AMD claims Radeon 8060S iGPU outperforms Nvidia RTX 4070 by up to 68.1% in gaming

[opckieran]

While it is generally up to the the DEVICE OEM to decide what (if anything!) they wish to pair an APU with, this level of performance in a iGPU pretty much negates the need or desire for dGPU in anything but professional CAD/CAM applications.

Sadly the technology & performance may never trickle down into sub $700 devices. (or might be a LONG TIME coming)

TL;DR, While I was writing my initial response below, it dawned on me that no Halo laptop will ever cost $700. That would severely undermine the entirety of AMD/Intel's mobile SOC lineups. Why would a non-Halo laptop be worth more than $700 if the Halo SOCs are far and away AMD/Intel's best chips? The answer would have to be nVidia, and neither AMD nor Intel would want to create a situation where they significantly undervalue their own lineups while simultaneously creating an opportunity by dint of confession for nVidia to market themselves as the premium or must-have option (i.e "it isn't a real laptop without nVidia"). There's no way they'd ever do such a thing as to allow the consumer to perceive AMD or Intel so poorly, and nVidia so highly at the same time.



In more detail:
AMD initially said their 300 series wouldn't allow expandable RAM, but this was proven to be incorrect/bypassed in the NUC market. As for Strix Halo not allowing a dGPU, they've already got the more CPU-oriented 9955HX(3D) SOC with 16 Zen 5 classic cores and a smaller iGPU specifically geared for maximum CPU performance and the gaming market, so there's no need for them to step on their own toes here.

AMD doesn't primarily see Strix Halo as a performance achievement. They've actually been making large iGPU SOCs since the PS4/XBOne, so nearly 15 years (they were taping out well before those consoles launched). To them, this is purely an engineering showcase. And that's why they feel justified in asking for supposedly $2200+ for Strix Halo laptops (again, my money is on a $1600-1800 launch). But in reality it won't be any faster than a 5060 on a good day at the same wattage, but, because it's a SINGLE chip, customers are supposed to be impressed (whoopee). Even though battery life isn't gonna be much better. For $2200, I'd rather just get a slim 5070Ti laptop with a really good screen and maybe even save $300-500 at the same time. Unless running 20-32b parameter LLMs or other similarly large AI workloads locally is that important to you. Even then, I'd demand a minimum of 48GB RAM to do that comfortably.

IMO, this is most likely AMD's way of exiting the laptop dGPU market, as Intel already has (they saw the writing on the wall after a single generation). They're struggling enough as is on the desktop dGPU market, and are almost certainly doing significantly WORSE in laptops. I'd estimate, if they've got 8-15% market share in the desktop dGPU market, they're probably at 0.2%-0.8% for laptop dGPU. You barely see ANY laptops with AMD dGPUs in them, probably because efficiency matters a lot more on laptops AND you don't get the crucial perf/$ and/or VRAM capacity advantage that you do in the desktop space. AMD's own site, TODAY, lists a grand total of NINE RX 7000 series laptops, with more than half of those listings being for the Chinese market. It's depressing. Not even the boutiques like Eurocom, Sager, nor Alienware offer AMD dGPU laptops today. Shoot, the 7800M launched back in September, and the only product it's currently offered in is an outrageously-priced $1150  OcuLink+USB4 Dock, which only gets you Desktop 4070 or Mobile 4080 performance anyways! No price advantage! I'd rather just DIY at that point and save the money!

Point is, that's just how non-competitive AMD are in the laptop dGPU space, and have been for easily a decade now, not even exaggerating. I'd bet money that AMD are finally throwing in the towel for laptop dGPUs, if not after RX 7000S/M(XT), definitely after RX 90M. But what sucks here is that even at a supposedly "low" $1800, AMD are STILL overestimating their brand compared to nVidia. Because who on earth is actually gonna spend $1800+ for a non-Apple device and be satisfied that it didn't come with nVidia? At that price it's probably 1-3% of the high-end laptop market. A niche, most likely high-end Linux laptop enthusiasts. And unless AMD provides an actually substantial alternative to CUDA, it won't get any better for what they're trying to sell to that userbase.

And that isn't even taking into account the up and coming Lunar Lake Halo! So, AMD really has their work cut out for them. Otherwise, in the event of a price war, and Strix Halo will be dragged, by Intel, with their iGPU ironically of all things, kicking and screaming down to the more feasible $1200-1300 price range, where it will then be seen as an "interesting" alternative to the 5070M at best. I doubt $700 would happen since Krakan Point is supposed to live in the $700s and below. Plus, AMD is currently dominating x86 CPU high-end perf and perf/watt anyways so they have a perfectly justifiable reason to charge more than they did back in the day. Who knows, maybe Lunar Lake Halo could hit $1100 in such a scenario (the increased BoM for on-package RAM might make that difficult). But then  that's a stretch for a different reason: it's always been easier for Intel to charge more for less in part due to their terrible ethics shutting AMD out of higher laptop build quality.

[paulzrecs]

That 4070 is running at 60w.

[Bennyg]

"more efficient than A 4070" just doesn't do the baiting of clicks that "more efficient than THE 4070" does...

A TDP gimped 4070 is NOT representative of the vast majority of 4070s that are >=100W

With this kind of misleading garbage AMD really is becoming the new Intel

[Russel]

The comparison is more about incremental upgrade of z13 from previous generation to the current one.

In performance? Sure. But I think people will be pleasantly surprised when it comes to efficiency and battery life. When you do things all on one chip on package and don't need to communicate to external chips through longer distances coupled with GDDR memory there is often huge energy savings cost. (Which benefit heat/thermals and noise of device)

I think for people in the US, you're better off getting a last year G14. As there are deals that can be had for $1000 there, so paying this much doesn't make much sense. Can buy a cheap tablet if really need stylus support. But for people in the Europe, where most RTX 4070 devices are over 2000 anyway, this doesn't seem like a bad deal, anymore.

I don't think of this as a barely improved 2023 rog flow z13 but more of a bigger ROG Ally with 3 times the performance, more memory and storage.

The only worry I've is warranty. I don't really trust Asus enough when it comes to theirs or their reliability. Hope this device is sold at retail stores with good after sale policies. (E.g. Costco)

I wasn't trying to discredit them of any progress made in efficiency.
I was merely stating that it's rather deceptive to portray it as a victory over 4070 mobile, because a 4070 mobile in a legion 5 pro would perform much better.
So rather than saying that amd's chip comes on top, it's more correct to say that Asus has improved what it offers with the z13 by a lot.

When amd claims superiority over 4070, it'd just invite ridicule just like what we see in the comment section. 😂

[GeorgeS]

When amd claims superiority over 4070, it'd just invite ridicule just like what we see in the comment section. 😂

For those of us that have memories of the last weeks/months/years might recall the fairly recent similar sounding claim that: "the PS5-PRO can outperform a 4070"

I think we all know how that turned out. :)

Granted, AMD has made GREAT strides in their APU's and I can dream of a day where I can get a 12-14" AMD powered laptop for < $1000 and play AAA games on it @1080P without needing a dGPU.

IMHO: out of AMD, Apple, Intel, Qualcomm & Nvidia (if they ever enter the APU laptop market) I'm betting on AMD to deliver APU's gamers can play games on.