Considering the clocks and naming, the 5700X3D sounds like a 65W part. It will be interesting to see how it fares in games.

The Ryzen 5000GT was mentioned in passing, but I couldn’t find nothing about it. Has AMD ever used GT as a suffix before?

Lots of good responses regarding why 128-bit isn’t a thing, but I’d like to talk about something else.

Extrapolating from two data points is a folly. It simply can’t work. You can’t take two events, calculate the time between them, and then assume that another event will happen after the same amount of time.

Besides, your points are wrong. (Edit: That also has been mentioned in another response.)

x86 (8086) came out in 1978 as a 16-bit CPU. 32-bit came with the 386 in 1985. x64, although described in 1999, was released in 2003.

So now you have three data points: 1978 for 16-bit, 1985 for 32-bit and 2003 for 64-bit. Differences are 7 years and 18 years.

Not that extrapolating from 3 points is good practice, but at least it’s more meaningful. You could, for example, conclude that it took about 2.5 times more to move from 32-bit to 64-bit than it did from 16-bit to 32-bit. Multiply 18 years by 2.5 and you get 45 years. So the move from 64-bit to 128-bit would be expected in 2003+45 = 2048.

This is nonsense, of course, but at least it’s a calculation backed by some data (which is still rather meaningless data).

First of all, it’s worth noting that FSR has several versions, namely 1,2 and 3. FSR 1 (what the Steam Deck, RSR and Hellblade offer) simply upscales. FSR 2 gets motion data, and extrapolates from multiple frames, so has significantly higher quality. FSR 3 adds frame interpolation.

When it comes to game vs. drivers when using FSR 1, then main difference is that the game will render the UI natively over an upscaled image, while the drivers will upscale the entire frame including UI. Therefore the UI in the game implementation will look better.

So it’s better to enable FSR in the game rather than the driver, and fall back to the driver only if the game doesn’t implement FSR. Most games implement FSR 2, which is significantly better than what the driver can do, but even with FSR 1, the UI difference will make the game implementation preferable.

4 more cores would take about 20 mm^(2) and 12 more CUs would take 28 mm^(2). So that would reach 210 mm^(2) (ignoring ROPs and the like, which might take some more space). However, space is cut in some other places. For example the Series S doesn’t have AV1 decoding while the Steam Deck does have it. Also, as mentioned, memory controllers will have a different size.

They could, but N5/N4 is significantly more costly than N7/N6 (was estimated to be 1.7x more costly, IIRC). That’s why AMD is trying to keep the size under control by using chiplets.

Upping the CUs from 96 to 128 (and the ROPs similarly) will increase the GCD size from ~305 to ~372 mm^(2), based on the die image (and leaving some blank space at the side), and the total to 596 mm^(2). Whether performance will increase performance enough depends on the RAM bottleneck.

It’s also worth nothing that RDNA 3 apparently didn’t reach the expected clocks, and if AMD managed to solve this problem it would be possible to get extra performance without much or any extra die space.

In general if you’re just aiming to reduce chip size by removing two MCDs, that’s not really that much of a cost saving. That won’t make the chip a mid-range chip as rumoured.

Now, for price-sensitive products (such as the Steam Deck, or the other game consoles), APUs seem to be the way to go. You can even make powerful ones, as long as they have enough bandwidth. It’d seem to me that it’d be clear that APUs provide a much better bang for the buck for manufacturers and consumers

It’s actually the other way round.

APUs make sense in a power-constrained product, not a price-sensitive one.

The Steam Deck is a good example. It has a pretty weak CPU/GPU combo (4 Zen 2 cores and 8 RDNA 2 CUs), but this doesn’t matter, because what matters is being able to run games on battery for a decent amount of time.

When everything is on one chip, power requirements are lower, because there’s no need for inter-chip communication. Space is saved because there’s only one chip to use. This is great for small mobile products.

What about price?

APUs sell for cheap on the desktop because their performance is lower than other CPUs, but they aren’t cheap to make.

For example, Raven Ridge was 210 mm^(2), while Summit Ridge / Pinnacle Ridge were 213 mm^(2). So the chip price was about the same, but the Ryzen 1800X debuted at $500 and then dropped to $330, where the 2700X also sold, but the top of the range Raven Ridge, the 2400G, was sold for $170.

So even though it cost AMD the same to make these chips, Raven Ridge was sold for half the price (or a third for the 1800X). AMD therefore made a lot less money on each Raven Ridge chip.

The console prices are deceptive. Microsoft and Sony subsidise the consoles because they make money on game sales and subscriptions. The consoles are often sold at a loss. If the consoles were not subsidised, they’d have sold for double the price, and would have likely lost to a similarly priced PC.

Flexibility

Even though laptops aren’t user-expandable, they are still configurable. When it comes to gaming laptops, there are a lot of CPU/GPU combinations. It’s impossible to create a special chip for each combination, and binning is hardly enough to create them.

Without having separate CPUs and GPUs, you’d get an ecosystem similar to Apple’s, or the consoles, where there is a very small number of models available for purchase. That would kill one of the benefits of the Windows ecosystem, the ability to make your purchase fit performance and price targets.

A silver lining

Chiplets do make it possible to pair different CPUs and GPUs on the same chip, even together with RAM. You could call that an APU.

Thanks for the effort. I’d rewrite the list like this:

Mobile SKU 7TFA#:

FA (family):

• FA=20: Mendocino (Zen2, RDNA2)
• FA=30: Barcelo-R (Zen3, Vega)
• FA=35: Rembrandt-R (Zen3, RDNA2)
• FA=40: Phoenix (and 7545U) (Zen4, RDNA3)
• FA=45: Dragon Range (except 7545U) (Zen4, RDNA2)

T (tier):

• T=3: Ryzen 3 (4 core)
• T=4: Ryzen 3 (4 core Phoenix)
• T=5: Ryzen 5 (4 core Mendocino, 6 cores otherwise)
• T=6: Ryzen 5 (6 core Phoenix or Dragon Range)
• T=7: Ryzen 7 (8 core)
• T=8: Ryzen 7 (8 core Phoenix, 12 core Dragon Range)
• T=9: Ryzen 9 (8 core Phoenix, 16 core Dragon Range)

I’d say that it’s true that the custom work done by AMD for Microsoft and Sony is more extensive than for Valve, but Valve did get a custom APU and a 6nm refresh of it. That’s something that no other companies except for Sony and Microsoft are getting.

So basically you have a project to write, and you’re asking us to do it for you.

On the question itself, I’d say that attempts to classify things as black and white (or, in this case, red, green and yellow) are simplifications and therefore pointless.

I’d say that it depends on whether you’re making money from it and how much you will use it.

If you’ll be making any money from it, and having a more powerful card will improve your output and therefore make you more money, then that’s the way you should go.

If you’re not making any money from it, but you’re using it a lot and having a more powerful card will allow you to be more productive then it’s a reasonable way to go, too.

Otherwise, if you’re not making money from it and you’re basically using it in the background for hobbyist things, then it might not be worth the extra money.

I’d have said that if you have the money then the 4090 may be worth it, but since you’re taking about 9 installments, it sounds like it will be stretching your funds.

Interesting. My speculation previously was that AMD would use Samsung to produce lower end GPUs and APUs, freeing TSMC for higher performance parts. Not sure what these reported Zen 5c parts will be.

In any case, I think it makes sense for AMD to diversify.

It would have been a good idea to read the article carefully. :)

Still, to save you the time, these are the real dual core ones, which are the ones compatible with Windows 11, and they’re getting a new box to ensure that people don’t by the other version by mistake.

I think it highly depends on what patents NVIDIA holds. Patents hold technology back, and one of the major obstacles for companies is finding ways around the patents other companies hold.

What you already see: conversation bots, art generation and manipulation…

But I think most of that power will go towards language models, and in a few years it will be standard to talk to computers using natural language. However, the other functionality will be included in this, like the computer being able to illustrate what you tell it, teach you things, create works of art (songs, pictures, videos) for you, … But for a start it will be mainly talk.

Would have been nice to see a newer entry level CPU than the 3000G.

Looking at that PCI list, I’m actually more curious about what Garfield, Arlene and Pooky refer to.

Anyway, since it’s not clear who’s updating this list and where these came from, I’m not sure that Phoenix3 and Phoenix4 actually mean anything.

Kind of. The Zen 4c cores aren’t really “LITTLE”, in the sense that they provide the same performance at the same clock speed, and can only reach lower clocks. When all cores are running together clocks drop anyway, so it’s not a big difference from having only Zen 4 cores.

It’s not completely the same, and for high power desktop chips there will likely be a difference, but it’s still nowhere near what Intel is doing with its P and E cores, and what ARM is doing in mobile.

In terms of tech, the 4300G and 5300G were the equivalents. The 3000G was 2 cores in the age where the mainstream APUs were 4 cores, but now they’re 8 cores, so the entry level being 4 cores makes sense.

AMD does have Mendocino for mobile (weak 4 core chip), but it can’t run on the desktop (it only supports LPDDR5).

I’d like to point out that the number of cores by itself is rarely the issue. Rather it’s how old the cores are. Two more modern cores could outperform four older cores.

I have an old Thinkpad with an AMD E-350 CPU, and I consider that not very usable (could be used in a pinch, but it’s certainly not fun to use). My Dell XPS 13 with a i5-6200U is still usable.

Regarding RAM, 8GB is still perfectly usable for light use.