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

[CuriousTomCat]

Apparently the Ryzen engineering sample on stage at CES had a 3.7GHz base clock with a 4.5GHz boost. The 9900k would have boosted to a slightly higher speed of 4.7GHz.

See the part at 24min 52 seconds

 

[4K8KW10]

Now you mention it, i actually clicked the link and read it

I cant simulate the 3.4Ghz base clock with a boost clock @ 3.6Ghz so i did 3.6Ghz all cores.

Hope i got this right, if so then its now something for the mathematicians to figure out

--------

Edit.. Added this > Userbench Possible 12c Ryzen 3xxx

Userbench 2700X at 3.6Ghz / 2666Mhz RAM

Cinebench 2700X at 3.6Ghz / 2666Mhz RAM

So, 8% IPC up in the application.
Multi-core float is also very strong on the Zen 2 ES.

 

[humbug]

Apparently the Ryzen engineering sample on stage at CES had a 3.7GHz base clock with a 4.5GHz boost. The 9900k would have boosted to a slightly higher speed of 4.7GHz.

See the part at 24min 52 seconds

That would give the Ryzen 3000 a 5% higher IPC, assuming the 4.5Ghz was an all core boost.

 

[4K8KW10]

That would give the Ryzen 3000 a 5% higher IPC, assuming the 4.5Ghz was an all core boost.

Zen+ --> Zen 2:
29% IPC up in DKERN + RSA;
8% IPC up in UserBenchmark;
5% IPC up in Cinebench R15.

 

[Scougar]

That would give the Ryzen 3000 a 5% higher IPC, assuming the 4.5Ghz was an all core boost.

Can't watch right now.. is it actually confirmed in the video with that speed?

 

[TrixX]

That would give the Ryzen 3000 a 5% higher IPC, assuming the 4.5Ghz was an all core boost.

Not sure they are right. Info from CES said the CPU used didn't have a boost clock just an all core overclock. RedGamingTech reported that after some maths the CES CPU was likely running 4.0GHz assuming a 15% IPC increase. 4.5GHz would be roughly correct if the IPC increase was only 4-5% too. Either way it's interesting to see which one gets validated.

 

[Scougar]

The question I have is.... Why would and choose to price these chips so cheap? Are they going to attempt to gain market and mind share again, as otherwise I just don't see why they would price such seemingly good chips at a low price when they could easily in my mind, make more profit whilst still taking from Intel.

I say this, because every time we hoped AMD would come in with something awesome for a cheap price, it came in at higher prices than we hoped.

Edited: for silly phone corrections.

 

[Potatowithearsontheside]

It took RedGakingTech nearly 3 weeks to work out that with a 15% IPC boost the ES would d have been running at 4.0GHz?
And these guys are some kind of knowledgeable folk?
Sheesh.
We all did the math about 5 seconds after the CES ES CB score was demonstrated.

 

[TrixX]

It took RedGakingTech nearly 3 weeks to work out that with a 15% IPC boost the ES would d have been running at 4.0GHz?
And these guys are some kind of knowledgeable folk?
Sheesh.
We all did the math about 5 seconds after the CES ES CB score was demonstrated.

Yeah but some people don't like forumites who do napkin maths to work stuff out. Some even go as far to say 'unless we have the actual information as a fact from the source, there's no point anything we do could be accurate as it's just speculation...'

Those of us who aren't afraid to analyze and speculate however can work out it was 4-4.2GHz with a 15% IPC gain and at 4.5GHz the IPC gain is as low as 4%. Though I don't doubt there's a ton of microcode optimization to go on before release.

 

[nashathedog]

why is it on the ryzen 2600 i can have a 4.1 oc at 1.36 volts thats stable, but to get to 4.2 i need to increase the voltage much higher

Maybe this is a situation where the difference between the X & non-X models rears it's ugly head, Have you tried leaving trhe voltage on auto?
All I did with my 2700x was change the multiplier to 42.50 (4.25 ghz) & then hit save & exit, job done, It's got to be worth a try.
I'm using my 2700x with two 8gb sticks of 3200mhz ram, an Asus X470 Prime Pro Motherboard & a Corsair GTX H100i All-In-One & temps max out in the low 60's under load...

 

[humbug]

I'm far more interested in what they go on to say, $99 50 watt chip beating the 9600K, matching the 9700K, $180 3600 beating the 9900K.... $500 Ryzen 9 3850 matching or beating the $1,700 i9 9980XE..... the CPU market is about to get turned on its head.

https://youtu.be/ew642t9_m88?t=1545

 

[nashathedog]

Can't watch right now.. is it actually confirmed in the video with that speed?

No

The question I have is.... Why would and choose to price these chips so cheap? Are they going to attempt to gain market and mind share again, as otherwise I just don't see why they would prices such seemingly good chips at a low price when they could easily in my mind, make more origin whilst still talking from Intel.

I say this, because every time we hoped and y would come in with something awesome for a cheap price, it came in at higher prices than we hoped.

Personally I think these nunbers being quoted are wide of the mark, I think we're more likely to get what their graph in the video calls the 3700 as the flagship, all the others are probably just wishful thinking,

In a recent interview Lisa Su confirmed that there would be more than 8 cores available but not how many cores or how many different models, For me a 12 core 24 thread cpu with a max clock of around 4.6 or 4.7 ghz sounds realistic. Unlike the higher examples. I suppose we might see a 16 core 32 thread model but again I can't see clocks going beyond 4.6 or 4.7 ghz.

 

[humbug]

No



Personally I think these nunbers being quoted are wide of the mark, I think we'll like get what their graph in the video calls the 3700 as the flagship all the others are probably just wishful thinking,

In a recent interview Lisa Su confirmed that there would be more than 8 cores available but not how many cores or how many different models, For me a 12 core 24 thread cpu with a max clock of around 4.6 or 4.7 ghz sounds realistic. Unlike the higher examples. I suppose we might see a 16 core 32 thread model but again I can't see clocks going beyond 4.6 or 4.7 ghz.

What is your reasoning behind that?

Putting two 8 core dies on a package and only selling it in upto 12 core form is incredibly wasteful, and don't forget AMD's 50'th anniversary is coming up in a few months.

On clock speed your still only thinking along Intel monolithic die terms, which of course they ain't...

 

[Jokester]

Especially when the 8-core die might be only pulling 40-50W!

 

[humbug]

For the 8 die alone probably a bit more than 50 watts but not much...

The 4.6Ghz 3700 12 core is 95 watts, that includes the IO die, take 20 watts? for the IO die. you're left with 70/75watts for the cores die, add that to a 95 watt package you've got 165/170 watts, use the best silicon for the flagship part and you can take another 20 watts off it leaving you with about 150 watts.

That's actually about the same power as Intel's "95 watt" 9900K.

The difference 3600 vs 9900K in the CES demo was 50 watts.

 

[nashathedog]

What is your reasoning behind that?

Putting two 8 core dies on a package and only selling it in upto 12 core form is incredibly wasteful, and don't forget AMD's 50'th anniversary is coming up in a few months.

On clock speed your still only thinking along Intel monolithic die terms, which of course they ain't...

Hi,
I'm being realistic, keeping my hopes & dreams in check, I want to be wrong but based on what we know about Ryzen I think anything beyond 4.7 is unlikely, I'm sure AMD would like to do a limited run of anniversary chips with top end clocks but wanting too and being able too are completely different things. If I'm right and they offer a 4.6 4.7 flagship maybe they'll be able to do some cherry picking to add a couple of exta points to a very limited run, who knows? When you look at Vega going from 14nm to 7nm the clocks haven't got a big boost & I don't think 7nm Ryzen will either, an additional .3, .4 or maybe .5 if we're lucky sounds likely. I hope their clocks work very differently to today's Ryzen but if not a claimed 4.7 is likely to mean an all core overclock of around 4.6 at best. On paper the 2700x's stated to have a max boost clock of 4.325 but that's a single core clock, Across all cores 4.2 or 4.25 is the norm. So far Ryzen comes close to it's limit in terms of core clocks out of the box, I'm expecting that won't change but as I've already said I'd love them to prove me wrong.

 

[humbug]

Hi,
I'm being realistic, keeping my hopes & dreams in check, I want to be wrong but based on what we know about Ryzen I think anything beyond 4.7 is unlikely, I'm sure AMD would like to do a limited run of anniversary chips with top end clocks but wanting too and being able too are completely different things. If I'm right and they offer a 4.6 4.7 flagship maybe they'll be able to do some cherry picking to add a couple of exta points to a very limited run, who knows? When you look at Vega going from 14nm to 7nm the clocks haven't got a big boost & I don't think 7nm Ryzen will either, an additional .3, .4 or maybe .5 if we're lucky sounds likely. I hope their clocks work very differently to today's Ryzen but if not a claimed 4.7 is likely to mean an all core overclock of around 4.6 at best. On paper the 2700x's stated to have a max boost clock of 4.325 but that's a single core clock, Across all cores 4.2 or 4.25 is the norm. So far Ryzen comes close to it's limit in terms of core clocks out of the box, I'm expecting that won't change but as I've already said I'd love them to prove me wrong.

I get that, TBH i am assuming that those scores are at 4.5Ghz all core Turbo because if not AMD have a significant IPC lead over Intel, "significant" like 15%. i and many analysing this find it far more believable that AMD have the clocks up, and a smaller jump in IPC.

For the 8 die alone probably a bit more than 50 watts but not much...

The 4.6Ghz 3700 12 core is 95 watts, that includes the IO die, take 20 watts? for the IO die. you're left with 70/75watts for the cores die, add that to a 95 watt package you've got 165/170 watts, use the best silicon for the flagship part and you can take another 20 watts off it leaving you with about 150 watts.

That's actually about the same power as Intel's "95 watt" 9900K.

The difference 3600 vs 9900K in the CES demo was 50 watts.

I actually have this wrong ^^^^ Jokester is right, i still have it in my head that Ryzen #600 are 6 core..... it is an 8 core at 4.5Ghz at 55 watts, that includes the IO die, so 135 Watts for the 16 core 3850... easily that or less, infact that suggests it is a 16 core.

 

[humbug]

Ryzen 9 3850 5.1Ghz 135 watt package

Ryzen 5 3600 4.5Ghz 55 watt package, 10 watts for the IO die? add that to the 55 watt package = 100 watts, that missing 35 watts could be for the extra Mhz.

Even if the clocks are lower i go back to what i have been saying all along, the clocks are irrelevent, we know what the performance is from the CES demo, be that at 4Ghz or 4.5Ghz.... its irrelevant it beat the 9900K so from "a performance stand point" this is happening.

Edit: AMD have form, if they can push 5Ghz clocks they will, remember the FX-9590?

 

[Jokester]

Yup, unless AMD seriously balls up the IO chiplets performance, and they’re already available I think in their higher end chips? There’s no reason to believe that these chips will be seriously capable.

 

[humbug]

Yup, unless AMD seriously balls up the IO chiplets performance, and they’re already available I think in their higher end chips? There’s no reason to believe that these chips will be seriously capable.

Rome, 'Zen2 64 core EPYC' haven been sampling with OEM's since November.

Ryzen 3000 single die 8 core for comparison.