• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

Associate
Joined
14 Dec 2016
Posts
958
really hard to say until you have it in front of you and its tested. I mean take the Ryzen 5 2600x for example, on paper it has the same number of cores and threads and a higher clock speed than Intel's i7 8700 (at base) yet in the majority of benchmarks, the i7 performs better.

Mainly because Intel boosts higher, Ryzen boost is capped extremely low, if Ryzen boosted as well as Coffeelake, then i think you'd see a totally different picture.
 
Caporegime
Joined
17 Mar 2012
Posts
47,382
Location
ARC-L1, Stanton System
really hard to say until you have it in front of you and its tested. I mean take the Ryzen 5 2600x for example, on paper it has the same number of cores and threads and a higher clock speed than Intel's i7 8700 (at base) yet in the majority of benchmarks, the i7 performs better.

Of course it does, its clocked much higher, the IPC is pretty much identical tho...

@4Ghz

8700K:
ST 174
MT 1325

2600:
ST 168
MT 1384

so at the same clocks the 8700K is 3% fast ST while the 2600 is 5% faster in MT.

dcQvZib.png
 

Deleted member 209350

D

Deleted member 209350

Mainly because Intel boosts higher, Ryzen boost is capped extremely low, if Ryzen boosted as well as Coffeelake, then i think you'd see a totally different picture.

Yep, but I would have assumed that the X chips would have been boosted better. But for now at least, Intel is doing something better

Of course it does, its clocked much higher, the IPC is pretty much identical tho...

@4Ghz

8700K:
ST 174
MT 1325

2600
ST 168
MT 1384

so at the same clocks the 8700K is 3% fast ST while the 2600 is 4% faster in MT.

dcQvZib.png

Yeah they are very similar, but the fact this is only a Ryzen 5 up against an i7 speaks volumes. Zen2 will take this to another level, I have a gut feeling
 
Caporegime
Joined
17 Mar 2012
Posts
47,382
Location
ARC-L1, Stanton System
Oh I get it, I just wasn't thinking broadly enough. One track mind today was just going "golden chips = EPYC pool" without factoring in a high clocker isn't needed for a server part.

Pretty much, the best chips are not about clock speed, they have a power budget and the best will simply fall into that power budget most efficiently, all the other chips it doesn't matter so much, they can be clocked high on a much higher power budget, tho the best of the mainstream will still be taken from higher quality stock, maybe not the top 20% but the top 20 to 30%.
 
Soldato
Joined
26 Sep 2010
Posts
7,146
Location
Stoke-on-Trent
You're digging yourself deeper again. You don't take criticism well at all, do you?

Stating previously I'd totally brain-farted and then making self-deprecating jokes about the egg on my face totally shows I can't take criticism :rolleyes:

I have a few minutes free at work, I can Photoshop you up a "LePhuronn gone fkd up" badge if it'll make you feel a bit more superior? Because I'm nice like that.
 
Soldato
Joined
13 Jun 2009
Posts
6,847
Of course it does, its clocked much higher, the IPC is pretty much identical tho...

@4Ghz

8700K:
ST 174
MT 1325

2600:
ST 168
MT 1384

so at the same clocks the 8700K is 3% fast ST while the 2600 is 5% faster in MT.

dcQvZib.png
It should be noted that this shows Cinebench IPC, not IPC in a general sense (which is almost impossible to calculate these days).

Pretty much, the best chips are not about clock speed, they have a power budget and the best will simply fall into that power budget most efficiently, all the other chips it doesn't matter so much, they can be clocked high on a much higher power budget, tho the best of the mainstream will still be taken from higher quality stock, maybe not the top 20% but the top 20 to 30%.
I'm still not sure about the whole binning situatio. I mean sure, server chips don't need to clock super high, they just need to remain energy efficient. However, we already know the top 5% bins become Threadripper chips so I am not convinced the cheaper desktop chips will boost to 5.1 GHz.
 
Caporegime
Joined
17 Mar 2012
Posts
47,382
Location
ARC-L1, Stanton System
It should be noted that this shows Cinebench IPC, not IPC in a general sense (which is almost impossible to calculate these days).


I'm still not sure about the whole binning situatio. I mean sure, server chips don't need to clock super high, they just need to remain energy efficient. However, we already know the top 5% bins become Threadripper chips so I am not convinced the cheaper desktop chips will boost to 5.1 GHz.

You're right, tho i do think its a pretty accurate overall representation, there is a video about somewhere, in it someone did a very detailed IPC comparison on all aspects of Ryzen + vs Skylake (Same as Coffeelake) and found they are pretty much identical across the board, but he also concluded Ryzen's IPC is heavily bottlenecked by a weak front end, i think we knew that, its why it responds so well to fast memory.

I would say Gaming is where its really all over the place, sometimes lower than Coffeelake IPC sometimes higher, depending on the game, for example Arma III in a single patch went from being typically AMD slow to getting a huge boost in performance equalling that of Coffeelake.
The thing is Arma III had never seen Ryzen, which is why it was patched.

I'll try and find the video, no promises its been a few months since i last knew where it was.

8400 boosts to 4Ghz.

0S3MLtf.png
 
Associate
Joined
29 Oct 2009
Posts
145
a lot of people are doubting high clockspeeds from the next chip, if it were still fabbed at GF then I'd agree, they have always been a let down. You could argue that their 14/12nm process had more in common with Intels 22nm than the current one. TSMC however, is a full node ahead of Intel, for the first time ever allowing AMD to leapfrog Intel hopefully.
 
Caporegime
Joined
17 Mar 2012
Posts
47,382
Location
ARC-L1, Stanton System
a lot of people are doubting high clockspeeds from the next chip, if it were still fabbed at GF then I'd agree, they have always been a let down. You could argue that their 14/12nm process had more in common with Intels 22nm than the current one. TSMC however, is a full node ahead of Intel, for the first time ever allowing AMD to leapfrog Intel hopefully.

GF's 14/12nm (Same thing) was designed for low power mobile chip's Tablets / Phones, 65 Watts for an 8 core 16 thread X86 CPU with the same IPC as Coffeelake, actually much more internal, just bottlenecked, that's impressive, unfortunately because the process was tuned for mobile getting them near 4Ghz required a lot of volts on the first 14nm iteration and not much past 4Ghz on the refined 12nm iteration.

TSMC's 7nm is more a proper high speed proccess, i still think only about 4.7Ghz on all cores but IMO with a small bump in IPC that's enough.
 
Caporegime
Joined
8 Jan 2004
Posts
31,994
Location
Rutland
GF's 14/12nm (Same thing) was designed for low power mobile chip's Tablets / Phones, 65 Watts for an 8 core 16 thread X86 CPU with the same IPC as Coffeelake, actually much more internal, just bottlenecked, that's impressive, unfortunately because the process was tuned for mobile getting them near 4Ghz required a lot of volts on the first 14nm iteration and not much past 4Ghz on the refined 12nm iteration.

TSMC's 7nm is more a proper high speed proccess, i still think only about 4.7Ghz on all cores but IMO with a small bump in IPC that's enough.

Hype train!
 
Caporegime
Joined
17 Mar 2012
Posts
47,382
Location
ARC-L1, Stanton System
Hype train!

By what measure? i know Intel have given us almost nothing from generation to generation in the past few years but it must be clear by now AMD are not Intel.
10% higher Mhz is hardly "Hype train" nor is a small bump in IPC, Zen to Zen+ was just a small tweak and it achieved both those things, i'm simply saying Zen 2 will achieve from Zen+ what it did from Zen, perhaps slightly more than that.
 

HeX

HeX

Soldato
Joined
20 Jun 2004
Posts
12,015
Location
Huddersfield, UK
By what measure? i know Intel have given us almost nothing from generation to generation in the past few years but it must be clear by now AMD are not Intel.
10% higher Mhz is hardly "Hype train" nor is a small bump in IPC, Zen to Zen+ was just a small tweak and it achieved both those things, i'm simply saying Zen 2 will achieve from Zen+ what it did from Zen, perhaps slightly more than that.

Should be more. Zen to Zen+ the entire improvement was down to the move from 14nm to 12nm process.

Zen2 is the 'tick' in old Intel terms, it's the proper architecture update, while at the same time having the 'tock' move to 7nm... it should give some real gains compared to current Zen+.
 
Soldato
Joined
18 Oct 2002
Posts
14,046
Location
West Midlands
a lot of people are doubting high clockspeeds from the next chip, if it were still fabbed at GF then I'd agree, they have always been a let down. You could argue that their 14/12nm process had more in common with Intels 22nm than the current one. TSMC however, is a full node ahead of Intel, for the first time ever allowing AMD to leapfrog Intel hopefully.

A lot of people also don't realise that AMD's current CPU's boost up to 4.4GHz out of the box, namely the Threadripper 2950X, which is the highest speed of any of the desktop or HEDT parts in the current generation using Zen+. Going to 7nm an adding 15% extra (660MHz) can't be that hard for a node shrink done correctly, taking it to 5GHz+ at least on a few cores. :)
 
Soldato
Joined
6 Jun 2008
Posts
11,616
Location
Finland
Yep, but I would have assumed that the X chips would have been boosted better. But for now at least, Intel is doing something better
Ryzen's current clocks aren't about AMD limiting them.
It's about AMD having been stuck with second rate GlobalFoundries' manufacturing node.
Which GloFo, after failing in their own R&D, actually had to buy/license from Samsung, who originally developed it for low clocks/power mobile SOCs.
Hence original expectations for first Ryzen iteration were pretty much for 3,5GHz clock speed.
Compared to that achieved clocks were actually good.

So Zen1 was basically betrayed by behind lagging manufacturing node.
Made on node clocking as high as Intel's tech, Intel would have been hard pressed to have lead in any benchmark.
 
Back
Top Bottom