Ryzen on 12nm might get a 50% core increase

[cj09beira]

So a few days ago, kitguru in one of there videos had a slide up for a few seconds, showing ryzen+'s skus, they claim that the slides are real, but take it with huge grain of salt.

(i didn't include a quote from what he says as he only glances at it, which might say something to his confidence on the slide)

here is the video (at the correct time stamp):

Quote

 

UPDATE: Now with a slide that is better for our eyes 

Quote

 

as we can see we have:  (assume boosts are single core)

ryzen 2700   with 12 cores 4ghz    base clock and 4.5 boost with a price of 329 (it actually said 12 cores  )

ryzen 2800   with 12 cores 4.4ghz base clock and 4.9 boost with a price of 399

ryzen 2800x  with 12 cores 4.6ghz base clock and 5.1 boost with a price of 449

 

 

it also talks about a launch on February 1st

 

now about the authenticity of the slide,

first the clocks:

•        global foundries's 14nm is around 5-10% slower than tsmc's 16nm, and global foundries does claim a 10% increase over competing 16/12nm products, which should put goflo's 12nm at 15-20% better clocks than what we have now with 14nm.
• a 20% clock increase would put today's xfr at 4.9Ghz (4.1*1.2)

 

50% more cores:

•        ever since we heard of 14nm+ and ryzen+, the increase in only clocks seemed a bit off for a company that did say no more incremental upgrades, so it seems within reason to have 12 cores on am4, i just thought it would come with ryzen on 7nm.
• Amd has already confirmed that the next gen of Epyc would support 48 cores, we though they were talking about 7nm, maybe not

 

 

my personal opinion:

My expectations of ryzen+ were simply a 10% clock increase with maybe 1-5% ipc gain just from small optimizations and fixes, with the same amounts of cores, so around 4.5ghz max clocks, which was enough to postpone my purchase to february.

deep down it always seemed odd that the perf increase was going to be small, especially with ryzen on 7nm being in 2019.

also people were asking why amd was going to spend the money porting all their designs to the new node, only to get around 10% higher clocks, if its more closer to 20% it makes far more sense.

the 4.9-5.1 boosts still seem far fetched, i guess we will have to wait and see. 

[Vode]

This does sound too good to be true, although it seems logical from what little information we have.

[Jade]

I'd be cool with more ipc gain rather than more cores. I didn't like the Bulldozer strat. Fewer fast cores > many mediocre cores.

[cj09beira]

1 minute ago, Jade said:

I'd be cool with more ipc gain rather than more cores. I didn't like the Bulldozer strat. Fewer fast cores > many mediocre cores.

well the increased clocks will reduce that problem alot, i dont expect significant ipc gains until ryzen on 7nm

[cj09beira]

Just now, Name Taken said:

32 core Threadripper?

 

24 core threadripper, 48 core epyc

[AresKrieger]

50% core number means nothing to me +50% core clock then I'm interested

 

Frankly unless you are constantly encoding video or something along those lines I don't see the point currently

[xg32]

4.5 was my dream scenario, i just don't see 5.1 SHIPPED to be true.

[cj09beira]

1 minute ago, AresKrieger said:

50% core number means nothing to me +50% core clock then I'm interested

 

Frankly unless you are constantly encoding video or something along those lines I don't see the point currently

people should read the whole article first

[DildorTheDecent]

11 minutes ago, cj09beira said:

ryzen 2700   with 10 cores 4ghz    base clock and 4.5 boost with a price of 329

ryzen 2800   with 12 cores 4.4ghz base clock and 4.9 boost with a price of 399

ryzen 2800x  with 12 cores 4.6ghz base clock and 5.1 boost with a price of 449

Yeah right....5.1 boost is going to require serious power. 5GHz on any unlocked chip is never that easy. Requires some proper tuning to get it right.

[AresKrieger]

2 minutes ago, cj09beira said:

people should read the whole article first

The only part I believe is the core count increase, because that has been their strategy from the get go

[cj09beira]

Just now, AresKrieger said:

The only part I believe is the core count increase, because that has been their strategy from the get go

i think it was more like the current node cant do more than 4ghz, than lets only have more cores,

a core clock increase this good also points to a very close collaboration between the two companies (same with tsmc's 12nm)

[AresKrieger]

1 minute ago, cj09beira said:

i think it was more like the current node cant do more than 4ghz, than lets only have more cores,

a core clock increase this good also points to a very close collaboration between the two companies (same with tsmc's 12nm)

It's not the node its more so the design itself, changing the node may help but the design is limiting in terms of clock frequency and memory frequency from what I've seen, however if they target servers more heavily its a non-issue from a business perspective however it doesn't interest me as a consumer.

[cj09beira]

1 minute ago, AresKrieger said:

It's not the node its more so the design itself, changing the node may help but the design is limiting in terms of clock frequency and memory frequency from what I've seen, however if they target servers more heavily its a non-issue from a business perspective however it doesn't interest me as a consumer.

all points to being a node problem, not a design problem, as even the gpus didn't clock much higher than the 28nm counterparts, and in the nodes documents available they talk of only 3ghz+

[AresKrieger]

4 minutes ago, cj09beira said:

all points to being a node problem, not a design problem, as even the gpus didn't clock much higher than the 28nm counterparts, and in the nodes documents available they talk of only 3ghz+

They weren't using 28nm though so they should be able to increase it beyond what they currently have it clocked to......actually maybe not I keep forgetting to factor in that an AMD chip at 16nm isn't actually a 16nm design if using the same standards of ibm or intel

 

I guess we'll see soon enough (though I still have my doubts)

[Bananasplit_00]

we dont NEED this amount of cores, like REALLY, but damn i cant say that i think a $400 12 core at potentially 5GHz is a bad chip

[Jito463]

1 hour ago, Jade said:

I'd be cool with more ipc gain rather than more cores. I didn't like the Bulldozer strat. Fewer fast cores > many mediocre cores.

IPC is fine with Ryzen, it's just the clock speed that needs improving.

[Trixanity]

I call bullshit. Sounds like someone making shit up for advertising revenue.

 

I'll gladly be proved wrong in 3 months time but there is pretty much no chance of this being real.

[Sakkura]

58 minutes ago, AresKrieger said:

The only part I believe is the core count increase, because that has been their strategy from the get go

AMD's strategy with Ryzen is not focusing on more cores. The top of the line Ryzen 7 1800X has the same core count as the FX-8150 they released in 2011.

[vanished]

A 50% jump in core count, plus a 27.5% boost to clockspeed (4 -> 5.1 GHz) and a 5% IPC bump works out to pretty much exactly 2x in multi-threaded performance.  If they can achieve that and while dropping price by 10% at the same time, that would be amazing.  My main concern would remain thermals though; you can't just throw 50% more cores on a package and boost clocks from 4 to 5 GHz without either cooking it, or making some serious improvements to efficiency, and to think they could pull off the other improvements and a vast efficiency improvement too at the same time would be impressive to the point of unbelievable.

 

Looks great, and I hope they manage this, but I'll believe it when I see it.

[AresKrieger]

Just now, Sakkura said:

AMD's strategy with Ryzen is not focusing on more cores. The top of the line Ryzen 7 1800X has the same core count as the FX-8150 they released in 2011.

I meant in general not specifically for ryzen, the only time I can think of where they lagged intel in core count on the consumer side was when intel released there first consumer dual core, either way I'm not a fan of rumors and am often very dismissive towards them unless an official statement addresses them.

[Bananasplit_00]

33 minutes ago, DildorTheDecent said:

Yeah right....5.1 boost is going to require serious power. 5GHz on any unlocked chip is never that easy. Requires some proper tuning to get it right.

no? 5GHz is easy as hell from my experience and from what i have seen people saying about 6th and 7th gen. put the multiplier at 50x and the voltage somewhere around 1.4V and done. power draw goes up but it dosent get insane. 12 cores at 5GHz probably would draw quite a lot, an 1800x is like 150W so 250W TDP or so?

[Drak3]

6 minutes ago, Ryan_Vickers said:

you can't just throw 50% more cores on a package and boost clocks from 4 to 5 GHz without either cooking it, or making some serious improvements to efficiency, and to think they could pull off the other improvements and a vast efficiency improvement too at the same time would be impressive to the point of unbelievable.

It's entirely possible that 4c/8t, 6c/12t, and 8c/16t SKUs remain.

[Sakkura]

4 minutes ago, Ryan_Vickers said:

My main concern would remain thermals though; you can't just throw 50% more cores on a package and boost clocks from 4 to 5 GHz without either cooking it, or making some serious improvements to efficiency, and to think they could pull off the other improvements and a vast efficiency improvement too at the same time would be impressive to the point of unbelievable.

 

Looks great, and I hope they manage this, but I'll believe it when I see it.

Adding cores does not make cooling that much harder, because added die area aids heat dissipation. Upping the clocks tends to be a bigger killer. Still, a good clock bump alongside a big jump in core count simultaneously is definitely going to be very, very difficult to pull off.

 

I'd personally be ecstatic if they achieve even half that increase in core count while actually getting to around 5 GHz.

Just now, AresKrieger said:

I meant in general not specifically for ryzen, the only time I can think of where they lagged intel in core count on the consumer side was when intel released there first consumer dual core, either way I'm not a fan of rumors and am often very dismissive towards them unless an official statement addresses them.

Heh, it's been a while since they added cores though. Just Intel decided quad cores were going to be the mainstream thing forever.

[AresKrieger]

Just now, Sakkura said:

Heh, it's been a while since they added cores though. Just Intel decided quad cores were going to be the mainstream thing forever.

Tbf without the rise of streaming it is likely more cores would still be fairly irrelevant to consumers, but I agree it was intel that lagged in that area rather than amd leaping

[vanished]

6 minutes ago, Sakkura said:

Adding cores does not make cooling that much harder, because added die area aids heat dissipation. [...]

Sure there's more area to pull heat from but there's still more heat