COMPUTEX 2018 Rumors and Info

[Ziip47]

Hello everyone,

As Computex 2018 is just around the corner I thought it is a good time to start a tread on current rumors and info. I am really looking forward to seeing what AMD has to bring to the table because they have been making up some lost ground in the GPU and CPU markets. I would also love to see what you guys know and what you are looking forward to seeing. Here is a link to a basic overview, from TR, of what to expect:

https://www.techradar.com/news/computex-2018

 

On a side note this is my first forum post ever. I am a relatively new member to the forum community so this is also a hello! 

[Crosseyed Sniper]

I was just thinking about this yesterday. I am hoping to see a few new mechanical keyboards and mice. Wonder if Corsair and Logitech will have anything to show? I am thinking we'll see new low-profile  keyboards with the new Cherry low-profile switches.

 

Always interesting to see all the latest new tech. Hopefully we'll get a lot of nice, new toys soon!  

[Skiiwee29]

This should have been posted under General and not in the Tech News forum. There are specific guidelines to posting in this forum. 

18 minutes ago, Ziip47 said:

Hello everyone,

As Computex 2018 is just around the corner I thought it is a good time to start a tread on current rumors and info. I am really looking forward to seeing what AMD has to bring to the table because they have been making up some lost ground in the GPU and CPU markets. I would also love to see what you guys know and what you are looking forward to seeing. Here is a link to a basic overview, from TR, of what to expect:

https://www.techradar.com/news/computex-2018

 

On a side note this is my forum post ever. I am a relatively new member to the forum community so this is also a hello! 

 

[laminutederire]

It would be almost mandatory to hear about RTGs advancements at this point. Communication is often key, and as of now, people tend to get a bit worried about the state of RTG considering how many people left and how Vega performed and how they seemingly have nothing to fuel any hype train of any sorts.

[Agonizel]

I think Corsair should up their game in the Big ATX Case department. Other than the 500D which felt meh (lacked design harmony imo) and the 1000D that's super over the top, there was nothing really significant. I'd like to see refreshes/remakes of the 500R, 750D & co

[Taf the Ghost]

1 hour ago, laminutederire said:

It would be almost mandatory to hear about RTGs advancements at this point. Communication is often key, and as of now, people tend to get a bit worried about the state of RTG considering how many people left and how Vega performed and how they seemingly have nothing to fuel any hype train of any sorts.

I doubt we're getting much out of AMD, minus TR2, this Computex. Though I expect it's going to be rumor central this year, especially since the big stuff is all coming next year, minus Nvidia's next generation.

 

One thing I am curious about: I think we're going to see Zen2-based CPUs across both TSMC & GloFo. I think we're going to see TSMC for Epyc and GloFo for Desktop/Mobile, which is why the rumors could be interesting. 

[laminutederire]

40 minutes ago, Taf the Ghost said:

I doubt we're getting much out of AMD, minus TR2, this Computex. Though I expect it's going to be rumor central this year, especially since the big stuff is all coming next year, minus Nvidia's next generation.

 

One thing I am curious about: I think we're going to see Zen2-based CPUs across both TSMC & GloFo. I think we're going to see TSMC for Epyc and GloFo for Desktop/Mobile, which is why the rumors could be interesting. 

Considering It seems clocks are often better on TSMC processes it'd make sense but would suck for us

Or it could not make sense as if it's going to be an overclocking wall like it is, it's better for higher end CPU since clocks are more limited by thermals at this point. So it'd make sense to use TSMC for high clocks on lower core count and reasonably high for high core count. Like 4.5-5GHz at that ipc on 16 cores wouldn't be unimpressive at all and still be quite competitive with Intel's parts, while less than 5Ghz on desktop would lead to AMD still lagging behind Intel because of clocks.

[Taf the Ghost]

32 minutes ago, laminutederire said:

Considering It seems clocks are often better on TSMC processes it'd make sense but would suck for us

Or it could not make sense as if it's going to be an overclocking wall like it is, it's better for higher end CPU since clocks are more limited by thermals at this point. So it'd make sense to use TSMC for high clocks on lower core count and reasonably high for high core count. Like 4.5-5GHz at that ipc on 16 cores wouldn't be unimpressive at all and still be quite competitive with Intel's parts, while less than 5Ghz on desktop would lead to AMD still lagging behind Intel because of clocks.

Clocks are always different between nodes. Highest "to market" node in the Desktop space still ends up being GloFo's 32nm, too bad it was with Piledriver. TSMC has been rolling out High Power then Low Power nodes, as a matter of cadence. I expect that the 7nm will work better for large die, moderate clocked parts. The 16c parts are going to need to stick pretty tight to the efficiency band, otherwise you're going to have 400w Server parts, which is a bit rough.

 

Which is why I think we're going to see two Zen2-based large core designs: 16c at TSMC and 12c at GloFo. It fits what AMD has put forward, the sudden move to Server-First, and it makes a lot more business sense.

[laminutederire]

2 hours ago, Taf the Ghost said:

Clocks are always different between nodes. Highest "to market" node in the Desktop space still ends up being GloFo's 32nm, too bad it was with Piledriver. TSMC has been rolling out High Power then Low Power nodes, as a matter of cadence. I expect that the 7nm will work better for large die, moderate clocked parts. The 16c parts are going to need to stick pretty tight to the efficiency band, otherwise you're going to have 400w Server parts, which is a bit rough.

 

Which is why I think we're going to see two Zen2-based large core designs: 16c at TSMC and 12c at GloFo. It fits what AMD has put forward, the sudden move to Server-First, and it makes a lot more business sense.

Well we will see  rumours are that it's hitting 5GHz, which if true would basically put them roughly on par or better than Intel, since they're roughly at the same IPC nowadays.

I still think RTG needs to say something about Navi-Vega 7nm anyway, solely for their reputation's sake. It's improbable they will, but they should.

[Taf the Ghost]

1 hour ago, laminutederire said:

Well we will see  rumours are that it's hitting 5GHz, which if true would basically put them roughly on par or better than Intel, since they're roughly at the same IPC nowadays.

I still think RTG needs to say something about Navi-Vega 7nm anyway, solely for their reputation's sake. It's improbable they will, but they should.

The RTG side of things has to be in one of the weirdest business positions. They're selling everything they're making, but no one is happy about it. Either the company or the normal consumers. Reminds me of GameCube Nintendo era: the designers felt like crap, even if the company was making a lot of money off the platform. (It couldn't do what all they wanted to, it was hard to expand the game library and party games weren't taking off.) 

 

The 5 Ghz part is why I think we're going to see 3 CCX and 4 CCX designs. 4 CCX goes to TMSC, first production servers show up around Dec 2018. Clocks roughly the same as Epyc 1, but with 64, 48 & 32 core parts as the main launch stack. The second design launches on GloFo's 7nm in Q1 for Server, bringing higher clocked parts at sizes 48, 36, 24 and 12. Somewhere late Q2 2019, we get the Ryzen 3000 series with up to 12 cores and Desktop/gaming clock speeds. (And maybe a very basic iGPU?)

 

Actually, running some quick numbers, a "Vega 6" iGPU should take up ~20 mm² on the 7nm process. Which would fit in roughly the same space as the cores (not the cache) for a full CCX on 7nm. Interesting.

[laminutederire]

38 minutes ago, Taf the Ghost said:

The RTG side of things has to be in one of the weirdest business positions. They're selling everything they're making, but no one is happy about it. Either the company or the normal consumers. Reminds me of GameCube Nintendo era: the designers felt like crap, even if the company was making a lot of money off the platform. (It couldn't do what all they wanted to, it was hard to expand the game library and party games weren't taking off.) 

 

The 5 Ghz part is why I think we're going to see 3 CCX and 4 CCX designs. 4 CCX goes to TMSC, first production servers show up around Dec 2018. Clocks roughly the same as Epyc 1, but with 64, 48 & 32 core parts as the main launch stack. The second design launches on GloFo's 7nm in Q1 for Server, bringing higher clocked parts at sizes 48, 36, 24 and 12. Somewhere late Q2 2019, we get the Ryzen 3000 series with up to 12 cores and Desktop/gaming clock speeds. (And maybe a very basic iGPU?)

 

Actually, running some quick numbers, a "Vega 6" iGPU should take up ~20 mm² on the 7nm process. Which would fit in roughly the same space as the cores (not the cache) for a full CCX on 7nm. Interesting.

I'm personally intrigued what they'll say now that their head of marketing has left.

Well 12 cores at 5GHz count me in

To be fair they don't even need 6, I'd argue 2 or 4 would suffice. Intel's igpu are just not good, but in a way the only thing you want from them is basically decoding Netflix videos and display the desktop an internet browsers as backup for anything but GPU intensive tasks why I still allowing the system to run on its on while a you is broken or waiting on the new GPU or something. Something cheap but useful when you rarely need it.

[Taf the Ghost]

21 minutes ago, laminutederire said:

I'm personally intrigued what they'll say now that their head of marketing has left.

Well 12 cores at 5GHz count me in

To be fair they don't even need 6, I'd argue 2 or 4 would suffice. Intel's igpu are just not good, but in a way the only thing you want from them is basically decoding Netflix videos and display the desktop an internet browsers as backup for anything but GPU intensive tasks why I still allowing the system to run on its on while a you is broken or waiting on the new GPU or something. Something cheap but useful when you rarely need it.

https://en.wikichip.org/wiki/amd/ryzen_3/2200u

 

This part, with "Vega 3" comes in as strong as a 8700k's iGPU. So, the reason for Vega 6 is that it would give built-in failure room. So if everything has a "Vega 4" with it, it'd all be good. They could then, over time, build out aspects like Intel has with their iGPU. I'm not sure the connection they'd use (IF?) rather then the PCIe lanes they use on the APUs, but, hey, maybe that extra x4 they didn't use could go that way for a basic iGPU. Intel actually puts upwards of 30% of their 6c die to the iGPU. AMD could put a lot less, by %, and provide more.

 

I'm not sure if they'll do it this generation, but it would open up a lot more OEM options in the future.

[ElZamo92]

I’m genuinely not that interested. My PC is fine as it is for now, specially because of the marginal improvements in performance that will most likely be revealed. The 1180 is rumoured to be barely faster than the 1080ti, which does not impress me in the slightest, specially after the two years they took to milk Pascal dry instead of developing significant improvements to the architecture. Intel’s rumored eight core CPU is out of the question for me, since I recently switched to the AM4 platform, and going back to Intel would be an expensive endeavour. The only thing I’m kind of interested in are new storage tech (one can never have too much storage), specially on the solid state side of things, and maybe developments in memory tech.