AMD Zen 3 (5000 Series), rumored 17% IPC gain.

[humbug]

I'm glad you brought this argument out. Dg has lost the plot.

CPU's do not work the way people think. The ZEN architecture will grab hundreds of opcodes and operands (instructions), figure out the dependency graph, do some branch prediction and then execute them out of order, several concurrently. You then stick the results all back together, maybe throwing away some of the incorrectly predicted branches. This is done to avoid the limitations of RAM and to increase parallelism without the coder needing to know the architecture in question. If you need to go back to RAM, it is REALLY slow, but again, executing tons of stuff allows this to proceed without the whole pipeline stalling for _ages_.

The above is the reason behind current CPUs having IPC's of non-integers, e.g. 3.2 for Zen 2 vs. 3.6 for Skylake in some workload [https://lemire.me/blog/2019/12/05/instructions-per-cycle-amd-versus-intel/]

So clocking the processor higher gives you linear scaling of processing, but improving your branch prediction and caching strategy can have much bigger impacts to your IPC.

TL;DR: Your processor isn't neat and tidy, it does things massively out of order and unpredictably. Clocking your processor higher nets you linear IPC scaling at best. Improving your deeply out-of-order execution strategy (read architecture), can net you non-linear IPC gains.

That article is a little odd, he's citing IPC comparison benchmarks that show a higher IPC on the Zen 2 architecture and then making the argument that his own little code that he wrote proves its not actually true, that in fact his benchmark code shows that Skylake actually has 50% higher IPC (2.1 vs 1.4)

Look, i don't pretend to understand any of this but much more often than not in any actual Desktop application workload Zen 2 is consistently 10 to 15% ahead clock for clock, i'm a little bit suspicious when someone comes along (academic or not) and says "look, despite the end result in the real world showing X result, if you use this software that i made here i can prove it all wrong"

Its a bit like Intel saying "ignore all of that software you actually use on your PC for your work and hobbies, instead here is this synthetic benchmarking tool, that we are not going to tell you we actually own the company who made it, and if you run that it shows that actually we are better"

Which is exactly what Intel have been doing since time immemorial but especially now, again.

He nor Intel today can ever explain away how it is the real world says one thing while they say another, he has been questioned about this in the comments and has basically ignored it.

 

[Slinky69]

Sorry to be the Grammar Nazi, but it's "time immemorial".

Please don't hate me

 

[Mr Evil]

That article is a little odd, he's citing IPC comparison benchmarks that show a higher IPC on the Zen 2 architecture and then making the argument that his own little code that he wrote proves its not actually true, that in fact his benchmark code shows that Skylake actually has 50% higher IPC (2.1 vs 1.4)

Look, i don't pretend to understand any of this but much more often than not in any actual Desktop application workload Zen 2 is consistently 10 to 15% ahead clock for clock.....

This is because there are two definitions of IPC, which he does mention:

However, it is not clear whether these reports are genuinely based on measures of instruction per cycle. Rather it appears that they are measures of the amount of work done per unit of time normalized by processor frequency.

The former is of academic interest, while the latter is what people who want to actually use the CPU should look for. It would be nice if the two measures had different names, but they don't, so confusion will persist.

 

[humbug]

Sorry to be the Grammar Nazi, but it's "time immemorial".

Please don't hate me

I don't. thanks for that Edited.
------------------

@0h_no_not_again, here is another academic writing a paper on the Zen Architecture vs Skylake, in which he's lays out in great detail why the Zen architecture has higher IPC.

Now we have two academics who disagree with eachother, only this one at least can explain himself.

Its worth watching, it compares the inner workings of the architecture, in summery Zen does a lot more inside than Skylake / Coffeelake...they are all the same lake, you know, more work per clock cycle...

 

[humbug]

Anyway... i posted that ^^^ in the original Ryzen 3000 thread before launch as a prediction of what Zen 2 might become.

At the back end Zen does a lot more work than Skylake, that work is bottlenecked by the front end, my argument was AMD would look to improving the front end to unlock more of the backend, it seems that is exactly what they have done as they have pushed the IPC up by 15%

I thought this might happen because Intel's core architecture has changed little since the original Core series 1, they have been able to increase the IPC little by little over the years by improving the front end, at this point and for the last few generations there is no front end bottleneck which is why they are now relying on clocks.
It costs a huge amount of money and time to create a new architecture from the ground up, its why its rarely done. AMD had to.

AMD has some way to go yet before they completely remove the front end bottleneck, Zen 3 will unlock more of the back end, another 15%, perhaps.

 

[Trig]

So Intel are like Porsche with the 911 then, its pretty much the same as its always been but with tiny tweaks over the years..

 

[Dominic-Toretto]

Zen 1 would be a good upgrade.

Oh wow, so Zen 3 will be a huge upgrade! Excited for the release of these

 

[cliffy]

I'm glad you brought this argument out. Dg has lost the plot.

CPU's do not work the way people think. The ZEN architecture will grab hundreds of opcodes and operands (instructions), figure out the dependency graph, do some branch prediction and then execute them out of order, several concurrently. You then stick the results all back together, maybe throwing away some of the incorrectly predicted branches. This is done to avoid the limitations of RAM and to increase parallelism without the coder needing to know the architecture in question. If you need to go back to RAM, it is REALLY slow, but again, executing tons of stuff allows this to proceed without the whole pipeline stalling for _ages_.

The above is the reason behind current CPUs having IPC's of non-integers, e.g. 3.2 for Zen 2 vs. 3.6 for Skylake in some workload [https://lemire.me/blog/2019/12/05/instructions-per-cycle-amd-versus-intel/]

So clocking the processor higher gives you linear scaling of processing, but improving your branch prediction and caching strategy can have much bigger impacts to your IPC.

TL;DR: Your processor isn't neat and tidy, it does things massively out of order and unpredictably. Clocking your processor higher nets you linear IPC scaling at best. Improving your deeply out-of-order execution strategy (read architecture), can net you non-linear IPC gains.

So i got it right? Was trying to dumb it right down but was unsure if i achieved that

 

[~>Dg<~]

Oh wow, so Zen 3 will be a huge upgrade! Excited for the release of these

thats the hope for us all. cant wait to see if its true.

 

[FoxEye]

Zen 1 would be a good upgrade.

He said gaming. So unless you can post some benchmarks showing a Zen 1 CPU beating a 7600K @ 5.2GHz in *gaming* workloads, then you're just shilling for AMD as usual.

You know you don't even help AMD by spreading the kind of disinformation that routinely comes from your mouth.

Oh wow, so Zen 3 will be a huge upgrade! Excited for the release of these

And the poor bugger believed you... (about Zen 1 being faster in gaming).

Just wait for reviews and try not to believe anyone here in this thread. Too much bias. People like @jigger are so blinded by bias for AMD that they will argue black is white and white is black all day long. Just ignore them.

 

[CAT-THE-FIFTH]

He said gaming. So unless you can post some benchmarks showing a Zen 1 CPU beating a 7600K @ 5.2GHz in *gaming* workloads, then you're just shilling for AMD as usual.

https://www.youtube.com/watch?v=97sDKvMHd8c

The Core i5 7600K even in 2019,was starting to loose to a Zen1 based Ryzen 5 1600AF in a number of newer titles. In 5 games the Core i5 7600K lost,with 2 being huge losses(SOTTR and BFV),in two games the Ryzen 5 1600 slightly lost,and in one it lost a decent amount. That is a Core i5 7600K at 4.8GHZ against a 3.9GHZ Ryzen 5 1600.In fact both the HUB samples were slightly below average overclockers. The average Core i5 7600K overclock on HWBOT was 4.9GHZ and the average Ryzen 5 1600 overclock was 4GHZ.

The problem with the Core i5 7600K,is that if a game actually needs more than 4 threads performance falls off a cliff,with very poor minimums,even if averages look OK. Now consider that Zen+ is basically a rebadged Zen1 with some cache fixes,I suspect a Ryzen 5 2600X or Ryzen 7 2700X is going to generally push past a Core i5 7600K.

This is why I never recommended a 4C/4T Core i5 for the last 5 years,and always suggested people either get a Core i7 or a Xeon E3. A Core i7 7700K wouldn't have the same problems!

 

[0h_no_not_again]

So i got it right? Was trying to dumb it right down but was unsure if i achieved that

Yeah dude, you got it in right

The point is that IPC is really complex and changes per workload.

 

[0h_no_not_again]

I don't. thanks for that Edited.
------------------

@0h_no_not_again, here is another academic writing a paper on the Zen Architecture vs Skylake, in which he's lays out in great detail why the Zen architecture has higher IPC.

Now we have two academics who disagree with eachother, only this one at least can explain himself.

Its worth watching, it compares the inner workings of the architecture, in summery Zen does a lot more inside than Skylake / Coffeelake...they are all the same lake, you know, more work per clock cycle...

Oh yeah dude. I was dubious of that article, where the actualy IPC numbers for each uarch seemed a bit meh. What it got across is that 1 instruction is not executed every cycle per core, otherwise we'd get an IPC of 4 for a 4 core part, etc.

I was trying to say that more Hz doesn't always equal more performance, particularly if you're flushing pipelines constantly. Improving your architecture means less flushing of pipelines, higher IPC and more performance.

 

[GregI]

Guy who translated the RDNA 2 leaks an hour or two ago has rumors for Zen 3 and it's not good news.

Right now AMD will not sell the 10 core (mentioned in the RDNA 2 leaks). It's going to be 8 12 and 16 core on launch. The 5800x will be the best chip for gaming but when the 5850x (10 core) does launch it will be VERY close in price to the 8 core part as it apparently costs less to produce. Limited stock of 5800x as AMD moves the good binned stuff for epyc and threadripper since margins are much better. Capacity is low and demand high so they are increasing prices.

I guess we will be really lucky to see £299 for the 8 core. Be prepared for £350 and if it's really bad... £375.

 

[Richdog]

Guy who translated the RDNA 2 leaks an hour or two ago has rumors for Zen 3 and it's not good news.

Right now AMD will not sell the 10 core (mentioned in the RDNA 2 leaks). It's going to be 8 12 and 16 core on launch. The 5800x will be the best chip for gaming but when the 5850x (10 core) does launch it will be VERY close in price to the 8 core part as it apparently costs less to produce. Limited stock of 5800x as AMD moves the good binned stuff for epyc and threadripper since margins are much better. Capacity is low and demand high so they are increasing prices.

I guess we will be really lucky to see £299 for the 8 core. Be prepared for £350 and if it's really bad... £375.

Please always provide links to sources with posts like this.

 

[Troezar]

Guy who translated the RDNA 2 leaks an hour or two ago has rumors for Zen 3 and it's not good news.

Right now AMD will not sell the 10 core (mentioned in the RDNA 2 leaks). It's going to be 8 12 and 16 core on launch. The 5800x will be the best chip for gaming but when the 5850x (10 core) does launch it will be VERY close in price to the 8 core part as it apparently costs less to produce. Limited stock of 5800x as AMD moves the good binned stuff for epyc and threadripper since margins are much better. Capacity is low and demand high so they are increasing prices.

I guess we will be really lucky to see £299 for the 8 core. Be prepared for £350 and if it's really bad... £375.

All sounds promising. Prices and availability will be fine by the time I move over. My 3700X is fine until AM5 drops. Then pop in the best upgrade until it becomes a bottleneck.

 

[michty_me]

All sounds promising. Prices and availability will be fine by the time I move over. My 3700X is fine until AM5 drops. Then pop in the best upgrade until it becomes a bottleneck.

Indeed!!

 

[Participant]

What motherboards / chipsets are we expecting with the Zen 3 release? Or is X570 still good for the next CPUs?

 

[grumpybeard]

What motherboards / chipsets are we expecting with the Zen 3 release? Or is X570 still good for the next CPUs?

Was here to ask same. X570 still the top tier platform or something else coming? Itching to upgrade from my 5820K.

 

[jigger]

He said gaming. So unless you can post some benchmarks showing a Zen 1 CPU beating a 7600K @ 5.2GHz in *gaming* workloads, then you're just shilling for AMD as usual.

You know you don't even help AMD by spreading the kind of disinformation that routinely comes from your mouth.


And the poor bugger believed you... (about Zen 1 being faster in gaming).

Just wait for reviews and try not to believe anyone here in this thread. Too much bias. People like @jigger are so blinded by bias for AMD that they will argue black is white and white is black all day long. Just ignore them.

You’ve clearly never used the two.

I would ignore this chap, he’s upset because he’s been pulled up on this nonsense.