Core 9000 series

[Silent_Scone]

Until AMD catch up with IPC and IMC strength there will always be a market for another iteration (9000).

I think he is probably refering to the rumoured x299 refresh cpus with up to 22 cores. They are expected to be released with new x299 refresh motherboards.

However I emailed AsRock about this and they stated that so far they have no plans to develop any further motherboard on x299 chipset.

So you are probably correct.

MB vendors are completely flat out as it is. If there's a HEDT refresh coming from Intel, at the very earliest you're probably looking around Christmas.

 

[humbug]

Until AMD catch up with IPC and IMC strength there will always be a market for another iteration (9000).

April 2019 AMD will overtake Intel, and Intel know it.

 

[Silent_Scone]

April 2019 AMD will overtake Intel, and Intel know it.

One for the quote bank.

 

[humbug]

One for the quote bank.

Vs Coffeelake

Ryzen 1### had a few % less IPC using just one core, the same IPC using multiple cores
Ryzen 2### has the same IPC using one core, a few % higher IPC using multiple cores

If Ryzen 3 gets a similar 4% IPC bump it will be significantly ahead of Coffeelake, add to that for 7nm Global Foundries are citing 40% higher performance vs 14nm, that's 40% higher clock speeds than Ryzen 1###, so if we take the 1800X at 3.6Ghz and add 40% you get 5Ghz.

That would make Coffeelake well and truly beat, Intel have no answer for Ryzen 3, their 10nm is way off, its why they are rebranding 14nm Coffeelake, the 9000 series is what will be competing with Ryzen 3.

Edit: oh... and Ryzen 3 will be 12 core mainstream, or possibly 16 core mainstream.

Stick a fork in Intel.

 

[eddiew]

If AMD pull out a 5ghz base clock next year I will be delighted - never a bad thing to have an upgrade option, regardless of whether you feel the need to immediately switch to it at the time

Whether they can actually cool 12 cores at 5ghz even with 7nm... remains to be seen. Would believe it if it's down around the 1.1vcore mark, which doesn't really feel impossible but I don't think there's any comment on as yet?

 

[ljt]

Vs Coffeelake

Ryzen 1### had a few % less IPC using just one core, the same IPC using multiple cores
Ryzen 2### has the same IPC using one core, a few % higher IPC using multiple cores

If Ryzen 3 gets a similar 4% IPC bump it will be significantly ahead of Coffeelake, add to that for 7nm Global Foundries are citing 40% higher performance vs 14nm, that's 40% higher clock speeds than Ryzen 1###, so if we take the 1800X at 3.6Ghz and add 40% you get 5Ghz.

That would make Coffeelake well and truly beat, Intel have no answer for Ryzen 3, their 10nm is way off, its why they are rebranding 14nm Coffeelake, the 9000 series is what will be competing with Ryzen 3.

Edit: oh... and Ryzen 3 will be 12 core mainstream, or possibly 16 core mainstream.

Stick a fork in Intel.

I think AMD will be within a knats whisker of Intel come Ryzen 3000 series, but I just can't see them beating them with clock speed.

Not least forgetting that even if AMD somehow do manage 5ghz, their mesh will still be slower than Intel ring bus, clock for clock, especially in gaming.

We can see that now in those clock for clock benchmarks.

 

[humbug]

I think AMD will be within a knats whisker of Intel come Ryzen 3000 series, but I just can't see them beating them with clock speed.

Not least forgetting that even if AMD somehow do manage 5ghz, their mesh will still be slower than Intel ring bus, clock for clock, especially in gaming.

We can see that now in those clock for clock benchmarks.

I think you're probably right with clock speed, but i do think into the mid to high 4Ghz next time round.

As for Infinity Fabric, as i said before there are outliers of old games that don't understand Ryzen at all but overall actually AMD's Infinity Fabric is proving not to be a hindrance to its performance.

This actually is pretty typical and when you consider there is no difference in gaming between the 6 core and 8 core Ryzen you realise just how good it actually is, the 8700K is clocked 19% higher and the performance overall in this 35 Game review is 15% to the 8700K.

Again this is pretty typical. its 15% for 19% higher clocks on a game that's not highly threaded, we all know that about Arma III. its very typical of the overall result of this 35 game review, 15% or a little less to coffeelake with 19% higher clock speed.



Imagine that ^^^^ with Ryzen 3 at 4.7Ghz and another 4% IPC.
Yes then you get a couple of games that look like this:
This is a very old game built solely for Intel.



And at its best:



Apart from a couple of outliers and cherry picking while ignoring others... i just don't see that clock for clock performance deficit, i really don't.

Intel enjoy a clock speed advantage, but that really is it.

https://www.techspot.com/review/1655-core-i7-8700k-vs-ryzen-7-2700x/

 

[ljt]

I was thinking more of the techspot article where they did 8700k/8600k/2600x/2700x all at 4ghz.

Core for core i.e. a 2600x @ 4ghz vs a 8700k @ 4ghz, showed around a 7-10% lead for the Intel chip in gaming. Which would be an indication even if both chips were clocked @ 5ghz. That would simply be down to a slight IPC advantage for Intel and the ring bus architecture.

If AMD can pull a decent IPC uplift and a significant frequency uplift with the 3000 series, then it'll be very close. I'm just hoping the 3000 series will work to full effect in my X370 motherboard!

 

[humbug]

I was thinking more of the techspot article where they did 8700k/8600k/2600x/2700x all at 4ghz.

Core for core i.e. a 2600x @ 4ghz vs a 8700k @ 4ghz, showed around a 7-10% lead for the Intel chip in gaming. Which would be an indication even if both chips were clocked @ 5ghz. That would simply be down to a slight IPC advantage for Intel and the ring bus architecture.

If AMD can pull a decent IPC uplift and a significant frequency uplift with the 3000 series, then it'll be very close. I'm just hoping the 3000 series will work to full effect in my X370 motherboard!

You're talking about this one https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/page3.html

In that one they only used 5 games and only one of them shows a real difference between them at 4Ghz, Farcry 5, you see the same in the 35 game benchmark, i think to ignore the 35 game one and propose a different one withy only 5 is cherry picking a little because it is much much more constrained.
But again like that review mostly there is little in it, the difference is this one is missing the games where Ryzen is clock for clock faster, like Dirt 4 and For Honour.... again only if you ignore those can you make that argument.

I think if anything if you look at these reviews side by side what you find is Ryzen scales better with higher clock speed than Coffeelake, 4Ghz vs 4Ghz it is slightly faster in games that we know from the 35 Game review yes Coffeelake is slightly faster in those, but turning the clock rate 20% higher than Ryzen that difference increases only marginally.

That i find really interesting, its like Coffeelake is hitting much more of a Mhz scaling wall than Ryzen.

 

[Panos]

You're talking about this one https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/page3.html

In that one they only used 5 games and only one of them shows a real difference between them at 4Ghz, Farcry 5, you see the same in the 35 game benchmark, i think to ignore the 35 game one and propose a different one withy only 5 is cherry picking a little because it is much much more constrained.
But again like that review mostly there is little in it, the difference is this one is missing the games where Ryzen is clock for clock faster, like Dirt 4 and For Honour.... again only if you ignore those can you make that argument.

I think if anything if you look at these reviews side by side what you find is Ryzen scales better with higher clock speed than Coffeelake, 4Ghz vs 4Ghz it is slightly faster in games that we know from the 35 Game review yes Coffeelake is slightly faster in those, but turning the clock rate 20% higher than Ryzen that difference increases only marginally.

That i find really interesting, its like Coffeelake is hitting much more of a Mhz scaling wall than Ryzen.

And 20% more power consumption.... Lets not forget that...

Intel already consumes well more power than Ryzen, and well in excess of the performance difference.

 

[humbug]

True ^^^^

 

[mmj_uk]

Yup it makes me laugh when all the nvidia fanboys go on about AMD having dubious power usage... which is kind of true until you start undervolting vega which AMD should have bloody well done in the first place.
However its a laugh because most of them are running a stupid amount of power through there cpu for next to no benefit... unless you really can feel a massive difference between 4.9 and 5.1 Ghz

If AMD didn't have to factory overclock their parts in order to be competitive they wouldn't need to use excessive voltages, if all Ryzen shipped at 3.6ghz they could probably knock 40% or more off the voltage. Both Intel and NVidia components generally have lots of overclocking headroom because they aren't configured out of the box at the very limit of the silicon.

 

[humbug]

If AMD didn't have to factory overclock their parts in order to be competitive they wouldn't need to use excessive voltages, if all Ryzen shipped at 3.6ghz they could probably knock 40% or more off the voltage. Both Intel and NVidia components generally have lots of overclocking headroom because they aren't configured out of the box at the very limit of the silicon.

Interesting you should say that when an 8700K, a 6 core, uses significantly more power than the 2700X. A higher performance 8 core.

 

[Panos]

Interesting you should say that when an 8700K, a 6 core, uses significantly more power than the 2700X. A higher performance 8 core.

 

[ljt]

You're talking about this one https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/page3.html

In that one they only used 5 games and only one of them shows a real difference between them at 4Ghz, Farcry 5, you see the same in the 35 game benchmark, i think to ignore the 35 game one and propose a different one withy only 5 is cherry picking a little because it is much much more constrained.
But again like that review mostly there is little in it, the difference is this one is missing the games where Ryzen is clock for clock faster, like Dirt 4 and For Honour.... again only if you ignore those can you make that argument.

I think if anything if you look at these reviews side by side what you find is Ryzen scales better with higher clock speed than Coffeelake, 4Ghz vs 4Ghz it is slightly faster in games that we know from the 35 Game review yes Coffeelake is slightly faster in those, but turning the clock rate 20% higher than Ryzen that difference increases only marginally.

That i find really interesting, its like Coffeelake is hitting much more of a Mhz scaling wall than Ryzen.

That was the article, I wasn't cherry picking, it's the only place that seems to have done a clock for clock comparison, I didn't have much else to go on. It would be useful if somewhere had done a more thorough game comparison

 

[Panos]

Interesting you should say that when an 8700K, a 6 core, uses significantly more power than the 2700X. A higher performance 8 core.

I don't think we will receive a reply

 

[humbug]

Nope ^^^^

Thats Fair enough @ljt itake the point

 

[sideways14a]

If AMD didn't have to factory overclock their parts in order to be competitive they wouldn't need to use excessive voltages, if all Ryzen shipped at 3.6ghz they could probably knock 40% or more off the voltage. Both Intel and NVidia components generally have lots of overclocking headroom because they aren't configured out of the box at the very limit of the silicon.

If ryzen ran at 3.6 when it came out it wouldnt have made such a big impression.
Now...... if Vega had been binned and many of them undervolted to run at 1600Mhz like a lot can.... well thats a different moan.

 

[smilingcrow]

Interesting you should say that when an 8700K, a 6 core, uses significantly more power than the 2700X. A higher performance 8 core.

The problem is that power figures tend to vary a lot depending on the review you read which is understandable as it can vary purely based on the software you are running.
So unless someone links to actual reviews I don't take it seriously.
I just looked at 2 reviews and got the info below.
These were the 1st two I looked at with power figures and I don't pretend to know if they are indicative as I prefer not to believe my own hype.

TechReport

Techpowerup

 

[4K8KW10]

The problem is that power figures tend to vary a lot depending on the review you read which is understandable as it can vary purely based on the software you are running.
So unless someone links to actual reviews I don't take it seriously.
I just looked at 2 reviews and got the info below.
These were the 1st two I looked at with power figures and I don't pretend to know if they are indicative as I prefer not to believe my own hype.

Ryzen 7 2700X has a TDP 105W, look at the pure 2700 and 1700 which are 65W parts.