Intel crosses the old bridge - new information regarding GPU line with 2021 release.

[williamcll]

VideoCardz has received new information regarding Intel's new graphics card, it will still take sometime before it appears on the market.

Quote

Intel has disclosed to the press that their upcoming 7nm Xe GPU is codenamed Ponte Vecchio. This is the first GPU to be named officially, but we expect other Xe GPUs to be released sooner.

The name “Ponte Vecchio” came from an old stone bridge located in Florence, Italy. The bridge basically refers to the interconnection between the GPUs, or as Intel calls it – CXL (Compute Express Link) – which Ponte Vecchio GPUs will use. Intel Ponte Vecchio is not a gaming GPU. The first Xe graphics are for exascale computing. On November 17th Intel will share details on the project “Aurora”. This exascale computer features Sapphire Rapids Xeon CPUs, Ponte Vecchio GPUs and Intel’s new initiative called oneAPI (unified programming model). In the new press deck (that we have totally not seen) Intel claims that Ponte Vecchio will use Foveros packing technology and will utilize CLX interconnection (this was already known). What’s new is that they confirmed that the Xe graphics feature: ultra-high cache and high memory bandwidth. Intel Ponte Vecchio will also have high double-precision FP throughput.

 

Source:https://videocardz.com/newz/intels-first-xe-graphics-processor-is-called-ponte-vecchio

Thoughts: 2021? It will be a difficult race for intel to reach the GPU performance like that of Nvidia and AMD.

[Flying Sausages]

Intel can pull 7nm on GPU but not on CPU?!

[LeSheen]

42 minutes ago, OlympicAssEater said:

Intel can pull 7nm on GPU but not on CPU?!

My best guess is that a CPU is way more complex.

Not an expert by any means only had the base from my computerscience education

[Arika]

was going to say that name was very Italian

 

 

48 minutes ago, OlympicAssEater said:

Intel can pull 7nm on GPU but not on CPU?!

completely different architectures

[Roswell]

15 minutes ago, Arika S said:

was going to say that name was very Italian

 

 

completely different architectures

Architecture matters little as far as their die shrinking woes go. It's the fabrication process itself that they're having issues with.

[Bombastinator]

1 hour ago, OlympicAssEater said:

Intel can pull 7nm on GPU but not on CPU?!

I also could use a summary of that summary regarding what it would mean to a consumer level gamer.

[thorhammerz]

3 hours ago, williamcll said:

It will be a difficult race for intel to reach the GPU performance like that of Nvidia and AMD.

Depends on how well Xe scales with larger (or more) GPU cores/slices - halo product performance in 2021 (3080Ti and Big Navi) will probably be around 2080ti + 50% (conservative estimate), not an impossible bar to reach if the architecture can be scaled up.

[pas008]

2 hours ago, OlympicAssEater said:

Intel can pull 7nm on GPU but not on CPU?!

 

1 hour ago, LeSheen said:

My best guess is that a CPU is way more complex.

Not an expert by any means only had the base from my computerscience education

my guess is commercial product money 

no expert either

[Mira Yurizaki]

2 hours ago, LeSheen said:

My best guess is that a CPU is way more complex.

That shouldn't have any bearing on how difficult it is to manufacture a chip. If anything, GPUs tend to be subject to higher failure rates when manufacturers go big because you can literally throw more execution units for performance with the workloads GPUs are expected to handle.

 

1 hour ago, Tegos said:

Correct me if I'm wrong, but isn't this the same as Xeon Phi?

No, it's an evolution of their existing GPU technology.

[SpaceGhostC2C]

22 hours ago, Tegos said:

Correct me if I'm wrong, but isn't this the same as Xeon Phi?

Xeon Phi is still x86. It's halfway GPU-like in terms of parallelization, while retaining the simplicity of using code meant for CPUs.

These GPUs have their own architecture, so Intel is developing this "oneAPI" environment to try and bridge the two, in search of the holly grail of code that you can write without knowing/caring where it's going to run, then scale as needed through CPU/FPGA/GPU.

 

In that sense, it's similar to what AMD was pushing some time ago. Can't remember the name, but was basically the excuse to start advertising APUs as "10 compute cores (4 CPU +6 GPU)", etc. As with everything software, AMD never go far with it, but just like Mantle resurrected through Vulkan, hopefully oneAPI brings the concept to fruition. My hopes on it not being very locked down to Intel are scarce, though .

(At leas they invited me to beta test it )

[mr moose]

6 hours ago, OlympicAssEater said:

Intel can pull 7nm on GPU but not on CPU?!

10nm and 7nm R+D are different teams, 10nm ran into issues but 7nm hasn't and is still on track.  We'll likely see CPU's on 7nm in the not too distant future baring some sort of  density issue or a incompatibility in design (given their history with density and attempts to stack, anything is possible).

[RejZoR]

Will Intel fab these GPU's themselves or will they use TSMC for it?

[Mira Yurizaki]

17 minutes ago, RejZoR said:

Will Intel fab these GPU's themselves or will they use TSMC for it?

Intel has plenty of fabrication capacity, so I don't see why they won't do it.

[Unixsystem]

8 hours ago, LeSheen said:

My best guess is that a CPU is way more complex.

Not an expert by any means only had the base from my computerscience education

 

9 hours ago, OlympicAssEater said:

Intel can pull 7nm on GPU but not on CPU?!

 

6 hours ago, pas008 said:

 

my guess is commercial product money 

no expert either

You guys miss the big "2021" in the title? 

[leadeater]

5 hours ago, SpaceGhostC2C said:

In that sense, it's similar to what AMD was pushing some time ago. Can't remember the name, but was basically the excuse to start advertising APUs as "10 compute cores (4 CPU +6 GPU)", etc.

AMD HSA

[Flying Sausages]

29 minutes ago, Waffles13 said:

 

 

You guys miss the big "2021" in the title? 

[JoshHendi]

This should be interesting