Tuesday, October 17th 2017

AMD Zen 2 Offers a 13% IPC Gain over Zen+, 16% over Zen 1

AMD "Zen" CPU architecture brought the company back to competitive relevance in the processor market. It got an incremental update in the form of "Zen+" which saw the implementation of an improved 12 nm process, and improved multi-core boosting algorithm, along with improvements to the cache subsystem. AMD is banking on Zen 2 to not only add IPC (instructions per clock) improvements; but also a new round of core-count increases. Bits n Chips has information that Zen 2 is making significant IPC gains.

According to the Italian tech publication, we could expect Zen 2 IPC gains of 13 percent over Zen+, which in turn posted 2-5% IPC gains over the original Zen. Bits n Chips notes that these IPC gains were tested in scientific tasks, and not in gaming. There is no gaming performance data at the moment. AMD is expected to debut Zen 2 with its 2nd generation EPYC enterprise processors by the end of the year, built on the 7 nm silicon fabrication process. This roughly 16 percent IPC gain versus the original Zen, coupled with higher clocks, and possibly more cores, could complete the value proposition of 2nd gen EPYC. Zen 2-based client-segment products can be expected only in 2019.
Source: Bits n Chips (Twitter)
Add your own comment

62 Comments on AMD Zen 2 Offers a 13% IPC Gain over Zen+, 16% over Zen 1

#2
dj-electric
+1
More woulds please, and less coulds.

Its nice dwelling in dreams about what could happen, market's reaction, the reaction to the reaction and an endless loop of market superiority fantasy.

AMD, time to bring the goods. We're waiting patiently.
Posted on Reply
#3
Deeveo
Hopefully the new process will allow higher clocks, with these speculated IPC gains it could add some nice prformance. We will see when the reviews are out I guess.
Posted on Reply
#4
noel_fs
Then it's only up to clock speed to compete with Intel at this point
Posted on Reply
#5
phill
I'm definitely looking forward to seeing the reviews on this :) Thank you!!
Posted on Reply
#6
FYFI13
noel_fs said:
Then it's only up to clock speed to compete with Intel at this point
That's from a discussion on Twitter:
Matan Mos - what clocks can we expect from 7nm?
it should be much higher than 14/12nm mobile processes
Bits And Chips - Not much higher
Posted on Reply
#7
lexluthermiester
This is going to be fun! Even if the gain is only 6% or 7%, that's enough to close the gap on Intel's single core IPC. AMD is very likely to continue forging ahead in advancement.
Posted on Reply
#8
B-Real
Seeing Intel prices, AMD wouldn't need a single IPC improvement as their CPUs are being sold like hell. If they really get these upgrades, even a part of the biased Intel owners will start thinking. Maybe.
Posted on Reply
#9
Imsochobo
noel_fs said:
Then it's only up to clock speed to compete with Intel at this point
if ipc is correct and they achieve 4.5 ghz there is no point in buying intel if price difference remains.
if they achieve 4.7-5ghz then sorry for intel and competetive market space.
Posted on Reply
#10
NC37
Imsochobo said:
if ipc is correct and they achieve 4.5 ghz there is no point in buying intel if price difference remains.
if they achieve 4.7-5ghz then sorry for intel and competetive market space.
Yep and Intel is very draconian towards adopting competitive pricing. Though I doubt they'll suffer much before they get their production problems settled. Even if AMD released a 32 core Ryzen that could do everything include wipe your butt after you poo, people would still be buying Intel. Their brand name is just too cemented. However, cement cracks and breaks. It takes time and bad management but eventually it breaks apart.
Posted on Reply
#12
SIGSEGV
no need to upgrade if it's turn out to be true.
wait...
I am confused with this part:
we could expect Zen 2 IPC gains of 13 percent over Zen+, which in turn posted 2-5% IPC gains over the original Zen..
Posted on Reply
#13
ShurikN
FYFI13 said:
That's from a discussion on Twitter:
The problem is, how much is "much", and how much "much" people want it to be.
There's a lot of talk on that twitter thread about clocks, but basically the first large cpu die to come out of TSMC will be Zen2 (and a server part at that), so there's no metric to predict max clocks (all or single core).

2700X has 3.7 base and 4.3 single core boost. With proper cooling it maintains 4GHz all core boost at all times.
Lets hypothetically bump the clocks by, lets say 300MHz across the board (and 300Mhz isn't really a lot considering a new process and it's also not that "much"). Couple that with large IPC gains, and most likely better memory support, and you got yourself a winner.
Posted on Reply
#14
GreiverBlade
good, good, now my retailer need to reflect the price drop on the 2700X and it will be an even better alternative that can compete quite well enough and my next upgrade path (well even a 2600X would do)


my 6600K turned into a 6600 (well with some mhz added but, bah...) ... i tried everything, it's not locked but it BSOD once you try to apply a little OC ... that's kinda a underhanded trick (nope i didn't change anything, compared to the time when it worked fine in OC ... and i paid quite enough for it, not totally since it was a replacement for a 4690K, to have a "permanent" OC'able and not a "time trial" OC CPU ... )


well i did wait quite a long time to see, again, being an excellent alternative, after a lot of K6-2, Athlon, Athlon XP, Athlon 64 and a FX6300 (which wasn't that bad ... ) i can finally get back to it (well i will let the end of year pass since it will be quite busy and kinda hard financially speaking)
Posted on Reply
#15
windwhirl
No hard data available and no products available yet. Take it with a little bit of salt. But, here's hoping AMD underestimated their own platform and pulls a 20% improvement.
Posted on Reply
#16
rvalencia
Metroid said:
Given this benchmark https://www.cpubenchmark.net/singleThread.html concerning singlethread, ryzen 2700x = 2190, so given 13%, +285 + 2190 = 2475, right now the second fastest is i7 8700k = 2704. Well this is only one test, gaming might be different.

https://www.cpubenchmark.net/cpu.php?cpu=AMD+Ryzen+7+2700X&id=3238
Add 30 percent for clock speed increase from TSMC's 7nm process tech.

https://www.anandtech.com/show/12677/tsmc-kicks-off-volume-production-of-7nm-chips
The 7 nm node is a big deal for the foundry industry in general and TSMC in particular. When compared to the CLN16FF+ technology (TSMC’s most widely used FinFET process technology) the CLN7FF will enable chip designers to shrink their die sizes by 70% (at the same transistor count), drop power consumption by 60%, or increase frequency by 30% (at the same complexity).
Posted on Reply
#17
First Strike
Although I personally welcome this change (damn, it seems I can finally consider Ryzen in my next computation rig), but scientific workload doesn't tell us anything about games. And people in the HPC field don't use the term 'IPC' for a reason. Why would some workload be called as 'scientific'? Because 1.they are usually pure FP, 2. they are MASSIVELY PARALLEL and extremely repetitive. So it stresses the FP pipeline in a streamlined way (good scientific code should always fill the pipeline without interupt it).

But that is nothing to do with game workload.
Posted on Reply
#18
R0H1T
dj-electric said:
+1
More woulds please, and less coulds.

Its nice dwelling in dreams about what could happen, market's reaction, the reaction to the reaction and an endless loop of market superiority fantasy.

AMD, time to bring the goods. We're waiting patiently.
The gains largely depend on the application, & optimization for Zen uarch. So the up to part will always apply, it's the same with Intel. You could think of it as a best case scenario wrt IPC btw the more important part to note will be the clock speeds. AMD can get very close to intel in MT tasks even with a 10~20% clock speed difference, thanks in large part to their SMT, so if they get closer to 5GHz OCed speeds I do think they can overtake Intel in most MT scenarios.
Posted on Reply
#19
qubit
Overclocked quantum bit
Doesn't this IPC improvement make AMD about equal to Intel for IPC performance? Maybe a little more?
Posted on Reply
#20
HTC
Apparently, this 13% extra IPC comes from tests in Epyc platform, which is why no game data is available as of yet: considering the jump in core count, the fact that it appears there are actual clock bumps @ all is quite remarkable. I was actually expecting the clock speeds on this platform to lower as the core count goes up, "as per norm".

Personally, i seriously doubt the IPC gains will be that high but maybe, just maybe i'm wrong: wouldn't mind that @ all :D

I said this in another thread:

HTC said:
What i wish for is for AMD to somehow manage to have the infinity fabric not directly tied to the RAM speed but be in a divider instead: this could bring extra speed to the IF, even with slower RAM. I'm thinking along the lines of an extra 1 / 8th or 1 / 6th speed VS RAM speed, via a divider of some sort. This would make IF run with 2400 MHz ram @ 2700-2800 MHz and with 3200 MHz RAM @ 3600-3733 MHz which, due to how RyZen currently works, would be quite a nice boost to performance, even if not changing anything else, VS current Zen +.
Still think this would be AMD's best bet. It would however force power usage up with the IF alone, possibly forcing to lower clocks a bit in order for it not to overshoot TDP but i think the newer 7nm process would more than compensate for that, though it could end up resulting in a lower overclock bump ceiling because of it. Still, the thought of being able to use much less expensive RAM while still having the "same IF speed / latency" is @ the very least drooling: think 2666 RAM with tight timings running @ either 2999 or 3110 MHz (via the divider: an extra 1 / 8 th and 1 / 6 th speed respectively). Currently, you need expensive RAM to take advantage of lower latency and this change would remove that, making the platform as a whole more affordable.
Posted on Reply
#22
ghazi
This "Bits n Chips" bloke has been caught talking out his ass on countless occasions. Why do his statements on Twitter merit such an article?
Posted on Reply
#23
HTC
Found this article interesting and concerning @ the same time.

No idea if any of it applies to Zen 2, @ this time.
Posted on Reply
#24
Mighty-Lu-Bu
The only thing I hope with Zen 2 is that AMD irons out the issues with gaming. I am using a Ryzen 7 1700X and make no mistake, compared to my old AMD FX CPU my gaming experience is pretty great, but I am still slightly disappointed that we are still a tad bit behind Intel when it comes to gaming.
Posted on Reply
#25
Durvelle27
Interesting

Higher clocks alone would close the gap but higher OC would put them slightly ahead with decent clocks
Posted on Reply
Add your own comment