AMD Zen 3 (5000 Series), rumored 17% IPC gain.

[Gerard]

This is from PCgamer but the link takes you to the source

Hopefully the case, a little more or a little less either way possibly.

 

[Panos]

I don't usually do this, but right now I want to point out how full of crap you are.

In the Intel thread you called me an idiot, because you claimed a September release of Zen3, and I questioned where you'd got that from.

You then said, "It's on AMD's Twitter, learn to Google." You kept on about September being the official release date.

Now you're saying "September-October at the earliest."

So hold on... why have you rowed back on the "official" September release date? That I was an "idiot" for questioning?

Maybe because it's YOU who can't read, or Google? Apology, perhaps? Or just admit you're clueless.

That was 2-3 weeks ago mate. Information changes. If we have Matisse 2 next month as some rumours flying around, more likely we will see Zen 3 early 2021.

 

[FoxEye]

That was 2-3 weeks ago mate. Information changes. If we have Matisse 2 next month as some rumours flying around, more likely we will see Zen 3 early 2021.

In fact nothing has changed. There was never any "information" in the first place.

There was never an announcement on AMD's Twitter that said September.

You still can't point to one because it never happened.

And yes I'm cheesed off because you chewed me out when I asked you for a source the first time.

 

[Panos]

In fact nothing has changed. There was never any "information" in the first place.

There was never an announcement on AMD's Twitter that said September.

You still can't point to one because it never happened.

And yes I'm cheesed off because you chewed me out when I asked you for a source the first time.

https://www.google.com/search?rlz=1...hUKEwiF38inrLTpAhVLDWMBHSvoDUcQ4dUDCAw&uact=5

As said, learn to use Google.

 

[CAT-THE-FIFTH]

Yeh the 3000 series isnt bad for DAW work at all, but there is little point in me upgrading to something with 12 or 16 cores, if the latency issues are still there for the large core parts

For example:

https://www.scanproaudio.info/2019/...00x-dawbench-tested-3-is-it-the-magic-number/

The 3900X with its new dual chiplet design was the only model to not come away with a completely clean sheet, although given we have an extra die section to deal with, this might not prove to be a huge surprise. I would expect it to mature in much the same fashion as the other chips below it in the range have done over future iterations and no doubt looking forward to them fine-tuning this new design further.

Models upto the Ryzen 7 3800X use a single chiplet AFAIK.

 

[FoxEye]

https://www.google.com/search?rlz=1C1CHBF_enGR859GR859&sxsrf=ALeKk02QiOnSos1UorxUZCPcVjE6t3frUQ:1589493391051&ei=j769XsXbAsuajLsPq9C3uAQ&q=AMD+Zen+3+september&oq=AMD+Zen+3+september&gs_lcp=CgZwc3ktYWIQA1CWYFiWYGDVYWgAcAB4AIABgQGIAYEBkgEDMC4xmAEAoAEBqgEHZ3dzLXdpeg&sclient=psy-ab&ved=0ahUKEwiF38inrLTpAhVLDWMBHSvoDUcQ4dUDCAw&uact=5

As said, learn to use Google.

You can't admit you're wrong, can you?

There was no OFFICIAL announcement or ANY announcement via AMD's Twitter of a September release.

You think I don't know there were rumours? What use are rumours? They are worthless.

You are STILL full of it. You were ADAMENT that AMD had given a September release for Zen3. They didn't. They haven't. You're WRONG. And in this thread you admit it.

 

[Grim5]

https://www.google.com/search?rlz=1C1CHBF_enGR859GR859&sxsrf=ALeKk02QiOnSos1UorxUZCPcVjE6t3frUQ:1589493391051&ei=j769XsXbAsuajLsPq9C3uAQ&q=AMD+Zen+3+september&oq=AMD+Zen+3+september&gs_lcp=CgZwc3ktYWIQA1CWYFiWYGDVYWgAcAB4AIABgQGIAYEBkgEDMC4xmAEAoAEBqgEHZ3dzLXdpeg&sclient=psy-ab&ved=0ahUKEwiF38inrLTpAhVLDWMBHSvoDUcQ4dUDCAw&uact=5

As said, learn to use Google.

Can you please explain how

1) Nvidia has managed to build 820mm2 die at 7nm when you said 500mm2 is the absolute maximum TSMC can do.
2) Nvidia has managed to fit in 54 billion transistors, therby by destorying AMD's transistor density by 66%.

 

[Panos]

Can you please explain how

1) Nvidia has managed to build 820mm2 die at 7nm when you said 500mm2 is the absolute maximum TSMC can do.
2) Nvidia has managed to fit in 54 billion transistors, therby by destorying AMD's transistor density by 66%.

Where is the 820mm2 7nm Nvidia GPU? Nowhere.

 

[Grim5]

Where is the 820mm2 7nm Nvidia GPU? Nowhere.

did you like just skip the last 24 hours

 

[Panos]

did you like just skip the last 24 hours

Yes. Yet you excuse a 400W 826mm GPU that would cost tens of thousands due to low yields?
The articles we argued were about consumer cards. Get your head around.

 

[alec]

Apparently the leaker is well respected so can be trusted.
https://www.reddit.com/r/Amd/comments/gjo2ad/rumor_zen3_will_exceed_expectations_just_like/

Funny translation. Too little information to guess what is going on. But any performance increase is welcome, be it single or multi core.
Smells like hype train is leaving the station.

 

[Panos]

did you like just skip the last 24 hours

FYI 826mm2 includes the 6 stacks of HBM ram!!!!!!!!

 

[Grim5]

FYI 826mm2 includes the 6 stacks of HBM ram!!!!!!!!

you sure, link to this info? If that's the case and the die is much smaller then it makes the transistor density even more impressive, amazing what they've achieved. Hope rdna2 can catch up

 

[Panos]

you sure, link to this info? If that's the case and the die is much smaller then it makes the transistor density even more impressive, amazing what they've achieved. Hope rdna2 can catch up

GV100 was marketed as "815mm2" but that included HBM also too.

 

[Deleted User 456458]

I can't find any evidence that the die size includes HMB2 whatsoever.

The die visuals don't show it either.

https://cdn.wccftech.com/wp-content/uploads/2020/05/NVIDIA-GA100-GPU-Block-Diagram-1030x516.png

** Do Not Hotlink Images **

 

[Grim5]

GV100 was marketed as "815mm2" but that included HBM also too.

Then its even more impressive than we thought this morning at the announcement, 54bill transistors in such a small die and the 63 billion transistor model is still coming, AMD has a lot of catching up to do.

 

[Panos]

Then its even more impressive than we thought this morning at the announcement, 54bill transistors in such a small die and the 63 billion transistor model is still coming, AMD has a lot of catching up to do.

TSMC has 0.09 sqm defect rate on the 300mm 7nm waffer. That means best case scenario 27-30 chips per wafer at that size.

5700XT in comparison has yield over 224 chips per wafer at same rate and Zen 2 chiplet over 762. (which is inline with AMD saying that Zen 2 has over 90% yield).
Each 7nm wafer costs $10,000

Let the sink in before you carry on trying to extrapolate that chip to 3000 series.

 

[CAT-THE-FIFTH]

Comparing transistor densities is not so easy between companies. Some companies report only active transistors,ie,those actually used in a fully functional chip,but others also include non-functional transistors(they exist in a die),and those included for improving yields.

Also last time I checked Navi was built last year,and is probably a more conservative design made for a far less mature TSMC 7NM. AFAIK,Nvidia is also using a newer and more denser version of the process.

We will need to wait and see how RDNA2 CPUs,stack up. We can kind of get a slight inkling from the consoles,but remember the CPU part will always be less denser than the GPU part.

 

[hyperseven]

GV100 was marketed as "815mm2" but that included HBM also too.

Link to this please. I tried use Google, nada, maybe I am doing it wrong? You seem to have the knack of finding things that never existed

 

[pete910]

GV100 was marketed as "815mm2" but that included HBM also too.

Are they including the transistor count then too ?