intel has for example for naming. Way better than the fake RTX 4000 laptop GPUs from NVidia.
Secondly nobody needs the U GPU. You can just lower the wattage on the HS model. Everyone wanted a 30W+ model. It allows the iGPU to work properly. See the Steam Deck.
All I can say is thank god there are no more crap U series CPUs. HS is much better.
Intel's U series are an abomination. 2 P core parts with half the performance of the hated quad core parts. Intel went from 6 to 4 to 2 for their U series. Horrible.
AMD is giving their CPU the correct amount of power to hit the efficiency curve in the right spot.
Hmm. Most average users (non-gaming, no video editing or Blender, 3D etc) don't need a 40W TDP laptop but they do need longer battery life and quieter fans/less heat etc. Hence an efficient 28W CPU is ideal for them, especially in a thin and light chassis. Granted 15W is probably quite limp but even so, probably not that bad for light users.
Irrelevant. So Mac users can wave their RDF generators for some benchmarks. Utterly irrelevant as they are not cross compatible so not in the same market in most cases.
Irrelevant to who? Clearly not to the rest of the market as Macs have been gaining record marketshare while PC has been declining, and as it gets more popular more devs will be incentivized to port their applications to the Mac.
Its not possible for apple to get any substantial market share. Apple's market share rose from 6.5% 2018 to 8.5% 2021. Why? Because they have limited wafers that can supply the market. Intel & amd can mass produce cheap i5s,r5 which are perfectly fine for everyday use -> people will gravitate towards perf/value.
If people gravitate towards perf/value, then they will buy Apple Silicon. This is precisely what is happening because Macs are growing while AMD/Intel are shrinking in laptops.
The AMD Phoenix mobile CPU (APU, whatever) is very exciting, indeed. But ultimately I'll still consider it vapor-ware until AMD can prove they can create these in sufficient numbers. I'm a frequent peruser of my local Best Buy, and over the last 4 years, the amount of Windows laptops with Intel chips vs AMD has always been around ~50:8 ratio, Intel to AMD. And most, if not all, of the AMD offerings are always previous-gen models. I rarely see a top of the line AMD mobile offering.
So while the Phoenix APU could very well be the pinnacle of mobile performance/efficiency, it means absolutely bupkis if it doesn't exist in meaningful numbers.
Couldn't agree more. It's hugely frustrating. I can only hope this new strategy of mixed nodes and generations will help alleviate that problem over the next couple of years. If AMD can't supply their CPUs to the OEMs, we can't buy them! Pretty simple.
Thanks Ryan. I know things are challenging at AnandTech but you guys still cover these big flagship announcements as well as anyone.
My only reaction to the hardware, aside from the bad naming scheme, is that AMD seems to think their HS chips will be placed in laptops without dedicated GPUs, hence the silicon budget spent on graphics. That has not generally been my experience.
you are over thinking it, the HS series is just the new U series
most U chasis can fit 35W anyways, Intel may call it a U series but that is not the max power draw because of Turbo, it is just the base power that is lower
It makes no sense that the Ryzen 7 7730U has 12 CUs, as the Vega iGPUs never had that much. I suspect it's a copy and paste error (on AMD's side) from the newer architectures.
There's little bit of confusion, in the fourth slide it says there is 32 megabyte of L3 cache in 7040HS and 8 CU Vega graphics in 7030U while the article says that there is 16 megabyte of L3 cache in 7040HS and 12 CU vega graphics in 7030U. What's true??
Those are, for the moment, the official specs as posted by AMD. There have been errors on both the website and their slides (inconsistently so). The Likely Most Correct answer is 32MB of L3 on the 7040HS, and 8/6/4 CUs of Vega graphics on the 7030U.
And just to update myself here, AMD has finally sorted out part of their Ryzen specs snafu this morning. The press info decks have all been updated to consistently indicate that Phoenix has 16MB of L3 cache, not 32MB. So no L3 cache increase for mobile this gen, sorry
Ok, so correct me if I´m wrong but are these PRO versions a move backwards? We have the Ryzen 7 & 5 PRO using Zen 3+ with RDNA 2 graphics and now these Ryzen PRO are ZEN 3? and Vega graphics?
Apple's neural processors already "Just Work" with some major AI workloads like Stable Diffusion text-to-image. They can already keep up with the (regular) M1 GPU using a fraction of the power, and rival some lower end Nvidia GPUs.
If AMD's AI cores really are that fast, that is extremely interesting. This give lower-end laptops the AI content creation power of a big, power hungry Nvidia GPU, not only because of speed but because they have access to a larger RAM pool than a low end Nvidia GPU.
Yes, never came to market where I live at all. And in the neighboring markets never in any form factor a normal consumer might want. 6600U/6800U completely failed to show up, so even OEMs which had stellar AMD lines switched to Intel (e.g. Envy X360 13). I just changed the battery, so hopefully in 2 years time I can get an AMD 15-28W RDNA3 SOC.
> If you only take away two things from AMD’s new product naming system, the important things are that the leading digit is the model year, and the third digit is the CPU architecture (e.g. Zen 4). The rest, while still important, are there to define segments and TDPs, rather than the hardware itself.
See but that’s not true at all. The 7x5x and the 7x4x are completely different hardware! The former is the desktop-class SOCs with RDNA2 on 5nm and the latter is the new mobile SOCs with Ryzen AI, RDNA 3, and on 4nm.
No matter what they say, this is super confusing for consumers.
That being said, those new Phoenix 7x4x SOCs look incredddiibllle. Can’t wait to see perf of the 780m with 12 RDNA 3 CUs running with LPDDR5x 7400…
God, it is so easy. 1st digit is year. 7. Don't even have to think about it, they are all 7s. You only have to look at the next 3. 2nd digit is the class. Slow, medium, fast. Ryzen 3, 5, 7. 3rd digit is the core tech. Look for a 4 for Zen4. 4th digit is for speed. The 5 models are faster than the 0 models. It isn't hard.
Why bother with the 1st digit at all? What useful info does a year convey? It's the 3rd digit that matters most for almost everybody, and this one you have to be in the know what does it actually mean (fab process, GPU gen..).
Till today I still don't understand whay there is a need for such confusing and dumb naming scheme... The only reason I can think of is to confuse the buyer thinking that he/she is getting the latest stuff because its 7000 series (compare with desktop counterpart).
Intel on the other hand is way more honest in the naming.... 12xx meanings its definitely alderlake and 13xx is raptorlake.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
46 Comments
Back to Article
isthisavailable - Wednesday, January 4, 2023 - link
Congratulations to AMD for the horrible naming so consumers can get tricked into buying a rebadged ryzen 4000 cpu as ryzen 7000.Also, why isn't there a flagship U series part with RDNA3 GPU on 4nm? Who cares about the IGPU on H series parts which will probably end up in a laptop with a dedicated GPU?
Ryan Smith - Wednesday, January 4, 2023 - link
"Also, why isn't there a flagship U series part with RDNA3 GPU on 4nm?"Most likely explanation: limited initial silicon, and H parts are more profitable.
AHA - Saturday, January 7, 2023 - link
Probably because they want to charge as much as possible for that node plus the additional AI tech. The nearest you'll get is the Ryzen 7 7736U which, uniquely in the range, breaks their new naming system. It's actually the much sought after, rarely spotted R7 6800U released in Apr 2022 but with a new moniker. Glad to see this sorely under-supplied ultra premium CPU might finally get the distribution it deserves.ballsystemlord - Thursday, January 5, 2023 - link
It is worth noting is that Intel has done this before. I know because I owned such a CPU in my laptop.Additionally, tech reviewers, IIRC like MLID, have often suggested using an older node for the lower end parts.
Kangal - Friday, January 6, 2023 - link
I agree.Perhaps the saving-grace could be from the likes of the r7-7840hs potentially underclocking from it's 35W TDP, all the way down to 15W TDP. I'd like to see something like this in the Valve Steam Deck.
Tams80 - Saturday, January 7, 2023 - link
It seems perfectly fine to me and possibly the best they could do while still remaining honest.The top chips get the latest architecture. The devices that will have dGPUs for certain get slightly older iGPUs because they don't need more as much as the premium thin and light do.
The cheaper and lower performance the part, the older the architecture. Seems fair to me when the lowest is Zen 2, and that runs very well.
Alistair - Saturday, January 7, 2023 - link
Wow I couldn't disagree more with everything you said.First of all, actually the naming is super simple. Look to the 3rd digit, if it is a 4, it is the current core. Way easier than anything Intel has for example for naming. Way better than the fake RTX 4000 laptop GPUs from NVidia.
Secondly nobody needs the U GPU. You can just lower the wattage on the HS model. Everyone wanted a 30W+ model. It allows the iGPU to work properly. See the Steam Deck.
All I can say is thank god there are no more crap U series CPUs. HS is much better.
Intel's U series are an abomination. 2 P core parts with half the performance of the hated quad core parts. Intel went from 6 to 4 to 2 for their U series. Horrible.
AMD is giving their CPU the correct amount of power to hit the efficiency curve in the right spot.
AHA - Saturday, January 7, 2023 - link
Hmm. Most average users (non-gaming, no video editing or Blender, 3D etc) don't need a 40W TDP laptop but they do need longer battery life and quieter fans/less heat etc. Hence an efficient 28W CPU is ideal for them, especially in a thin and light chassis. Granted 15W is probably quite limp but even so, probably not that bad for light users.Tams80 - Saturday, January 7, 2023 - link
They also don't need the latest architecture.lemurbutton - Wednesday, January 4, 2023 - link
Will Anandtech finally pit AMD's mobile chips against the M2, Pro, Max in a review?Apple Silicon has been the most exciting event to happen in chips in a long time but we get so little coverage of it here.
dontlistentome - Thursday, January 5, 2023 - link
Irrelevant. So Mac users can wave their RDF generators for some benchmarks. Utterly irrelevant as they are not cross compatible so not in the same market in most cases.caribbeanblue - Thursday, January 5, 2023 - link
Irrelevant to who? Clearly not to the rest of the market as Macs have been gaining record marketshare while PC has been declining, and as it gets more popular more devs will be incentivized to port their applications to the Mac.VioletKim1998 - Thursday, January 5, 2023 - link
Its not possible for apple to get any substantial market share. Apple's market share rose from 6.5% 2018 to 8.5% 2021. Why? Because they have limited wafers that can supply the market. Intel & amd can mass produce cheap i5s,r5 which are perfectly fine for everyday use -> people will gravitate towards perf/value.lemurbutton - Thursday, January 5, 2023 - link
If people gravitate towards perf/value, then they will buy Apple Silicon. This is precisely what is happening because Macs are growing while AMD/Intel are shrinking in laptops.VioletKim1998 - Friday, January 6, 2023 - link
Shrinking in laptops? by what 1.7%? No these cheap acer, hp @ $500 are better perf/value since they're mostly used for casual work.Alistair - Saturday, January 7, 2023 - link
"if people gravitate towards value, they will buy Apple"LMAO, can't even, can't even respond to that
Dante Verizon - Thursday, January 5, 2023 - link
Real PC runs windows. period.Farfolomew - Thursday, January 5, 2023 - link
The AMD Phoenix mobile CPU (APU, whatever) is very exciting, indeed. But ultimately I'll still consider it vapor-ware until AMD can prove they can create these in sufficient numbers. I'm a frequent peruser of my local Best Buy, and over the last 4 years, the amount of Windows laptops with Intel chips vs AMD has always been around ~50:8 ratio, Intel to AMD. And most, if not all, of the AMD offerings are always previous-gen models. I rarely see a top of the line AMD mobile offering.So while the Phoenix APU could very well be the pinnacle of mobile performance/efficiency, it means absolutely bupkis if it doesn't exist in meaningful numbers.
lemurbutton - Thursday, January 5, 2023 - link
Phoenix APUs cannot be the pinnacle of mobile performance/efficiency because Apple Silicon exist.Wereweeb - Thursday, January 5, 2023 - link
Sorry, I'll fix it for them: "Phoenix is the pinnacle of mobile performance/efficiency for people who have fully developed higher brain function"nandnandnand - Thursday, January 5, 2023 - link
Maybe that's why they didn't announce Phoenix-U APUs. So nobody gets their hopes up for RDNA3 gaming handhelds until much later in the year.Alistair - Saturday, January 7, 2023 - link
agree we need volume!!!AHA - Saturday, January 7, 2023 - link
Couldn't agree more. It's hugely frustrating. I can only hope this new strategy of mixed nodes and generations will help alleviate that problem over the next couple of years. If AMD can't supply their CPUs to the OEMs, we can't buy them! Pretty simple.Sunrise089 - Thursday, January 5, 2023 - link
Thanks Ryan. I know things are challenging at AnandTech but you guys still cover these big flagship announcements as well as anyone.My only reaction to the hardware, aside from the bad naming scheme, is that AMD seems to think their HS chips will be placed in laptops without dedicated GPUs, hence the silicon budget spent on graphics. That has not generally been my experience.
brucethemoose - Thursday, January 5, 2023 - link
IGP performance is still important, as the dGPUs are basically unusable when trying to save battery life.Alistair - Saturday, January 7, 2023 - link
you are over thinking it, the HS series is just the new U seriesmost U chasis can fit 35W anyways, Intel may call it a U series but that is not the max power draw because of Turbo, it is just the base power that is lower
ET - Thursday, January 5, 2023 - link
It makes no sense that the Ryzen 7 7730U has 12 CUs, as the Vega iGPUs never had that much. I suspect it's a copy and paste error (on AMD's side) from the newer architectures.Ryan Smith - Thursday, January 5, 2023 - link
Quite likely. There have been a few errors on the specs posted this evening.Alistair - Saturday, January 7, 2023 - link
it was just a typo, it says 8 on the AMD website, the specs are on there nowThefuturistic - Thursday, January 5, 2023 - link
There's little bit of confusion, in the fourth slide it says there is 32 megabyte of L3 cache in 7040HS and 8 CU Vega graphics in 7030U while the article says that there is 16 megabyte of L3 cache in 7040HS and 12 CU vega graphics in 7030U. What's true??Ryan Smith - Thursday, January 5, 2023 - link
Those are, for the moment, the official specs as posted by AMD. There have been errors on both the website and their slides (inconsistently so). The Likely Most Correct answer is 32MB of L3 on the 7040HS, and 8/6/4 CUs of Vega graphics on the 7030U.Ryan Smith - Thursday, January 5, 2023 - link
And just to update myself here, AMD has finally sorted out part of their Ryzen specs snafu this morning. The press info decks have all been updated to consistently indicate that Phoenix has 16MB of L3 cache, not 32MB. So no L3 cache increase for mobile this gen, sorryKhanan - Thursday, January 5, 2023 - link
AMD becomes more and more serious in the mobile space, good to see. Take away that money from intel.wenart - Thursday, January 5, 2023 - link
Ok, so correct me if I´m wrong but are these PRO versions a move backwards?We have the Ryzen 7 & 5 PRO using Zen 3+ with RDNA 2 graphics and now these Ryzen PRO are ZEN 3? and Vega graphics?
brucethemoose - Thursday, January 5, 2023 - link
Apple's neural processors already "Just Work" with some major AI workloads like Stable Diffusion text-to-image. They can already keep up with the (regular) M1 GPU using a fraction of the power, and rival some lower end Nvidia GPUs.If AMD's AI cores really are that fast, that is extremely interesting. This give lower-end laptops the AI content creation power of a big, power hungry Nvidia GPU, not only because of speed but because they have access to a larger RAM pool than a low end Nvidia GPU.
brucethemoose - Thursday, January 5, 2023 - link
Here is a link to what I am talking about: https://github.com/apple/ml-stable-diffusion#-perf...I would hope AMD is working on similarly easy PyTorch integration for this thing.
qlum - Thursday, January 5, 2023 - link
Let's hope availibility is better this time around, the 6000 series was rare and availibility only became worse over time.markiz - Monday, January 9, 2023 - link
Yes, never came to market where I live at all.And in the neighboring markets never in any form factor a normal consumer might want.
6600U/6800U completely failed to show up, so even OEMs which had stellar AMD lines switched to Intel (e.g. Envy X360 13).
I just changed the battery, so hopefully in 2 years time I can get an AMD 15-28W RDNA3 SOC.
Dante Verizon - Thursday, January 5, 2023 - link
The lack of RDNA3 iGPU benchmarks tells me that there aren't any big advances in this regard.Alistair - Saturday, January 7, 2023 - link
well they won't be out in volume until at least April, probably don't want to brag about that yetbwhitty - Friday, January 6, 2023 - link
> If you only take away two things from AMD’s new product naming system, the important things are that the leading digit is the model year, and the third digit is the CPU architecture (e.g. Zen 4). The rest, while still important, are there to define segments and TDPs, rather than the hardware itself.See but that’s not true at all. The 7x5x and the 7x4x are completely different hardware! The former is the desktop-class SOCs with RDNA2 on 5nm and the latter is the new mobile SOCs with Ryzen AI, RDNA 3, and on 4nm.
No matter what they say, this is super confusing for consumers.
That being said, those new Phoenix 7x4x SOCs look incredddiibllle. Can’t wait to see perf of the 780m with 12 RDNA 3 CUs running with LPDDR5x 7400…
bwhitty - Friday, January 6, 2023 - link
And look, I screwed up my comment: it’s the 4th digit which indicates the major change from 5nm+RDNA 2 to 4nm+RDNA 3.Decoder ring needed, indeed…
Alistair - Saturday, January 7, 2023 - link
God, it is so easy. 1st digit is year. 7. Don't even have to think about it, they are all 7s. You only have to look at the next 3. 2nd digit is the class. Slow, medium, fast. Ryzen 3, 5, 7. 3rd digit is the core tech. Look for a 4 for Zen4. 4th digit is for speed. The 5 models are faster than the 0 models. It isn't hard.markiz - Monday, January 9, 2023 - link
Why bother with the 1st digit at all? What useful info does a year convey? It's the 3rd digit that matters most for almost everybody, and this one you have to be in the know what does it actually mean (fab process, GPU gen..).Tams80 - Saturday, January 7, 2023 - link
Nah, that's just you overthinking it.See Alistair's comment. It's quite simple and as simple as they could make it while still being informative.
escksu - Wednesday, March 1, 2023 - link
Till today I still don't understand whay there is a need for such confusing and dumb naming scheme... The only reason I can think of is to confuse the buyer thinking that he/she is getting the latest stuff because its 7000 series (compare with desktop counterpart).Intel on the other hand is way more honest in the naming.... 12xx meanings its definitely alderlake and 13xx is raptorlake.