AMD Previews 7nm Ryzen CPUs at CES

AlphaAtlas

[H]ard|Gawd
Staff member
Joined
Mar 3, 2018
Messages
1,713
At their 2019 CES keynote, AMD's CEO Lisa Su revealed a new generation of Ryzen processors built on TSMC's 7nm process. First, AMD demo'd an "early version" of the upcoming processor on Forza 4, and AMD claims it consistently ran the game at over 100 FPS. Then, they pitted the 8 core, 16 thread processor against a Core i9-9900k in Cinebench. The 3rd gen Ryzen processor managed to hit a slightly higher Cinebench score while only consuming about 133 watts on average, while the 9900K ate over 179 Watts running the same workload.

Like AMD's 7nm Epyc processors, 3rd gen Ryzen uses a multi-chip design. Lisa mentioned that the I/O die is actually the larger of the 2 dies on the chip she held up, and I would bet it's fabricated on a node larger than 7nm. 3rd gen Ryzen will allegedly be the first PC platform to support PCIe 4.0, but it will still run on existing AM4 motherboards, as previously promised. AMD says 3rd gen Ryzen will launch "in the middle of 2019," and that more details will come as the launch date gets closer.
 
A little disappointed that the launch will be close to the midyear, this would mean we have products towards back to school/early summer which is later than I expected.

However given that its a sample with non finalized score and it beats the 9900k is impressive.
 
Interesting... they went with a chiplet design, with 1/4 I/O die, as I suspected...

...but this ES only has one chiplet, despite the clear room for a second chiplet.

I wonder if AMD plans to release the 8 core, single chiplet first... and a 12 and 16 dual chiplet (and/or 8c + Vega GPU) later. Or if the best working engineering samples are only the single chiplet ones, for now.
 
Interesting... they went with a chiplet design, with 1/4 I/O die, as I suspected...

...but this ES only has one chiplet, despite the clear room for a second chiplet.

I wonder if AMD plans to release the 8 core, single chiplet first... and a 12 and 16 dual chiplet (and/or 8c + Vega GPU) later. Or if the best working engineering samples are only the single chiplet ones, for now.

She never specified that the ES was a Ryzen 3,5, or 7...
 
She never specified that the ES was a Ryzen 3,5, or 7...

I know. I'm speculating why she chose to demo the single chiplet design, when room was clearly made for a second chiplet. BoiseTech's reply is on point, though... probably means dual chiplet designs will be released later.
 
She never specified that the ES was a Ryzen 3,5, or 7...

This is why I'm wondering if they showed the Ryzen 5 version in that Cinebench test. Can a Ryzen 5 beat the Intel chip or are those leaks wrong and the Ryzen 7 is only 8c/16t?
 
This is why I'm wondering if they showed the Ryzen 5 version in that Cinebench test. Can a Ryzen 5 beat the Intel chip or are those leaks wrong and the Ryzen 7 is only 8c/16t?

Clearly room for a second chiplet. AMD stuck the chiplet in the top right for a reason... that doesn't happen by accident.

That being said, no idea if this is a Ryzen 5, 7, or some other thing. No idea if/when dual chiplet designs will be released either.
 
She never specified that the ES was a Ryzen 3,5, or 7...

the 3/5/7 is used to compare against intel class for class.. if they do end up doing a 16/32 am4 chip it'll most likely be called ryzen 9.

my prediction is 16/32 core and 12/24 core ryzen 9, unannounced 10/20 core and 8/16 core ryzen 7, 6/12 core and 4/8 ryzen 5 and 4/4 ryzen 3.
 
Last edited:
This is why I'm wondering if they showed the Ryzen 5 version in that Cinebench test. Can a Ryzen 5 beat the Intel chip or are those leaks wrong and the Ryzen 7 is only 8c/16t?

It'll be one or the other, but I would guess this will end up being the Ryzen 5, due to all the space on the CPU for another chiplet.
 
AMD Raedeon Navi $699 actually selling for $1200 on e-tailer's store. Coming February 7th.......... Coming to acomputer no where near you, because limited quantities are a thing.

How is this the opposite of Ryzen.
 
Maybe they want to make ThreadRipper a more mainstream element? I can't quite see a 12/16c making sense for a gaming rig, save for streamers and enthusiasts who the ThreadRipper might address.
 
Clearly room for a second chiplet. AMD stuck the chiplet in the top right for a reason... that doesn't happen by accident.

That being said, no idea if this is a Ryzen 5, 7, or some other thing. No idea if/when dual chiplet designs will be released either.

I agree there seems room for something else on the chip. However, if a ~$240 Ryzen 5 is competing with the 9900k in multithreading? That's huge.
 
Maybe they want to make ThreadRipper a more mainstream element? I can't quite see a 12/16c making sense for a gaming rig, save for streamers and enthusiasts who the ThreadRipper might address.

eh i could, basically it's the same thing intel did with i think it was those useless 8/16 dual channel 2066 socket chips(can't even remember what they were called anymore, lol) except on am4 and not requiring a 400+ dollar motherboard.

threadripper has basically moved on to be AMD's epyc processor enthusiast class and no longer is an "experiment".
 
Maybe they want to make ThreadRipper a more mainstream element? I can't quite see a 12/16c making sense for a gaming rig, save for streamers and enthusiasts who the ThreadRipper might address.

You can't see it *now* but now that game devs will have systesm with 8+ cores to develop to, we should see a significant boost in multicore utilization for single applications while people may not need the increased cache, PCIe lanes, and quad channel memory.
 
Like everyone else said no mention of specs as far as clock speed or performance increase but just guesstimating on what they showed us and looking at other reviews about a 10% increase in clk to clk IPC performance over the 2000 series. If she's telling the truth and the clock speeds are from an ES and not set I would have to guess they are at least running at the same speed of a 2700x if not little higher. I'm guessing final clocks will be around 4.8ghz at the highest. If they can get those chips to 5GHz or higher they're going to be killing Intel.
 
  • Like
Reactions: GHRTW
like this
I feel like a huge thing people are missing that the powers listed are total system powers (180W vs 130W). According to Anandtech the chip power for the i9 is 125W but the AMD CPU is at 75W...

At Just Over Half The Power…?!
Also, in that same test, it showed the system level power. This includes the motherboard, DRAM, SSD, and so on. As the systems were supposedly identical, this makes the comparison CPU only. The Intel system, during Cinebench, ran at 180W. This result is in line with what we’ve seen on our systems, and sounds correct. The AMD system on the other hand was running at 130-132W.

If we take a look at our average system idle power in our own reviews which is around 55W, this would make the Intel CPU around 125W, whereas the AMD CPU would be around 75W.

Same performance with numbers of core/threads at about HALF the power means some pretty serious increases in IPC or frequency.
 
A little disappointed that the launch will be close to the midyear, this would mean we have products towards back to school/early summer which is later than I expected.

Did you seriously expect AMD to launch Matisse (desktop Ryzen 3000 Series) at CES?

I have been saying that whole time that the launch is probably going to happen sometime around May.
 
Did you seriously expect AMD to launch Matisse (desktop Ryzen 3000 Series) at CES?

I have been saying that whole time that the launch is probably going to happen sometime around May.


You may have been saying that, but the rumormongers said CES:

upload_2019-1-9_13-15-6.png
 
Intel is already ready and anyone who put faith in AdoredTV is an idiot.
It clearly stated TBA , that means: To Be Announced.
It is not about faith in AdoredTV it is where the leaks are coming from. Beside his technical guess work pretty much got Ryzen 3000 down.
Even the IO die and the chiplets.
 
Intel is already ready and anyone who put faith in AdoredTV is an idiot.

Compelling argument...

I hope this turns out as good as it looks, autumn is right around the time I was hoping to do a hardware refresh, and it's been far too long since I've built myself an AMD system.
 
Intel is already ready and anyone who put faith in AdoredTV is an idiot.

He wasn't all wrong either. There's clearly room for a second chiplet.

I suspect the Zen 2 leak was correct in general terms, but was incorrect about launch timing/order.

Rumors/leaks should always be taken with big grains of salt. Talk about them - bullshit about them - don't take them too seriously.
 
Intel is already ready and anyone who put faith in AdoredTV is an idiot.
I don't think anyone just accepts rumours/leaks as facts, however I am not sure the leak is so far off from facts as it seems the on stage performance matches the Ryzen 5 3600X in the leak.

It is interesting to see it actually beat the 9900k which indicates either an IPC advantage, Cinebench bias or a clock frequency advantage.... Assuming ofc. that the 9900k didn't throttle...
 
I don't think anyone just accepts rumours/leaks as facts, however I am not sure the leak is so far off from facts as it seems the on stage performance matches the Ryzen 5 3600X in the leak.

It is interesting to see it actually beat the 9900k which indicates either an IPC advantage, Cinebench bias or a clock frequency advantage.... Assuming ofc. that the 9900k didn't throttle...


He wasn't all wrong either. There's clearly room for a second chiplet.

I suspect the Zen 2 leak was correct in general terms, but was incorrect about launch timing/order.

Rumors/leaks should always be taken with big grains of salt. Talk about them - bullshit about them - don't take them too seriously.

If you extrapolate 2700x performance and use some common sense, ANYONE could have predicted these performance results with cinebench that was shown today.
 
This is why I'm wondering if they showed the Ryzen 5 version in that Cinebench test. Can a Ryzen 5 beat the Intel chip or are those leaks wrong and the Ryzen 7 is only 8c/16t?

You say "only 8c/16t," its not long ago that 4c/8t was the best desktop chip you could get. How far we have come when 8c/16t is considered to be nothing special.
 
Looks like I'll have to hold onto my 4790k a little longer, was hoping to get something sooner than mid next year and I'm sure as hell not giving Intel any more of my money being you cant ever upgrade to a new chip without a mobo swap etc.
 
You say "only 8c/16t," its not long ago that 4c/8t was the best desktop chip you could get. How far we have come when 8c/16t is considered to be nothing special.
Back in my day we had 1 core, 1 thread and cooled everything with 8 fins of aluminium without these fancy "fans". Lot of innovation since my 2700+ days.
 
Been I while since my last hype train. The high is good but sooner or later reality hits in. It's been way, way worse than than.

May is good though. Two 8 core zen 2 cpu's to table 5 please
 
For reference,

A 2700X OCed to 4.4ghz with CL14 3200mhz DDR4 scores 1964 in CB:
untitled-13.png

(Guru3d)
The 16 thread Zen 2 CPU scored 2057:
Capture+_2019-01-09-14-07-33.png
(9900k left, AMD 16 thread right)

If your only variable is clockspeed, then a 2700X clocked at 4.6ghz would achieve a similar score that was shown off today as 16 thread Zen 2.

Zen 2 is evidently quite power efficient. I hope that means there is a lot of headroom for overclocking.

Curious to see how this one plays out.
 
Back in my day we had 1 core, 1 thread and cooled everything with 8 fins of aluminium without these fancy "fans". Lot of innovation since my 2700+ days.

You are right JosiahBradley, my first CPU was an Intel 286 which didn't have a co processor !!
 
Back
Top