AMD RyZen CPU Architecture for 2017

Very interesting indeed. Even Ryzen 1700 is 12% faster than 6900K. Can't be explained by clock rate advantage. Ryzen has 2x larger L2 caches, but Intel's L3 is bigger and faster than Ryzen's, so I doubt it matters much.
Yes but I;d like them to use 3200 memory since it would increase performance on ryzen more than on Intels.

Also I think the R5 use 2-2 CCX config to avoid the advantage that the 1400 would have over the 1500.(the 1500 would need to be 2+2 to have access to the other 8BM of L3 while the 1400 could be 4-0 but then have an advantage of not having the penalty of the CCX interconnect. Test of the R7 on 2-2 and 4-0 vs R5 would really interesting to see.
 
Ah! I did not know about Snowy Owl. That does seem like a sound way to go. By the time Coffee Lake is out and the dust has settled, top Ryzen 7 SKUs will probably sell for somewhere around $350, leaving a nice $500–$1000 space for Snowy Owl, which could be called Ryzen 9, or Ryzen X or whatever. A couple of 12- and a couple of 16-core SKUs would fill that space quite nicely.
I would think so too. But keep in mind it isn't officially confirmed, so still rumor territory.
 
Deus Ex is very interesting. I can see how a multi-threaded game might like Ryzen, but the latter is way ahead of even the i7-6900K
Very interesting indeed. Even Ryzen 1700 is 12% faster than 6900K. Can't be explained by clock rate advantage. Ryzen has 2x larger L2 caches, but Intel's L3 is bigger and faster than Ryzen's, so I doubt it matters much.
It could have something to do with Toms testing the game at the Very High preset, which is a step down from Ultra, it removes contact shadows, some of the lighting and reflections. Other media outlets tested Deus Ex @ Ultra and most Intel CPUs came out on top by a decent margin.
http://techreport.com/review/31366/amd-ryzen-7-1800x-ryzen-7-1700x-and-ryzen-7-1700-cpus-reviewed/9
http://www.hardwarecanucks.com/foru...-amd-ryzen-7-1800x-performance-review-16.html
https://www.computerbase.de/2017-03.../4/#diagramm-deus-ex-mankind-divided-dx12-fps
http://pclab.pl/art73169-17.html
 
Last edited:
That'd be because I failed to refer back to original quote & thought the core count referred specifically to the APU :oops:
But if AMD needs higher core counts next year bear in mind Ryzen/CCX architecture is already intended to scale to 16 & 32 core chips even if not (currently) planned for desktop.

Yeah, that's why I think it would relatively easy to do, and a pretty good idea too. APUs probably ought to stick to 4 cores for now, at least until AMD can afford to make separate dies (with 6 and 4 cores) otherwise the CPU cores would take too much silicon area away from graphics, and add too much cost for most laptops.
 
That are rumors, Intels philosophy is process -> Arq -> Optimization. SK is Opt. So cannon lake is process(same arq, different process) and coffe lake is Arq and the next will be Opt. So Intel will need to wait until Arq time again and doing to math that would be 5 years from now.
If we take into consideration the time that takes to make a CPU its pretty clear that CL and CL are already type out and everything is decided, Intel could tweak here and there but can't do much to them. So Until Intel magically(or prophetically?) wanted to make a huge jump with cofee lake years ago they will need to change the philosophy and launch a new Arq before is "schedule" right now.
According to Wikipedia Cannonlake is process (10nm), Icelake is architecture and Tigerlake is optimization.
Coffeelake is a 14nm stopgap since Intel is having trouble with 10nm process transition.
I think Ryzen is a nice CPU that will force Intel to finally lower its prices, but I don't think that Intel desperately needs a brand new architecture in order to compete with AMD.
If it took AMD half a decade to go from Bulldozer to Ryzen, I find it hard to believe that it would take Intel more than a decade to go from Sandybridge to something really new. But then again, Bulldozer sucked while Sandybridge did not.
 
According to Wikipedia Cannonlake is process (10nm), Icelake is architecture and Tigerlake is optimization.
Coffeelake is a 14nm stopgap since Intel is having trouble with 10nm process transition.
I think Ryzen is a nice CPU that will force Intel to finally lower its prices, but I don't think that Intel desperately needs a brand new architecture in order to compete with AMD.
If it took AMD half a decade to go from Bulldozer to Ryzen, I find it hard to believe that it would take Intel more than a decade to go from Sandybridge to something really new. But then again, Bulldozer sucked while Sandybridge did not.

CPU design takes several years to be complete and ready to manufacture, the CPUs that Intel and AMD will launch in this 3 years are already in the pipeline of finish/debugging and optimization.Yes Intel could throw 10 billion of dollars to make a and sell a CPU in a year but then where is the gaining? you just can't throw money without worrying about how you gonna get that back.

As far as i know 10nm is right on track but it doesn't matter the names on it what matter is that this year Intel will make the jump to 10nm, then sell a new Arq, then optimize that arq for the new process, then make another jump in process and then make a new arq. this are 5 steps before getting a new Arq aim to fight in a world with ryzen(remember the new 2018 arq was finish a while ago and its being debugging and optimizing) and each step is a year long, so 5 years.

Intel will need to break this and launch a new arq w/o having to follow these steps because AMD will change arq every year and will eventually catch up(wich is not really far away, if AMD manage to make an improvement of 10 or 15% in IPC that would put ryzen on top of SK).
 
CPU design takes several years to be complete and ready to manufacture, the CPUs that Intel and AMD will launch in this 3 years are already in the pipeline of finish/debugging and optimization.Yes Intel could throw 10 billion of dollars to make a and sell a CPU in a year but then where is the gaining? you just can't throw money without worrying about how you gonna get that back.

As far as i know 10nm is right on track but it doesn't matter the names on it what matter is that this year Intel will make the jump to 10nm, then sell a new Arq, then optimize that arq for the new process, then make another jump in process and then make a new arq. this are 5 steps before getting a new Arq aim to fight in a world with ryzen(remember the new 2018 arq was finish a while ago and its being debugging and optimizing) and each step is a year long, so 5 years.

Intel will need to break this and launch a new arq w/o having to follow these steps because AMD will change arq every year and will eventually catch up(wich is not really far away, if AMD manage to make an improvement of 10 or 15% in IPC that would put ryzen on top of SK).
I know it takes several years to design and manufacture a high-end CPU, which means that Intel already have its brand new architecture(s) almost ready at this time, and AFAIK it's Icelake.
You seem to imply that Intel just slept during all these years and suddenly decided to do a new architecture only after learning about Ryzen. In your second paragraph you state that Intel still needs two process shrinks and two new architectures in order to compete with Ryzen, which is just not true. And after saying that it takes Intel several years to design, test, manufacture and optimize a CPU, you then say that AMD will somehow "change arq every year".
You seem to be extremely optimistic about AMD and extremely pessimistic about Intel.
 
I get the impression that some people look at the processors Intel has been marketing and assume that's the best their architecture(s) can be binned to do. My thoughts are more along the lines of those products being extremely profitable and the yields are simply magical for them. AMD was able to take advantage of it to a degree. Intel's next product launch will certainly be interesting.
 
Last edited:
I get the impression that some people look at the processors Intel has been marketing and assume that's the best their architecture(s) can be binned to do. My thoughts are more along the lines of those products being extremely profitable and the yields are simply magical for them. AMD was able to take advantage of it to a degree. Intel's next product launch will certainly be interesting.

I doubt it , AMD has to claw back through a lot of bad years of negative press . I'm sure intel wont care for a few more years if even. They may move the 6 core chips down in price but I still am not sure on it .
 
I know it takes several years to design and manufacture a high-end CPU, which means that Intel already have its brand new architecture(s) almost ready at this time, and AFAIK it's Icelake.
You seem to imply that Intel just slept during all these years and suddenly decided to do a new architecture only after learning about Ryzen. In your second paragraph you state that Intel still needs two process shrinks and two new architectures in order to compete with Ryzen, which is just not true. And after saying that it takes Intel several years to design, test, manufacture and optimize a CPU, you then say that AMD will somehow "change arq every year".
You seem to be extremely optimistic about AMD and extremely pessimistic about Intel.

You need more comprehensive reading. Intel had a pipeline of CPU design which contemplates minimal increase in performance from one gen to another and unless Intel have some...Intel on what AMD was doing that pipeline didn't take into account that AMD suddenly came out with a BW-E level of performance so the CPUs that Intel was developing were not design to a significance boost in performance. Yes with time Intel can make that happen but my whole point was that with intel philosophy of launching CPU: process - arq- opt. Intel would force themselves to wait 5 years to make that happen and they they will have to take a step back to the tick-tock or even and AMD like tock-tock in order to remain on the top.

In AMD case from the beginning they commit to a new arq every year and that is while Zen2 and 3 are in different stages of developments while on Intel they were working on adapting CL to new process and optimizing it instead of working in a new arq with a significant jump in performance.

The next gen won't really be the problem, with the advantage that Intel have and if we take into account historical jumps we may see 5 to 7% difference, maybe more with the 10~% difference that SL already have we are talking about around 20% from Z1 and if we imagine Z2 to have 15% increase that would place AMD still slightly behind(taking into account that games wont fully use 8C or 16Ts) But if Intel keeps with its current launching and use that Arq and just "optimize" it gaining a 5% increase then AMD will catch up or even surpass their CPU. So Intel will have to change the way the develop CPUs to once again start the performance race with AMD because AMD won't stop running and trying to catch and win that race, whether or not they will win we can't know but they won't surrender this time.

And yes I do think intel felt asleep in their bed of dollars and forgot about AMD the proof is in the all the innovation that Intel has being bringing to the market(exactly, what innovation?) they were just making ridiculous jumps and increasing the pricess of their products so the new ones weren't replacing the old ones.
 
You need more comprehensive reading. Intel had a pipeline of CPU design which contemplates minimal increase in performance from one gen to another and unless Intel have some...Intel on what AMD was doing that pipeline didn't take into account that AMD suddenly came out with a BW-E level of performance so the CPUs that Intel was developing were not design to a significance boost in performance. Yes with time Intel can make that happen but my whole point was that with intel philosophy of launching CPU: process - arq- opt. Intel would force themselves to wait 5 years to make that happen and they they will have to take a step back to the tick-tock or even and AMD like tock-tock in order to remain on the top.

In AMD case from the beginning they commit to a new arq every year and that is while Zen2 and 3 are in different stages of developments while on Intel they were working on adapting CL to new process and optimizing it instead of working in a new arq with a significant jump in performance.

The next gen won't really be the problem, with the advantage that Intel have and if we take into account historical jumps we may see 5 to 7% difference, maybe more with the 10~% difference that SL already have we are talking about around 20% from Z1 and if we imagine Z2 to have 15% increase that would place AMD still slightly behind(taking into account that games wont fully use 8C or 16Ts) But if Intel keeps with its current launching and use that Arq and just "optimize" it gaining a 5% increase then AMD will catch up or even surpass their CPU. So Intel will have to change the way the develop CPUs to once again start the performance race with AMD because AMD won't stop running and trying to catch and win that race, whether or not they will win we can't know but they won't surrender this time.

And yes I do think intel felt asleep in their bed of dollars and forgot about AMD the proof is in the all the innovation that Intel has being bringing to the market(exactly, what innovation?) they were just making ridiculous jumps and increasing the pricess of their products so the new ones weren't replacing the old ones.

I think you need to calm down a bit.
 
the best their architecture(s) can be binned to do.
Indeed.
7700K is a decent example of what they can do & from my understanding last several generations have pretty much guaranteed OC to well over 4Ghz.
This is why I hope AMD can push out a high clocked quad.
 
You need more comprehensive reading. Intel had a pipeline of CPU design which contemplates minimal increase in performance from one gen to another and unless Intel have some...Intel on what AMD was doing that pipeline didn't take into account that AMD suddenly came out with a BW-E level of performance so the CPUs that Intel was developing were not design to a significance boost in performance. Yes with time Intel can make that happen but my whole point was that with intel philosophy of launching CPU: process - arq- opt. Intel would force themselves to wait 5 years to make that happen and they they will have to take a step back to the tick-tock or even and AMD like tock-tock in order to remain on the top.

In AMD case from the beginning they commit to a new arq every year and that is while Zen2 and 3 are in different stages of developments while on Intel they were working on adapting CL to new process and optimizing it instead of working in a new arq with a significant jump in performance.

The next gen won't really be the problem, with the advantage that Intel have and if we take into account historical jumps we may see 5 to 7% difference, maybe more with the 10~% difference that SL already have we are talking about around 20% from Z1 and if we imagine Z2 to have 15% increase that would place AMD still slightly behind(taking into account that games wont fully use 8C or 16Ts) But if Intel keeps with its current launching and use that Arq and just "optimize" it gaining a 5% increase then AMD will catch up or even surpass their CPU. So Intel will have to change the way the develop CPUs to once again start the performance race with AMD because AMD won't stop running and trying to catch and win that race, whether or not they will win we can't know but they won't surrender this time.

And yes I do think intel felt asleep in their bed of dollars and forgot about AMD the proof is in the all the innovation that Intel has being bringing to the market(exactly, what innovation?) they were just making ridiculous jumps and increasing the pricess of their products so the new ones weren't replacing the old ones.
Well, my english sucks as it's not my first language, so it's hard for me to discuss eloquently in english and sometimes I can be a bit blunt. For that I apologize, but my reading comprehension is perfect. Now back on topic.

Considering that the Bulldozer architecture sucked so hard, having worse single-thread performance at launch than its predecessor while using more power, it was obvious to Intel and everybody else that AMD would start designing and eventually release a truly new and better architecture than Bulldozer (aka Ryzen). But why did AMD keep optimizing the Bulldozer architecture (Piledriver and Excavator) instead of quickly releasing the truly new and better architecture in the next one or two years?
Because it's impossible to do otherwise, and that's why "Zen2" and "Zen3" are going to be optimizations of their current Ryzen architecture, as their names imply. And their performance improvements are unknown at this time, just like Intel's.

Right now Intel don't need a brand new architecture in order to counter Ryzen, they just need to reduce prices.
IMO, Intel's real problems are things like Dennard scaling (RIP), Moore's law (soon to RIP), the speed of light, cost vs benefit of new process, etc. Those are the things that will allow AMD and ARM designs to come closer and closer to Intel's, not some "launch philosophy" that is mostly marketing bullshit anyway.

But it's obviously better (for us) to have some market competition than to have none, as Intel is too accustomed to big fat margins at our expense.
 
No it wasn't obvious, ryzen is a jump that no1 expected that AMD could make, when was the last time you saw a cpu jump in performance of 52%? AMD is right now about 10% behind of Intel in arq. and even when its behind in process its still more efficient. Zen is a bran new arq with plenty of improvement to make so yes AMD have a base to fight against Intel.

Scaling is the same for everyone, intel is not that far way from the rest in process tech and everyone is looking at how to surpass the limit in the current silicon that is about 5nm and we will be making chip at 10nm this year so yes its a real problem but not just for intel but for everyone.
 
About the pricing of course Intel can lower them, but how do you think that negative growing will appear for investors? Intel needs to keep their CPU in a big lead over AMD to be able to keep their prices. When you see CPUs like the 1600 that will offer the level of performance of the 6800k for half the price you see there is a problem for Intel there.
 
One really hopes that AMD will increase single threaded performance.
If I was looking for a reasonably priced gaming CPU today (knowing about Ryzen 5) I would still buy the i5-7500 at its current price.
Now if I wanted a workstation CPU, I'd buy the Ryzen in a heartbeat.
And since AMD is smart, I expect that they've calculated that there is no use lowering their prices further – either they know they'll be making a lot on workstation users, or their supply is limited, or...
 
Why 5 years?
AFAIK Icelake is a brand new architecture for 10nm to arrive around 2019?
It might be, it might not be. According to Intel Sandy Bridge, Haswell and Skylake were "brand new architectures" too, while in reality they're all small evolutionary steps from Nehalem

Kind of a bummer. Does it make sense for potential buyer of ~$150 CPU to wait for true quad core? So far one can hope they will be cheaper, faster, clocked higher, more efficient...
Most likely the only "true quad cores" you'll see will be Raven Ridges and possible variants with GPU fused off but still on the chip
 
It might be, it might not be. According to Intel Sandy Bridge, Haswell and Skylake were "brand new architectures" too, while in reality they're all small evolutionary steps from Nehalem


Most likely the only "true quad cores" you'll see will be Raven Ridges and possible variants with GPU fused off but still on the chip

Most likely apparently it was cheaper and easier to focus into making a single chip than 2 or 3 variants. But I really don't agreed in the 2-2 config, using the 2-2 and having lower clock speed than the 8 core is like asking users to but the 8 core instead(maybe that was the intention). losing cache hurts but interccx latency hurts more, unless AMD is confident that game optimization will get ride of this. But since this 6 and 4 core are 8 cores I can see AMD wanting to offer a full range of products but also wanting to sell more 8 cores than 6 or 4.

There is a rumor of an "confirm and reliable source" that says AMD fix that for the silicon of their server parts. this is see to believe.

But it's obviously better (for us) to have some market competition than to have none, as Intel is too accustomed to big fat margins at our expense.

Sorry my mother is sick and sometimes I feel very stressed and let that out, didn't pretend to be offensive. I'm here to argue with people with brain and knowledge that actually knows what they are talking about and use facts instead of opinions not to fight about who is right and who is wrong, again sorry if I was too aggressive. .
 
There is a rumor of an "confirm and reliable source" that says AMD fix that for the silicon of their server parts. this is see to believe.
As it stands now, Naples should be Zeppelin MCM's, so exact same issues as Ryzen going CCX to CCX, and more due added latency from going chip to chip in the MCM-package
 
Back
Top