Lisa Su confirms Q3 launch for Ryzen, Epyc and NAVI

#5669900 Posted on: 05/16/2019 02:28 PM

---

some people hear "new" and use it as a placeholder for their wishes.

Navi is, was, and will be a mid-range card with mainstream cut-down cards.

don't get it twisted, i'm a big fan of Ms. Su. this is not a slam at AMD or Nvidia, it is what is.
i bought a radeon VII (which i love) to replace my 1080ti for a period of 18 months to two years, until the full iteration of the "chiplet" design, "Arcturus" is on the market.

that said, when they have a full Navi/Ryzen laptop, i'll buy it.

#5669949 Posted on: 05/16/2019 04:17 PM

---

Both of these slides are from this year; 2 things. They clearly go out of their way to designate their next GPU to be their "next gen" architecture, and they also specifically indicate that we won't see any real efficiency gains until that architecture. It's not in quotes either, so "next gen" is not the name.Efficiency gains are what AMD needs to be able to do much of anything at all at this point, and they indicate we shouldn't really expect much in that department.. They are literally TDP limited to 1080 Ti performance on a smaller node at this point. Navi definitely won't be that fast.

I can't find the slide, but there has been one floating around for a couple of years that has Polaris/Vega/Navi listed before "Next Gen". Common sense would tell you that Navi is still clearly GCN.

There is also this:

https://www.pcgamesn.com/amd/navi-linux-confirms-gcn-design


edit: One thing I will say though, which is entirely possible, is that Navi will actually be as big as Vega. I don't find this to be very likely given how expensive early 7nm is, but it's possible I suppose, and it would line up with rumors (save for the TDP).





That's all, you got there? AMD always puts "Next Gen"/"Next-Gen" to products for which they do not have fitting code name. Please understand that NextGen != Next uArch. Generation is not specific in anything. It is very generic.
Then those 4 mentioned things are not time schedule, but approach and it applies globally. Not to certain time frame or GPU generation.
As for Linux thing, it does not confirm anything else than Navi being mostly driver compatible with GCN which has been stated in one of videos long time before someone found America again on Phoronix.

And to fill you in. 7nm is far from early since Radeon 7 been made for quite some while and has 13B transistors. Phone chips available have more transistors than RX-580. IIRC Apple is having made 10B+ chips. 7nm is doing well in yields as mentioned few times before.

So, do you actually have something that's not misinterpretation of source material?

#5669955 Posted on: 05/16/2019 04:27 PM

---

I currently have no reason to upgrade my 1500X, and that will probably remain the case until the next-gen consoles are released. But, I am still very interested in Zen2's performance. The die shrink, increased clock speeds, higher core/thread count, better RAM support, and Intel's security mitigations leads me to believe Zen2 will be the first truly all-around better CPU for the first time in over a decade.

All that being said, I'm mostly interested in Navi. All I want is something that can play 4K@60FPS with AA off for a reasonable price.

#5669977 Posted on: 05/16/2019 05:25 PM

---

Been waiting for something to replace my GTX1080 but NVIDIA is too expensive, and from what i've been reading even top Navi isn't that good either, sigh.

Guess i'll have to wait for Intel GPUs so that NVIDIA starts putting fair prices and better hardware (7nm RTX hopefully).

#5669981 Posted on: 05/16/2019 05:34 PM

---

Hmm, I wonder what they mean by "all new GPU architecture". Everything they've shown until now has indicated Navi was just going to be another GCN iteration.

It can easily be an all-new architecture and be GCN at the same time...just like the fact that a Ryzen 2700 and an Intel 8088 are both considered "x86" because of backwards compatibility even though the designs are totally unlike each other. We'll see...but you have to also realize this is an "investor" press release...not a consumer spec-sheet.

OTOH, it was always obvious to me that RVII was a pro-sumer product (16GBs ram) and that another line of gaming GPUs was slated for release in 2H. Lisa Su practically said exactly that to the "pundits" on the RVII release Interviews she did--but as usual many of them didn't catch it. I'll be surprised if AMD doesn't skewer nVidia again in the volume discreet GPU markets as for some reason nVidia simply isn't competitive in the < ~$300 discrete GPU volume markets. Navi will probably be 8GBs for the gaming market--but I really couldn't say.

It may be ground breaking, but possibly not quite as useful. There is only so much hardware every core has, and while a second thread can make use of scheduling deficits to use idle hardware quite well, there are extremely strong diminishing returns from adding more threads to individual cores - nevermind software limitations in even keeping more threads busy.

In a consumer line, like Ryzen, I really don't see the added value for more SMT threads.

Yes, If the 1 SMT thread capability isn't enough to keep the core going at 100%, then more than one might be needed. But since AMD has little trouble implementing lots of cores, there doesn't seem to be any room for a second SMT thread per core.