AMD might announce their new CPU lineup on Tuesday

[Taf the Ghost]

1 minute ago, leadeater said:

Well we can only use what ever products actually have Gen-Z and the easiest right now is SCM, Gen-Z NVMe would be the next easiest. Those are just the same products with updates Gen-Z controllers in front of the DRAM/NAND etc. I do expect to see in chassis SCM though (Page 20).

Yup. Intel loves their TAM (Total Addressable Market), but AMD produces an entirely new one for itself by getting GenZ switches to market. The really interesting part is if Vega 20 can attach to it. That would be...interesting.

[RejZoR]

Biggest bottlenecks these days are interconnections between subsystems. CPU by itself has massive internal bandwidth. GPU's have massive internal bandwidth (even outside of GPU to the onboard DRAM). Problem are connections betwen CPU, RAM and GPU (as well as storage).

 

Ideal would be when memory becomes an unified thing so we won't have RAM and storage separately anymore. It would just be "memory" in which you store and execute stuff. And then linking that to CPU and GPU using a massive highway. That's really the only way to address system wide bottlenecks that we are fighting for basically decades.

[Trixanity]

5 hours ago, leadeater said:

Better not be an R7 GPU, those were crap.

 

I don't think there will be much talk about GPUs though, it seems both too early and I hope they will just focus on one segment. AMD's just not good at juggling multiple things at once.

Possibly a more firm Vega 20 launch date if not an outright launch to go along with any Zen 2 updates. The 7 should just mean that it's all about 7nm but that's of course pretty vague.

[cj09beira]

5 hours ago, leadeater said:

I'm not too sure about timing either, I just know Vega 7nm compute cards are coming soon and so is 7nm EPYC2. Which will actually come first, or both at same time don't know. Unless this event is going to have a large focus on EPYC2 I don't see much value in talking about 7nm Vega or 7nm Navi which is further away than both as far as I know.

pretty sure we got confirmation that it would be vega 7nm, epyc 2, ryzen 3xxx, navi

[porina]

2 hours ago, RejZoR said:

Biggest bottlenecks these days are interconnections between subsystems. CPU by itself has massive internal bandwidth. GPU's have massive internal bandwidth (even outside of GPU to the onboard DRAM). Problem are connections betwen CPU, RAM and GPU (as well as storage).

 

Ideal would be when memory becomes an unified thing so we won't have RAM and storage separately anymore. It would just be "memory" in which you store and execute stuff. And then linking that to CPU and GPU using a massive highway. That's really the only way to address system wide bottlenecks that we are fighting for basically decades.

My concern with this is less so about bandwidth, as that can be brute forced in some way. Latency remains an open question. As "low latency" as Gen-Z claims to be, I'd have to assume it would be a significant increment over native DRAM class devices.

 

It then comes down to optimising for the workload. A scaleable design would be great for some enterprise level applications, but I'm not sure it would make sense at consumer level in near to mid term.

[Brooksie359]

10 hours ago, ravenshrike said:

Could be an EPYC reveal and a basic Zen+ to Zen2 CCX performance comparison.

 

*shrugs* Pretty sure they still are obligated to produce something through GloFo, even if its no longer their primary processors.Could be an easy way to meet that fab's obligations.

Didn't they get to renegotiate because global foundries were supposed to deliver 7nm as part of their agreement?

[Doobeedoo]

Can't wait to see these, 7nm desktop chips are looking to be very interesting. 

[Humbug]

7 hours ago, RejZoR said:

Biggest bottlenecks these days are interconnections between subsystems. CPU by itself has massive internal bandwidth. GPU's have massive internal bandwidth (even outside of GPU to the onboard DRAM). Problem are connections betwen CPU, RAM and GPU (as well as storage).

latency, not bandwidth is the problem. I.e. we can transfer data at massive rate per second (bandwidth). We need to reduce the time it takes in Nano seconds to get from point a to b (latency).

[Taf the Ghost]

Rumors are flying fast & furiously this morning. The weird thing is there might be something related to Raytracing included with the announcements. Which either might be interesting or just as a way to placate some investors that don't know what that means.

[ravenshrike]

4 hours ago, Brooksie359 said:

Didn't they get to renegotiate because global foundries were supposed to deliver 7nm as part of their agreement?

Yeah, which is why they're no longer obligated to use them for Ryzen. But the agreement wasn't entirely eliminated IIRC and AMD still needs to produce something through GloFo to keep a certain percentage of their fabs running.

[Jito463]

A bit of clickbait in the title.  It should have just been about the upcoming show, and then a side note about the possibility of a Zen2 announcement.

On 11/4/2018 at 1:42 PM, leadeater said:

Wonder if we'll get a semi quick Zen2+ that has both DDR4 and DDR5 support or just DDR5 support. It's just so damn close to DDR5 it's a shame not to be able to use it right away. If we're really lucky Zen 2 actually does have support for both, with new motherboard required for DDR5 when they come available.

On 11/4/2018 at 2:01 PM, Ben Quigley said:

AMD has a long history of supporting multiple memory types on one chip... Am4+ anyone?

Really, the only time they did that was with AM3 chips, which were backwards compatible to AM2+ boards.  I can't think of another single instance, nor can I think of an instance where they made a chip that was forwards compatible with different RAM (AM2+ chips were not compatible with AM3 boards, for example).  I highly doubt we'll see AM4 chips with a DDR5 controller, though it's possible we could see AM5 chips be backwards compatible to existing AM4 (or AM4+?) motherboards.

[Nicnac]

Here's what the rumour mill has been spilling out so far:

"Unexpectedly high core frequencies"

"better XFR"

"Increased IPC, focus is now on better latency"

oh yea and that weird raytracing rumour...

 

if anyone speaks chinese: https://www.chiphell.com/thread-1926186-1-1.html

[williamcll]

8 minutes ago, Nicnac said:

Here's what the rumour mill has been spilling out so far:

"Unexpectedly high core frequencies"

"better XFR"

"Increased IPC, focus is now on better latency"

oh yea and that weird raytracing rumour...

 

if anyone speaks chinese: https://www.chiphell.com/thread-1926186-1-1.html

Also
"2. focus on improving memory latency"
"3. AVX performance increase (unsure about AVX 512 support)"
"6. uncertain ryzen support for PCIe-4"

 

 

There is also a mention of a Nvidia GA100 that may be released end of next year and maybe on the 7nm scale as well.

[williamcll]

EDIT: wrong thread

[Taf the Ghost]

17 minutes ago, williamcll said:

Also
"2. focus on improving memory latency"
"3. AVX performance increase (unsure about AVX 512 support)"
"6. uncertain ryzen support for PCIe-4"

 

 

There is also a mention of a Nvidia GA100 that may be released end of next year and maybe on the 7nm scale as well.

Ampere is Nvidia's 7nm design, and the GA parts are going to be Professional/HPC. They aren't bringing consumer 7nm parts next year unless something goes really, really wrong.

[leadeater]

How long until event starts?

[Taf the Ghost]

1 minute ago, leadeater said:

How long until event starts?

It's in San Fran, so it's at least 7+ hours, though I can't find an exact start time.

[Taf the Ghost]

36 minutes ago, Nicnac said:

Here's what the rumour mill has been spilling out so far:

"Unexpectedly high core frequencies"

"better XFR"

"Increased IPC, focus is now on better latency"

oh yea and that weird raytracing rumour...

 

if anyone speaks chinese: https://www.chiphell.com/thread-1926186-1-1.html

Apparently the slang for Intel's desktop platform gets translated by Google as "toothpaste factory". That's some meme context I'd love to know, haha.

[leadeater]

6 minutes ago, Taf the Ghost said:

It's in San Fran, so it's at least 7+ hours, though I can't find an exact start time.

I hate time zones, it's almost Wednesday damn it 

[Taf the Ghost]

2 minutes ago, leadeater said:

I hate time zones, it's almost Wednesday damn it 

https://www.timeanddate.com/worldclock/converter.html?iso=20181106T090000&p1=264&p2=224

 

One downside with the Internet is the classic, "Okay, what time is it where?". I like to keep tools handy.

[cj09beira]

41 minutes ago, Taf the Ghost said:

https://www.timeanddate.com/worldclock/converter.html?iso=20181106T090000&p1=264&p2=224

 

One downside with the Internet is the classic, "Okay, what time is it where?". I like to keep tools handy.

some said something about 2pm, which if true, means i will have to wait until late today (just woke up a few hours ago) 

https://www.timeanddate.com/worldclock/converter.html?iso=20181106T220000&p1=264&p2=224&p3=1422

[cj09beira]

1 minute ago, cj09beira said:

some said something about 2pm, which if true, means i will have to wait until late today (just woke up a few hours ago) 

https://www.timeanddate.com/worldclock/converter.html?iso=20181106T220000&p1=264&p2=224&p3=1422

cj dont forget that they said some info would be released at 9am too, @Taf the Ghost though for you that will be in prime sleep time

[Taf the Ghost]

18 minutes ago, cj09beira said:

cj dont forget that they said some info would be released at 9am too, @Taf the Ghost though for you that will be in prime sleep time

Whether I am Stateside or not this week, one shouldn't assume I am ever on the correct time schedule for the region.

[GoldenLag]

34 minutes ago, Taf the Ghost said:

Whether I am Stateside or not this week, one shouldn't assume I am ever on the correct time schedule for the region.

i dont know how timezonenames work, but +1 time its 15:00. hopefully we will have something as AMD`s site is empty, which it really shouldnt be at this time. there is confusion as the event is a San Francisco which would make it 06:00 for them..........

 

http://ir.amd.com/events/event-details/amd-next-horizon

 

so much info..........

[cj09beira]

correction amd's event is starting at 9am PST

https://www.anandtech.com/show/13547/amd-next-horizon-live-blog-starts-9am-pt-5pm-utc