• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Should AMD include a small iGPU die variant for their upcoming 4C/8T Zen APU?

Should AMD include a small iGPU die variant for their upcoming 4C/8T Zen APU?

  • Yes

  • No


Results are only viewable after voting.
Should AMD include a small iGPU die variant for their upcoming 4C/8T Zen APU?

I voted YES.

I would go further, and suggest that they consider making it also be able to carry out basic gpu compute (openCL etc) and/or IPMI2 functionality (i.e. KVM over ethernet, for servers). If sensible/practicable to do.

The existing FX chips, are probably "hurt", for people wanting to make cheap, basic/simple PCs for web browsing/office use, because they still need external (to the cpu) video cards. Which tend to be power hogs, potentially pricey and usually a lot less reliable than Igpus.
 
No, pure CPU is what we need. We dont need the APU/intel iGPUs in the High-End enthusiast class of CPUs.
 
No, pure CPU is what we need. We dont need the APU/intel iGPUs in the High-End enthusiast class of CPUs.

I'm referring to mainstream laptops (see link in post #2 ) and mainstream desktops.

1.) Performance wise we don't need large iGPUs for those products.

2.) Also consider the cheapest way to buy 8GB in the future may be as a single 8GB DDR4 3200 DIMM. (So the core count and iGPU should work well enough with that 25.6 GB/s bandwidth)
 
I'm referring to mainstream laptops (see link in post #2 ) and mainstream desktops.

1.) Performance wise we don't need large iGPUs for those products.

2.) Also consider the cheapest way to buy 8GB in the future may be as a single 8GB DDR4 3200 DIMM. (So the core count and iGPU should work well enough with that 25.6 GB/s bandwidth)

Even if you are talking about mainstream APUs, i would also say no. Either you buy an APU for its strong iGPU performance or you by a CPU.

So if you want a CPU, I would prefer the die area occupied by any iGPU to be used for more cash/extra core or to make the CPU smaller = cheaper.
Why add a small iGPU when in the same area you can put another Core and sell a 6C 12 T CPU for the same price as a 4C 8T + small iGPU ??

edit: A small iGPU will not give you enough performance to pair it with a dGPU. You better have a large fast iGPU and pair it with the same performance dGPU or you better have a pure CPU and pair it with a strong dGPU. For 14nm FF APUs, the smallest iGPU should be 384 shaders and that should be small enough for a 2C 4T 14nm APU. For 4C 8T APU 512 Shaders should be the minimum.
 
Last edited:
High end mainstream chips (large die size, lower volume chips):

1. 8C/16T without iGPU (dual channel memory)
2. 4C/8T with large iGPU (dual channel memory)

Mainstream (small die size, higher volume chips):

1. 4C/8T with small iGPU (single channel memory with DDR4 3200 or dual channel memory with slower DDR4 RAM). This processor used with and without dGPU.
 
Should AMD include a small iGPU die variant for their upcoming 4C/8T Zen APU?

You are a few years too late for asking that question.
Zen's design is most likely already complete, and they are bug fixing now along with doing early validations.
 
Why add a small iGPU when in the same area you can put another Core and sell a 6C 12 T CPU for the same price as a 4C 8T + small iGPU ??.

Remember this chip (4C + small iGPU) is for a high volume laptop/desktop.

Take away the iGPU and now its 6C/12T + mandatory dGPU.

6C/12T + mandatory dGPU is going to be in a higher price class than 4C/8T small iGPU.
 
You are a few years too late for asking that question.
Zen's design is most likely already complete, and they are bug fixing now along with doing early validations.

They can always come out with a second design if a large iGPU Zen APU comes out first.
 
They can always come out with a second design if a large iGPU Zen APU comes out first.

Ahh, so you should have said should future designs...

I am no fan of integrated GPU on the same die, that is, until they are around 20x more powerful than what we have now.
It would be much cooler if they would have had multiple sockets on the motherboard that we could have swapped out GPU cores, like the way we can upgrade CPUs.
 
A weak iGPU covers about 95% of usage cases (including your main GPU dying) and doesn't add complexity nor price. Intel figured as much, AMD might as well.
Practically the only desktop usage scenario for not including an iGPU is 3D gaming.
 
A weak iGPU covers about 95% of usage cases (including your main GPU dying) and doesn't add complexity nor price. Intel figured as much, AMD might as well.
Practically the only desktop usage scenario for not including an iGPU is 3D gaming.

+1
 
I really think they should focus on Zen CPU cores FIRST...and then link it to a new GPU once they finally made a new GPU arch.


So my answer is yes and no.

AMD already said that most of their R&D budget is going into Zen....so I'd prefer if they deliver a fat fat CPU first and give us a bandaid Carrizo/DDR4 APU ....and then give us a Zen APU later.


If the CPU ends up being mediocre again because their already tiny/ split budget was split once again...then no amount of awesome iGPU is gonna save them...especially while Intel is improving theirs with every iteration as well.


I really think AMD is gonna go down guns blazing if that CPU fails getting anywhere near at least Haswell-E. And right now they can probably afford taking a step back from Laptops, anyway....not like they ever really had a foot in there, anyway. Their APUs in those products are cute n all...but with that marketing it's not gonna sell anywhere.
 
Last edited:
An iGPU in a top Zen based model might be a question of balance: Provide some parallel compute capabilities close to the cores at high bandwidth or let the dGPU behind the PCIe do the stuff?
 
I'd say that 128 GCN cores would be really nice. not too much die area and enough power for common tasks, so people that needs good CPU performance but just basic graphics doesn't have to buy an APU or a dedicated card; it would be pretty useful for testing and fallback purposes.
 
An iGPU in a top Zen based model might be a question of balance: Provide some parallel compute capabilities close to the cores at high bandwidth or let the dGPU behind the PCIe do the stuff?

My hope is that the four Zen cores would be plenty enough compute for the average person. Then the small iGPU merely needs to serve common task graphics needs. (This is probably 98% of consumers)

Then for those people that want additional parallel compute close to the cores (for specialized tasks) AMD can offer an APU model with a larger iGPU. (This is probably 2% of consumers)

In the case of dGPU, this can be used with both small and large iGPU APUs.
 
I'd say that 128 GCN cores would be really nice. not too much die area and enough power for common tasks, so people that needs good CPU performance but just basic graphics doesn't have to buy an APU or a dedicated card; it would be pretty useful for testing and fallback purposes.

The front-end and the uncore for a 128 Shader GCN iGPU will take more space than the 128 Shaders alone. So at the end, you will dedicate a big space for a worthless 128 shader iGPU when you could install at the same space more CPU cores or larger Cache etc.
 
The front-end and the uncore for a 128 Shader GCN iGPU will take more space than the 128 Shaders alone. So at the end, you will dedicate a big space for a worthless 128 shader iGPU when you could install at the same space more CPU cores or larger Cache etc.

AMD already has a 8C/16T Zen with large cache planned.

So there is no reason to make another chip like that.
 
AMD already has a 8C/16T Zen with large cache planned.

So there is no reason to make another chip like that.

I was talking hypothetically, why dedicate a large die size area for a worthless 128 Shader iGPU ??? APUs are all about iGPU performance, smallest Kaveri APU already has 256 Shaders (A6-7400K) at 28nm, why spend money/resources for a 14nm 128 Shader iGPU ?? Not worth it.
 
Not having an iGPU, would kill the vast majority of any business AMD could get in the laptop space, except for Puma-like CPUs. There has to be an iGPU of some sort. 128 GCN SPs is a bit archaic in total capability for a higher end quad core Zen part unless it's so good that it manages to gain multiple wins in the high end/gaming laptop market to be paired with dGPUs, which is to be frank, quite unlikely.

I would just about encourage AMD to avoid "big APUs" to focus on producing dies that are better focused on CPU core area because that is what most consumers really need at the end of the day. AMD can't afford multiple production lines for multiple dies, even though that would be best if they could achieve perfect lithography, yields, and excellent sales.

Depending on the die sizes and practical performance/Watt achievable, it seems that sticking to a native quad core Zen APU with 512/768 GCN SPs makes sense for majority of the mobile and low end desktop market (where it can fight Pentiums, and i3s), since it can cover such a broad range of systems via harvesting and binning.

It would be nice to see a bunch of nicely laid out native dies for every purpose but I see three sets of dies for a prospective Zen family:

4 Core/768 SPs - To fight Pentiums, i3s, and i5s in the laptop and low end and mini desktop space. Could be HBM ready or at least refreshed for it.
8 Core/256 SPs - Fight i7s in the laptop space, i5s and i7s in the desktop space.
16 Core/No SPs - Fight very high end i7s and Xeons in the server space.
 
Last edited:
Back
Top