Leaked Patch Confirms AMD Zen Will Have 32 Cores Per Socket?

Nope I am sure that SMT is a different implementation then what Intel calls Hyper Threading. What you are saying that there enough commonalities to blur them. I am sure that you can find yourself a few examples where they discuss the difference between SMT and HT that they approximately do the same job is something we both know.
Usually processors are designated as cores and threads :)

Er, no. Simultaneous multithreading (SMT) and hyperthreading (HT) are exactly the same thing.

www.agner.org/optimize/blog/read.php?i=6
Agner Fog said:
Some processors take multithreading even further by running two threads in each core. This is what Intel calls hyperthreading (also called simultaneous multithreading).
 
That's why I always stop to keep answering when someone say HT isn't the same as SMT.. but thanks Tonkatsu that just back my early post. some people just think AMD is the better and greater and what everything what AMD make is the best and different from intel so not matter if they are the same thing just because AMD implemented it it will be better, when the fact is totally opposed, this is specially applicable to this case regarding the improved "Hyper-threading" used by intel with skylake with everything I said in my early post in this topic.

Intel have improved Hyper-threading 3 times since the Pentium 4 HT Failure, the first with nehalem, the second one with Ivy Bridge and the third with Skylake.
 
That's why I always stop to keep answering when someone say HT isn't the same as SMT.. but thanks Tonkatsu that just back my early post. some people just think AMD is the better and greater and what everything what AMD make is the best and different from intel so not matter if they are the same thing just because AMD implemented it it will be better, when the fact is totally opposed, this is specially applicable to this case regarding the improved "Hyper-threading" used by intel with skylake with everything I said in my early post in this topic.

Intel have improved Hyper-threading 3 times since the Pentium 4 HT Failure, the first with nehalem, the second one with Ivy Bridge and the third with Skylake.

The problem with pentium 4 was the massive branch prediction miss penalty. There was little or nothing wrong with the P4 hyper-thread implementation. Good use of hyper-threading was able to hide some of the issues with branch prediction misses, which resulted in insanely high performance gains when hyper-threading was optimized in P4 designs. This was part of the reason, some apps that used Hyperthreading were awesome on the P4 otherwise AMDs solutions, at that time, were superior most everywhere else.
 
as that report state, we have already that kind of performance with Haswell-E, more important thing is, at what price will those 8c/16t will be launched.. specially with the consideration of the low price of the actual 5820K and its future replacement with Broadwell-E.
 
The problem with pentium 4 was the massive branch prediction miss penalty. There was little or nothing wrong with the P4 hyper-thread implementation. Good use of hyper-threading was able to hide some of the issues with branch prediction misses, which resulted in insanely high performance gains when hyper-threading was optimized in P4 designs. This was part of the reason, some apps that used Hyperthreading were awesome on the P4 otherwise AMDs solutions, at that time, were superior most everywhere else.

well yeah isn't about only performance, but efficiency, the high resource cost, and specially the temperature, those chips were hot as hell at the time... hyper-threading at that time was inviable for that reason, contrary to the re-implementation with nehalem..
 

There is something funny ;) SemiAccurate Forums - View Single Post - Official AMD Zen uarchitecture thread

So this was not something serious about double the performance :

muahahah, echo chamber picked up on my joke about Zen being twice as fast as Orochi. Even bigger problem is that they assume that Orochi is 8350, and not 8150
biggrin.gif

I only was completely joking about counting pixels to estimate performance advantage

SemiAccurate Forums - View Single Post - Official AMD Zen uarchitecture thread

My estimate:

If true that Zen core is 40% faster than EX core then Zen is ~60% faster than PD core in ST tasks.

Come MT workloads 8C/16T Zen is 1.6(IPC) x 1.25(SMT) / 0.8(CMT penalty that PD has) = 2.5x faster than 8350 at the same clock. Since the same clock will not happen we have to adjust for clock difference. I expect that AMD will manage to hit 3.3Ghz base and 3.7Ghz for turbo (ST/low thread workloads).

So with clock adjustment , 8C/16T 3.3/3.7Ghz Summit Ridge could be 2.5x3.3/4~=2 times faster than 8350, stock Vs stock. In ST tasks it could be 1.6x3.7/4.2~=1.4 or 40% faster than 8350, stock Vs stock.

This does not sound that far fetched ....
 
Yes, that's why I just left a link without a comment. It sounded far fetched, but it's a new rumor so best to add it to the rumor thread. ;)
 
Strange....those CPU guesses and IPC outcomes....... Look exactly like what I posted earlier. :whistle:
 
Strange....those CPU guesses and IPC outcomes....... Look exactly like what I posted earlier. :whistle:

I'm still waiting on the good news where we finally get to hear what is happening with Zen it seems like such a long time ago when Jim Keller left and they days are counting down to Q4 but little to nothing is revealed through AMD. Something as amount of cores and maybe price indication would be nice , it is not that anyone at Intel will lose sleep over it ....
 
This thing is going to come in just way too late for anybody to care. It'll be old tech before it hits the market.
 
I'm still waiting on the good news where we finally get to hear what is happening with Zen it seems like such a long time ago when Jim Keller left and they days are counting down to Q4 but little to nothing is revealed through AMD. Something as amount of cores and maybe price indication would be nice , it is not that anyone at Intel will lose sleep over it ....


You need someone to tell you that Zen is coming in 2, 4, 6 and 8 cores for the desktop? (not including SMT) "AMDs version of hyperthreading"
Server CPUs will arrive later with up to 16 cores. two dies on one chip, under one heat spreader just like the pentium D.

But this is old news... pricing is the BIG unknown.
 
You need someone to tell you that Zen is coming in 2, 4, 6 and 8 cores for the desktop? (not including SMT) "AMDs version of hyperthreading"
Server CPUs will arrive later with up to 16 cores. two dies on one chip, under one heat spreader just like the pentium D.But this is old news... pricing is the BIG unknown.
I don't see anything but 8 cores and the lower core chips prolly come after the Zen APU made an appearance.

This thing is going to come in just way too late for anybody to care. It'll be old tech before it hits the market.
I promise you one day you might actually make a post that might make sense keep trying ;) .
 
actually jwacalla it's right, zen may bring haswell performance, great right? that's 2013 intel technology, so at the point Zen arrive to the market it will be like just buying a Z97platform with a 4790K that's just if also AMD is able to reach 4790K clocks... what will be the point, at that moment kaby lake will be also in the market, so intel will still have 3 CPU architecture ahead of that same "haswell" like performance.

I see AMD guys always saying, Intel quad cores + HT are POS, they are a sidegrade at best, even skylake i7 offer nothing to me to make me sidegrade from my bulldozer truly octa core.. then zen will arrive with 2013 intel-like performance and I pretty sure every AMD fan will scream that zen is the best CPU ever made, that they now have a truly reason to upgrade and so on..

And im just keep thinking.. the initial release of Zen will be 4 cores and 4 cores + SMT (HT).. the best scenario include 6 cores and 6 cores + SMT...
 
actually jwacalla it's right, zen may bring haswell performance, great right? that's 2013 intel technology, so at the point Zen arrive to the market it will be like just buying a Z97platform with a 4790K that's just if also AMD is able to reach 4790K clocks... what will be the point, at that moment kaby lake will be also in the market, so intel will still have 3 CPU architecture ahead of that same "haswell" like performance.
I see AMD guys always saying, Intel quad cores + HT are POS, they are a sidegrade at best, even skylake i7 offer nothing to me to make me sidegrade from my bulldozer truly octa core.. then zen will arrive with 2013 intel-like performance and I pretty sure every AMD fan will scream that zen is the best CPU ever made, that they now have a truly reason to upgrade and so on..
And im just keep thinking.. the initial release of Zen will be 4 cores and 4 cores + SMT (HT).. the best scenario include 6 cores and 6 cores + SMT...

Is this a serious post ?
Why don't you link each and every single post you can find describing "Intel 4 cores + HT is shit". Or do you want me to go over to the Intel forum and make some of those I can do that I don't have anything better to do really ...
 
as that report state, we have already that kind of performance with Haswell-E, more important thing is, at what price will those 8c/16t will be launched.. specially with the consideration of the low price of the actual 5820K and its future replacement with Broadwell-E.

It is true that there are alerady performance from Intel, but at the end of the day, it is all about pricing. If it is price right, it can steal some market shares back from Intel if the performance is there.
You need someone to tell you that Zen is coming in 2, 4, 6 and 8 cores for the desktop? (not including SMT) "AMDs version of hyperthreading"
Server CPUs will arrive later with up to 16 cores. two dies on one chip, under one heat spreader just like the pentium D.

But this is old news... pricing is the BIG unknown.

Indeed, if it is price correctly, then AMD will definitely gain marketshare again. Price anywhere near Intel offering, there will be little traction.
 
what will be the point?

http://cpu.userbenchmark.com/Compare/Intel-Core-i7-6700K-vs-Intel-Core-i7-4790K/3502vs2384
http://cpu.userbenchmark.com/Compare/Intel-Core-i7-6700K-vs-Intel-Core-i7-4770K/3502vs1537

The point is quite clear isn't it? (unless Haswell is crap now because of it's "age")

Performance that will most likely match or exceed the top Haswell chips at reduced prices.

On a new, supported platform with all the bells and whistles i.e. arguably better than buying old socket 1155, 1150, AM3, FM2 etc

How can 8 or 16 cores at Haswell speeds (with SMT/HT) not equal awesome?


Maybe
there will be slim pickings for those of you who can afford the very best Broadwell-E, Xeon v3/v4 chips seemingly every year but Zen is promising for the rest of us mere mortals.

I've yet to read a post where anyone claimed they gained "significant life changing features" or got "bang for buck" upgrading an "old" i7 to a "new" i7 i.e. I needed a i7 6700K because my 2600K, 3770K, 4770K weren't good enough.

But I've seen plenty of posts to the contrary. And it's likely that Kaby Lake will be the usual 5% better than Skylake.


If AMD fail to either:
> Do the above
> Release a bad design
> Fail to improve on the (good) design with revisions over time

Then I'd agree with you - utterly pointless.

It'd be lovely for someone to reinvent the wheel and bypass the known limits of silicon but it's not gonna happen is it? Not by October any way.

I think plenty of people will make the jump from old or weaker AM3 & FM2 setups.

Likewise for anyone with weaker i3 or i5 chips that really wanted an i7 but couldn't due to cost.

Other than performance the only other concern I can see would be the AMD brand bogeyman; some people just won't touch them.
 
Last edited:
as that report state, we have already that kind of performance with Haswell-E, more important thing is, at what price will those 8c/16t will be launched.. specially with the consideration of the low price of the actual 5820K and its future replacement with Broadwell-E.

Yeah, the problem with the 5820K - however - is that it has a gimped number of PCIe lanes, which - to me - eliminates the entire reason for going with a -E platform. I certainly hope Zen comes with more lanes.
 
There is something funny ;) SemiAccurate Forums - View Single Post - Official AMD Zen uarchitecture thread

So this was not something serious about double the performance :



SemiAccurate Forums - View Single Post - Official AMD Zen uarchitecture thread



This does not sound that far fetched ....


I'll have to quote myself again :p

Here is my little linear interpolation based forward looking performance analysis from another thread (in which the OP was asking about video rendering performance)

I don't know what your video rendering looks like, but let's assume for a moment that it performs similarly to Cinebench 11.5:

A 6700k at max turbo of 4.2ghz scores 2.07 right now in the single core test. That's ~0.49/ghz.

A 7870k at max turbo of 4.1 scores 1.06 right now in the single core test. That's ~0.26/ghz.

So lets add the much talked about 40% to that figure. We get a predicted Zen performance of ~0.36/ghz. This is still only ~73.5% of the score of a Skylake core per ghz. Maybe AMD will be able to ramp up the clocks a little bit? Their current APU's are held back in clock because they have to use a GPU friendly manufacturing process. On pure FX chips this won't be an issue and maybe they'll be able to squeeze more clock speed out of them? If we assume this, and say maybe they can get a 4.8ghz turbo out of an FX based Zen, where does that leave us? Well, at 4.8Ghz, the predicted Zen score would be 1.74. This is a huge improvement, but it is still only 84% as fast as a current skylake.

That being said, a 1.74 score, ties Zen with a Haswell i7-4770K which is a respectable leap forward.

In other words, these calculations demonstrate what we have been guessing all along (or at least were, before the hype machine took over), that Zen will allow AMD to catch up to where Intel was a generation or two ago core for core. Being competitive with Haswell is actually better than I was expecting. I was thinking more along the lines of Ivy Bridge.

With your type of workloads (video rendering) you should scale very well with core count too, and if history is any judge AMD will release many core chips. Zen might just be the platform of choice for video rendering / encoding type workloads if they add many cores. (though I wouldn't count on consumer 32 core chips as have been rumored)

Now, this analysis is to be taken with TRUCKLOAD of salt as it is based on many assumptions and both educated and unedcuated guesses. Also, there is a lot of criticism of Cinebench as a benchmark, as it is believed to strongly favor Intel chips and not be a fair comparison. I used it because it is one of the few benchmarks that I can find results from that isolate single core results (as throwing in extra cores muddies the waters), and because you spoke of doing video rendering. The intent - however - is to put things in perspective, of what we can expect from Zen IF they live up to the official marketing slides from AMD.
 
Other than performance the only other concern I can see would be the AMD brand bogeyman; some people just won't touch them.

If they're selling yesterday's technology at half the price today, then you can't expect too many people to get excited and open their wallets. Good tech companies are expected to be on the cutting edge. All it does is hurt their brand even more.

It would be like if some ARM SoC vendor came out with an announcement today, "Look at our first, very-own shiny new 64-bit ARM chip!" Yeah, welcome to a few years ago.

All I'm saying is that the thing is coming in way too damn late, and like it or not, Intel is setting the standard for performance and features.
 
I think the point that most of us with AMD FX 8 cores are making is that the current FX is capable well enough within certain criteria (gaming 60hz-16:10). Doesn't mean better or that one shouldn't get Intel or should get AMD. But when Zen comes out and it does offer 40-60% performance increase and 8cores/16threads then yes for sure we will upgrade. There are more than enough stories proving that even Skylake was not an upgrade worth the investment over the existing CPU users had (within that aforementioned gaming 60hz) be it FX or 2500k or i7-7xx. Now if you work on your PC where time is money then you have no choice but get the latest greatest. But for the majority that just game or live on forums then sometimes the reasons to upgrade have precious little to do with what level of tech the current CPU offering is.
 
IMO Here's where Desktop is going to struggle.

They're going to be 600-800Mhz behind on the top turbo speed when compared to Intel's best.
They're probably going to be near 15-20% slower IPC than whatever Intel is releasing next spring
Intel's probably still going to be more efficient per Watt especially in the higher clock bin'd chips
Intel is well positioned to respond to almost ANYTHING AMD could release on the desktop


The interesting stuff is...
Mobile Zen has lots of potential to give intel massive heartburn.
 
IMO Here's where Desktop is going to struggle.

They're going to be 600-800Mhz behind on the top turbo speed when compared to Intel's best.
They're probably going to be near 15-20% slower IPC than whatever Intel is releasing next spring
Intel's probably still going to be more efficient per Watt especially in the higher clock bin'd chips
Intel is well positioned to respond to almost ANYTHING AMD could release on the desktop


The interesting stuff is...
Mobile Zen has lots of potential to give intel massive heartburn.
Maybe, maybe not. IPC isn't the end-all-be-all, but obviously more helps. The part that I see being the biggest pain is the APU moving forward. Sure Intel kept it close (but be honest here, Intel iGPU even the Irispro usually sucks for gaming). But they were at 16nm/10nm where AMD will now be at 14nm and able to pack more GPU horsepower under the hood rather than the 28nm iGPU they were at before with limited die space and TDP constraints.

Laptop positives galore.
 
The crew at TechFrag have been digging around in what is believed to be a leaked Linux patch and have come to the conclusion that AMD Zen based processors will feature up to 32 physical cores.

A leaked Linux patch on LKML.org, first spotted by The New Citavia Blog, suggests that AMD Zen based processors will feature up to 32 physical cores. The patch also hints at the similarity of parts of the “Zen” and “Zeppelin” codenames. The Zeppelin codename was first mentioned back in August last year, and parts of the patch identify it as a “family 17h model 00h” CPU.
Upper limit I'm sure. Especially in Server Class (operton?) CPUs they might hit that limit.
 
Glad to see they are going all out on the server chips. They've been out of the game so long on the server side they need to bring their A-game. The only thing that worries me is the 180w TDP I saw listed. That is a lot of heat.

However, Intel's current best is the 24 core Xeon E-7 8890 v4 at 165w TDP... so I guess 32 cores at 180w TDP is within expectations assuming these are running in the 2ghz range like the Intel chip.
 
I fail to see how Haswell level performance with 8 cores will be a failure for AMD with Zen. The last few chips from Intel have people complaining why should I upgrade for a 5% increase in performance. Also the latest prices for Intel chips is starting to get pricey. I think Zen will sell quite well if it lives up to the performance rumored so far.
 
I fail to see how Haswell level performance with 8 cores will be a failure for AMD with Zen. The last few chips from Intel have people complaining why should I upgrade for a 5% increase in performance. Also the latest prices for Intel chips is starting to get pricey. I think Zen will sell quite well if it lives up to the performance rumored so far.

Me as well.

Get me a desktop part with 8 cores that has IPC 80% of haswell, ecc (unbuffered) support & cheap motherboards (~ $100'ish) and I will happily replace my 1090T with 32G + ddr3 ecc without a second thought. I'd happily spend $500 pretty much immediately to upgrade.

$200 cpu, $100 board, (2*16GB) ddr4 ecc unbuffered $170.

Barring this, my next upgrade may be something will be one of those cheap TS140s or Dell's for $200-$250 that I can drop my e3-1240v3 into or possibly a Z420 or Z620 with 1 or 2 E5-2670s.
 
Last edited:
I fail to see how Haswell level performance with 8 cores will be a failure for AMD with Zen.

I think most people predicting doom and gloom don't believe AMD's performance estimates of "Haswell-like" performance. I don't trust them either, I'll believe it when I see it because AMD has broken a lot of promises recently.
 
I expect the 8 core / 16 threaded zen to be a > $400 CPU if it is competitive (its competition would be the 6 core / 12 threaded lga2011-3 CPUs and the lowest Skylake-e). If it is a $200 CPU I expect AMD (will be in no better condition than they are now) will go bankrupt.
 
I expect the 8 core / 16 threaded zen to be a > $400 CPU if it is competitive (its competition would be the 6 core / 12 threaded lga2011-3 CPUs and the lowest Skylake-e). If it is a $200 CPU I expect AMD (will be in no better condition than they are now) will go bankrupt.

You know AMD would be filthy stinking rich if they go 1 cent for each and every time people said this.
 
I fail to see how Haswell level performance with 8 cores will be a failure for AMD with Zen. The last few chips from Intel have people complaining why should I upgrade for a 5% increase in performance. Also the latest prices for Intel chips is starting to get pricey. I think Zen will sell quite well if it lives up to the performance rumored so far.

the success or failure of zen will depend entirely of prices.. we have already in the market "Haswell" level of performance since 3 years ago.. we have actually skylake which it's a ~20% average of performance above haswell.... Even if AMD truly reached all the estimations about IPC that will mean still not much if Zen can't reach Intel-like clocks which may still keep'em out of the competition, that's why most people think zen will still be a failure, because they are aiming 3 years old performance..

Only people that are going to buy Zen are the same people have AMD chips today, so very very little amount... they need to be appealing to intel users, they need to make people jump the fence and buy their zen processors, they need to convince intel people they have a competitive product with latest intel products which at that moment probably they will be faced against kabylake, if not then no matter what, it will be a failure, that's why prices are what can save truly AMD. .



What make you think Zen 8c/16t will cost 200$? that's almost absurd, 200$ will be about what a 4c/8t zen chip could cost..
 
What make you think Zen 8c/16t will cost 200$? that's almost absurd, 200$ will be about what a 4c/8t zen chip could cost..

~$175 for the 4c / 8t is my expectation. The 4c / 8 threaded Zen should be competitive with i5s. If 8 core / 16 threaded zen can only muster i5 level performance AMD is in trouble.
 
Glad to see they are going all out on the server chips. They've been out of the game so long on the server side they need to bring their A-game. The only thing that worries me is the 180w TDP I saw listed. That is a lot of heat.

However, Intel's current best is the 24 core Xeon E-7 8890 v4 at 165w TDP... so I guess 32 cores at 180w TDP is within expectations assuming these are running in the 2ghz range like the Intel chip.


180w TDP for 32 cores is pretty damn good IMO, their 8 core 9370 cpu is a 220w TDP chip.
 
the success or failure of zen will depend entirely of prices.. we have already in the market "Haswell" level of performance since 3 years ago.. we have actually skylake which it's a ~20% average of performance above haswell.... Even if AMD truly reached all the estimations about IPC that will mean still not much if Zen can't reach Intel-like clocks which may still keep'em out of the competition, that's why most people think zen will still be a failure, because they are aiming 3 years old performance..

Only people that are going to buy Zen are the same people have AMD chips today, so very very little amount... they need to be appealing to intel users, they need to make people jump the fence and buy their zen processors, they need to convince intel people they have a competitive product with latest intel products which at that moment probably they will be faced against kabylake, if not then no matter what, it will be a failure, that's why prices are what can save truly AMD. .




What make you think Zen 8c/16t will cost 200$? that's almost absurd, 200$ will be about what a 4c/8t zen chip could cost..
Sorry, meant 4 core 8 thread. I said 8 core but meant 8 thread.

Would really like to see it under 95w as well @ at least 3.4ghz
 
Last edited:
180w TDP for 32 cores is pretty damn good IMO, their 8 core 9370 cpu is a 220w TDP chip.

Those 32 cores might not even reach 2.0 GHZ.
It will be pretty decent especially if it can be around 1.8-2.2ghz speed on all cores. The 24 core Intel chip I mentioned is around 165w with all cores at 2.2ghz. If they can pull off near similar speeds with that many cores it means GloFo's processes are caught up and in good state.

We gotta keep in mind that these massive core count CPU's are for specific tasks that get more results with more cores than mhz. That doesn't mean they can be dog slow, but it means they don't have to be lightning quick either. If they hit 2.0ghz with 32 cores, I suspect it could be a winner in the server markets where the high core count Xeon's currently reign... at least until Intel releases their own.

As I mentioned before, my real concern is heat. Even if these chips are better than the 24 core Xeon, 180w TDP is a LOT of heat. 220w in an enthusiast desktop is one thing, 180w (x1, x2, or x4 arrangement) in a multi socket server board shoved into a 4U case while sitting in a rack mount with other servers will require some interesting cooling setups. Simply put, I'm concerned it might be a step too far. At least with multiple graphics cards the cooling goes front to back, fresh air goes over one chip and straight out the PCI slots. They only have to deal with the residual heat from sitting so close to each other. With multiple CPU arrangements and rack mount cases, sometimes the CPU's are right behind each other, feeding off each other's recycled air before getting blown out the back.

I would love to hear some thoughts on this from people who have experience with rackmount servers and whether or not this is a real issue. I'm sure these will be in server rooms with massive a/c units, but I'm sure there are limits.
 
Last edited:
Back
Top