AMD manages to make the Ryzen 7040U look worse than it is

Not that good to begin with either

AMD Ryzen LogoEven by AMD’s low bar, the messaging surrounding the Ryzen 7040 series ‘launch’ was pathetic. SemiAccurate is rather stunned at how AMD took an underwhelming part and made it looks significantly worse.

Lets start this out with the elephant that won’t leave the room, security. It is our informed opinion that no AMD (and Qualcomm laptop too) consumer CPU since the Ryzen 6000 generation is safe to deploy. They have intentional remote hardware backdoors and malware that preclude security and encryption on the device. You can not secure any device with the Microsoft Pluton block even if it has an external TMP chip. Take this seriously, it is about as bad as it gets.

Ok rant over, back to the real rant. AMD launched the Ryzen 7040 chip line, something they are not proud of. Why do we say that? They showed the silicon off at CES, the author held one, and then crickets. Three delayed dates later we have the ‘launch’ of the chip. Here is where we usually say something about the briefing but, well, there wasn’t any. Literally none. AMD sent out an email less than 17 hours before the embargo lift with a link to a PDF about the new line. I didn’t see the email until 4 hours after the embargo lifted which is why we didn’t have a story up at that time.

<sarcasm>Luckily there was all the time in the world to ask questions and get replies for the new paradigm of Xilinx based AI bulls…functionality on the devices.</sarcasm> We are getting tired of this refrain but AMD obviously isn’t because when you get an email at 3:21pm, even if you saw it immediately, you have how long to go over the material, ask a question, and expect a first pass reply? If you are the type that sleeps on a somewhat regular basis, that window is <9 hours, <2 of which are ‘working’ hours. But AMD did a CYA move and said we could ask questions any time full well knowing that the embargo had already gone up.

This is a cynical and pathetic PR move to intentionally preclude questions. How does it work? They give you the choice of writing what they send and hitting the deadline or doing your job and missing it. They know the majority of sites won’t question things and will just puke back the official line and never correct things when they realize how crappy the product is. More on this later. It is just a shit thing to do and you do it when you know you have a turkey, if the product was good there would be a full briefing. There was more than enough time to do the right thing but AMD seems allergic to that path of late.

On paper the Ryzen 7040 series is supposed to open up new worlds with dedicated AI hardware, something that AMD really doesn’t want you to know about. Why? Because it is crap and unsupported. What do we mean by that? Well you would think from the hype that borders on lying through omission that it is a block like a GPU that you can send AI work to, right? Nope. It is a block of something with no disclosed features or capabilities but you can’t actually access it and despite our best efforts, SemiAccurate can’t find any documentation on it. Basically it is a closed black box at the moment, probably forever, that only specific OEMs can use.

This ‘powerful’ block is so powerful and game changing that the OEM comments on it were mind blowing. Actually scratch that, there were none. But the software shown on it was… nope, none of that either. Yup, a ‘launch’ with no claims of models coming, no software supporting it, and even the more sniveling of OEMs wouldn’t give a quote, not even Microsoft. The entirety of the information on this block, misleading though it is, is in the slide below.

AMD Ryzen 7040U AI descriptions

Can you feel the obfuscation?

Honestly, this is the most pathetic example of a briefing on a much hyped feature that SemiAccurate can remember, and I for one have been doing this job for 20+ years. Literally no examples of what it can do, only a caveat-ed example of support in Windows, sort of, if the OEM supports it on the platform. What a joke. “Amazing, upgradeable capabilities…” that are so good we can’t tell you about them. Really? Do they expect use to not only buy this crap but to shovel it back to you? Oh wait, most sites did.

Then there is the bit about the Ryzen 7040 being the ‘first processor of it’s kind’. Depending on what you mean by ‘it’s kind’, that statement may be able to pass scrutiny at AMD legal. If you are thinking it means the first Windows capable CPU that has AI hardware, err, Apple, Intel, Qualcomm, and even VIA/Centaur got there first. See why AMD was deathly allergic to real briefings and questions? See why they wanted to get the info out without real scrutiny? Now if you are thinking they meant the first consumer CPU that precludes security, well their own Ryzen 6000 beat them to the punch and Qualcomm did too. Intel is wisely staying out of that race for now.

So lets pivot and give you a bit of unvarnished tech about why AI on a consumer laptop chip. AI hardware is good for some tasks at a consumer level. Cameras are a big part of it, both on the image acquisition side and on the image processing/search side. Unfortunately no one is dumb enough to take a picture with their laptop when they have a phone handy, and do recall phone penetration is much higher than laptop penetration among literally every demographic in the world. Once the phone uploads the picture to a repository, all that categorization, processing, and search/recognition can be done better and faster on the server. Sure you CAN do it on your laptop but by the time you download it, it should have already been done ‘professionally’ in the cloud.

To make matters worse, nearly all the AI related processing tasks depend on having lots of trained information. Do you think your laptop has a better model than, oh say, Facebook or Google? AI hardware is great for acquisition devices but borderline useless on laptops. Anything that hardware would accelerate is borderline trivial to do on a modern CPU anyway, is it worth buying a new device for something you may do every few weeks for 15 seconds at a time even if the hardware AI saves you 3/4 of that 15 seconds? Would 90% faster sway your purchasing decisions?

There are other use cases like the ones Qualcomm showed off last year. Things like tweaking avatars in Teams and better background removal. The bunny ears you digitally put on your head are more realistic, a little more realistic anyway, all for the price of a new laptop. And it could be done on your CPU just as well if Microsoft put the coding effort into it which they are willing to do for a new block but not for an already purchased piece of software. Other than corner cases that don’t matter and don’t actually need AI hardware, SemiAccurate is at a loss as to why any laptop needs AI hardware. Based on the pathetic ‘briefing’, AMD is at a loss too, mainly because it is nearly pointless for consumer hardware/laptops. AI is a buzzword that investors love so it has to be there, useful or not. At the moment, literally all of AMD’s competition does it better and has broader support though.

AMD Ryzen 7040U lineup

AMD Ryzen 7040U lineup

So back to the devices themselves, there are four. The top end 7840U is the fully active part with the 7640U, 7540U, and 7440U being progressively crippled from there. All are in the 15-30W range, support DDR5/LPDDR5 of undisclosed frequencies, and are built on a TSMC 4nm process although AMD didn’t deem it fit to tell us that. Nor the die size. Or transistor count. Or give us time to ask such basic questions much less real ones like why they added such a useless block and why they called it AI when it is just an FPGA with features removed. But I digress.

AMD Ryzen 7040U block diagram

AMD Ryzen 7040 ‘details’

Here are the ‘details’ that AMD gave out, basically the upgrades you expected last year are here now. The GPU is up to 2022 standards and the memory catches up to last year too. Note the crippling in the bottom right, basically only the high end part is worth it if graphics matter, and since this chip is insecurable, it is our strong recommendation that you don’t do anything more trivial than game on it.

Other than that there were a few benchmarks, the most notable was how AMD ‘won’ against Apple’s M2 on things not AI related. Of the benchmarks ‘shown’, gaming was not among them, nor was it for the comparison against Intel’s i7-1360P. Needless to say while we might believe the cherry picked category wins against the Intel part, we don’t buy the wins against the Apple CPU for a second. AMD went out of their way to NOT show any system despite multiple opportunities to do so, to NOT allow any independent testing, or to NOT give any chance to question or back up their claimed numbers. Any guesses why?

So overall what do we have? AMD took a chip that could possibly be useful for things where security is not a question and made it look like crap. There was no reason for them to do this but of late they seem to be getting stunningly adept at digging their own hole, not stopping, and then blowing both feet off with a footgun when they realize they are stuck at the bottom.

They ‘launched’ a device without giving out the bare minimum details and doing so in a manner that actively precluded questions. This was done because they have a lot to hide. Their vaunted ‘AI’ hardware block is basically a user-inaccessible FPGA with functionality removed that is so powerful that they won’t even hint at its capabilities. It is so good they couldn’t even get one shill to stand up and claim it was useful for their wares, Microsoft included. No partners, no examples, no hardware, no details, and a lot of purposeful omissions leads to a having a lot to hide. Truly pathetic AMD.S|A

The following two tabs change content below.

Charlie Demerjian

Roving engine of chaos and snide remarks at SemiAccurate
Charlie Demerjian is the founder of Stone Arch Networking Services and SemiAccurate.com. SemiAccurate.com is a technology news site; addressing hardware design, software selection, customization, securing and maintenance, with over one million views per month. He is a technologist and analyst specializing in semiconductors, system and network architecture. As head writer of SemiAccurate.com, he regularly advises writers, analysts, and industry executives on technical matters and long lead industry trends. Charlie is also available through Guidepoint and Mosaic. FullyAccurate