*** AMD "Zen" thread (inc AM4/APU discussion) ***

[Orcvader]

Well to be fair while Bristol Ridge is Excavator based, the desktop variant does use the AM4 socket, shared with Zen.

 

[Panos]

Does anyone knows if OCUK going to bring any? HP UK isn't going apparently...

 

[AlamoX]

wow APUs are memory bandwidth hungry, assuming HP engineers are oblivious to the tech they are selling and how it works, AMD must have guide lines underlining the critical importance of bandwidth to the performance of the APU.
AMD shouldn't sell their APUs to partners like these who hold back the performance of their parts by a good 30% after AMD spends millions to squize extra 10%.

 

[Orangey]

Correct me if I'm wrong, but Bristol Ridge isn't Zen is it?

Now I'm sure we've had a similar situation before, but are people expecting Zen to hit this year, given they've just launched Bristol Ridge? I really can't see it (Yet the entire Internet is telling me Zen's Q3 2016, which starts in 27 days.)

Dunno where you heard 3, Q4 if anything is all I've heard since over a year now.

Quite possible it could be delayed or paper'd, however I don't see a problem as they would only be launching high-end CPUs as opposed to the APU thrust of Bristol. In fact Zen APUs will take quite some time extra IIRC, so around a year between those releases.

 

[Emlyn_Dewar]

I'm sure we'll never see another feasible AMD laptop.
They have some great mobile chips just now, but they are ALWAYS gimped with either 15w limit, single channel memory, and a horrific 1366x768 screen.
It's not going to change.

 

[Orangey]

I read something the other day about yet more Intel shenanigans in the laptop market.

No way are they going to let AMD get a foothold.

 

[Orcvader]

wow APUs are memory bandwidth hungry, assuming HP engineers are oblivious to the tech they are selling and how it works, AMD must have guide lines underlining the critical importance of bandwidth to the performance of the APU.
AMD shouldn't sell their APUs to partners like these who hold back the performance of their parts by a good 30% after AMD spends millions to squize extra 10%.

There were rumours some Zen based APUs will include HBM memory as well, which should allow the inbuilt GPU to fly along without being bandwidth bottlenecked. Similar to the eDRAM cache on the Intel Iris Pro graphics.

http://www.kitguru.net/components/a...n/amd-said-to-be-working-on-zen-apu-with-hbm/

 

[d_brennen]

Contra-revenue has to come to an end. Intel has mothballed fabs and the ones running are below capacity... Shareholders won't stand for it much longer

 

[CAT-THE-FIFTH]

Interesting:

http://semiaccurate.com/2016/06/06/amd-reveals-polaris-10rx480-zen-apus/

Apparently AMD might have had boxes with Zen APUs running at CES and it seems to have a new logo too.

 

[Orangey]

There's nothing about APUs in that, just CPUs.

 

[humbug]

Interesting:

http://semiaccurate.com/2016/06/06/amd-reveals-polaris-10rx480-zen-apus/

Apparently AMD might have had boxes with Zen APUs running at CES and it seems to have a new logo too.

I wonder if that box was powering the big screen

 

[Martini1991]

If zen had been there, I'd have expected it to be the machine they ran the 480 benchmark on, they could name it or not, but it was stated as an Intel high end machine.

 

[humbug]

If zen had been there, I'd have expected it to be the machine they ran the 480 benchmark on, they could name it or not, but it was stated as an Intel high end machine.

That would be spilling the beans, they ain't ready to publish performance figures yet, or they would have done it.

 

[SiDeards73]

I doubt it was Zen, I still don't think we will see any Solid Zen info for a few more weeks

 

[humbug]

I doubt it was Zen, I still don't think we will see any Solid Zen info for a few more weeks

More probably if Polaris 10 is anything to go by we wont see anything tangible until release.

 

[Martini1991]

That would be spilling the beans, they ain't ready to publish performance figures yet, or they would have done it.

They had no problem will multiple appearances of Zambezi in systems being publicly displayed backed in the day with games etc (Although obviously didn't give figures, which frankly they wouldn't have been doing so with the swarm benchmark, no one would associate the result with the CPU.)
So, it either wasn't Zen, or Zen's not the Intel killer it's being hyped too. (I still think Zen will be good, but there's far too many rumours of landscape changing chips and landscape changing prices)

 

[d_brennen]

I read an analysis on the 8 core die shot the other day. The author was pretty sure it had a chip-chip interface. Am I the only one dreaming of a dual socket AM4 (undoubtedly named G4x series) board, 16c/32t workstation for the cost of one top end Xeon? Hmmm...

 

[h4rm0ny]

If zen had been there, I'd have expected it to be the machine they ran the 480 benchmark on, they could name it or not, but it was stated as an Intel high end machine.

Definitely not. If you're going to show off your new graphics card with a comparative test, you're going to do it on a platform other people can recognize and be familiar with. You're not going to report "...and our 480 was this good against Nvidia's 1080 when running on some stuff that we wont tell you about". Instantly casts (valid) suspicion on the benchmarks that you're trying to make look good.

No way would they have done those benchmarks on Zen even if they had it available which I have doubts about.

 

[h4rm0ny]

I read an analysis on the 8 core die shot the other day. The author was pretty sure it had a chip-chip interface. Am I the only one dreaming of a dual socket AM4 (undoubtedly named G4x series) board, 16c/32t workstation for the cost of one top end Xeon? Hmmm...

You are definitely NOT the only one. And the server market is a big part of the goal with Zen (bigger than desktop, imo) so I think dual-socket AM4 workstation boards are a strong probability.

 

[Martini1991]

Definitely not. If you're going to show off your new graphics card with a comparative test, you're going to do it on a platform other people can recognize and be familiar with. You're not going to report "...and our 480 was this good against Nvidia's 1080 when running on some stuff that we wont tell you about". Instantly casts (valid) suspicion on the benchmarks that you're trying to make look good.

No way would they have done those benchmarks on Zen even if they had it available which I have doubts about.

The benchmarks are shrouded in confusion anyway because of the AMD failure that was stating 51% utilization, and the other Swarm oddities.

Besides, I don't particularly think that the test was a good comparison, it scores the same as much as my stock 290X does if you go off the 1.83 scaling that AMD mentioned etc.

Like I say, and I'll stick to it, it either wasn't there, or Zen's not the Intel slayer.