AMD Zen Performance Preview

[noko]

Pricing of manufacturing of these kinda of parts, doesn't matter if its in China or the US most of it is automated. The steps of designing and testing the chip can be done anywhere in the world.

Just like how AMD and nV work design is done in one of their many design groups around the world, they have their parts made at any fabs they want but testing is done in India, cheaper for them, to do it there. Intel can be doing this too, fabs are here in the US but that doesn't mean they are designing and testing here in the US (although I think design is done here in the US, not sure about testing). But even if Intel is doing everything here in the US, AMD being fabless, off sets the cost (GF needs to make a profit too)

I am thinking of China underselling it like they do dumping steel in the US was the chain of thought there.

Really if you make a whole GPU with more transistors, ram, power circuits, plugs, board and sell it cheaper then Intel's processors by a long ways. Even a Titan X Pascal is cheaper then some of Intel's CPU's and it is at a very profitable price. In other words AMD is maybe getting back to reality of pricing and hopefully making a decent profit to move on and continue in business. We will just have to wait for official AMD MSRP pricing for Zen vice expecting something that maybe ridiculous. Well maybe by March of next year will be my upgrade point for one of my rigs, was expecting June but if launched in January that will be great.

 

[noko]

Going into a price war when products are competitive from a performance/feature/power perspective. We know both companies are on 14nm, which are fairly equal (if anything Intel has an advantage but for this discussion lets ignore that).

Just say you have a product that is equal to another product done by someone that is a monopoly and they have set the price (price isn't elastic in a monopoly). Would you cut your price down when you know full well the competitor will match you?

This is what the current CPU marketplace looks like (not numbers just over all view)

price elasticity of demand, is none existent in a monopoly which the CPU market is virtually a monopoly.

Now a new competitor (virtually AMD is a new competitor) would take advantage of this and price just under or equal to the monopoly's products, that would maximize their profits even for a short amount of time and let the monopoly start the price war.

The other way to look at it is, if AMD starts the price war, they can gain marketshare quickly, but that doesn't work because the first thing any company or person buying the monopoly's product will ask the monopoly "hey can you match their price?" All of this is because the products are EQAUL.

How many times have you got to Best Buy and pulled out your phone and showed the sales man, look they are selling this product you have for this much less, if you can beat the price I'll get it here?

Both of these situations drops the price of the goods that are viaing for our money and the monopolistic price will come down automatically to the competitive price, and both of these situations don't favor the newcomer in making any ground. But one of them has an advantage of the newcomer in possibly making more money up front.

Now lets flip this around and look at what happened with Athon generations. AMD had an undeniable advantage in performance and power usage. This forced Intel to do shady tactics and down right cheat AMD out of the market. Even with that going on AMD was able to gain marketshare (just not as fast or as much as they should have gained). With out that advantage this time around, we can't expect that to happen, not only that, we can't even expect them to gain much of anything at all in a saturated and mature marketplace. Because

A) there are no new consumers
B) there are no growing markets
C) there are no advantages to switching over to the new product

To A) So AMD has to do is steal Intel's customers, which is not easy to do in this market
To B) Create a niche because there is something special about their product (which we know there is nothing special, what they can do Intel processors can too.
To C) The increased cost of OEM's, server providers etc need to be offset by the hardware they use or increase in consumers buying those products need to make the purchasing of the new product viable. Which points to A, the market already has its set consumers so that doesn't work.

There are growing markets and the big one is China followed by India - in fact a very huge growing market and if AMD can tap into those markets it could dwarf US sells overall. AMD server market in China with China made AMD CPU's for example. China team designed GPU - Vega and so on. For those markets Intel priced themselves out of it, AMD would be sitting good. That is if they can actually produce and supply the demand.

 

[razor1]

There are growing markets and the big one is China followed by India - in fact a very huge growing market and if AMD can tap into those markets it could dwarf US sells overall. AMD server market in China with China made AMD CPU's for example. China team designed GPU - Vega and so on. For those markets Intel priced themselves out of it, AMD would be sitting good. That is if they can actually produce and supply the demand.

Think you are getting confused with markets that upgrade slower then our market or european market vs what a mature and saturated market is. The PC industry as a whole and its current individual components are saturated and mature, they are not a growing market hence why we have seen over all PC, CPU volume sales drop over the past few years.

 

[razor1]

I am thinking of China underselling it like they do dumping steel in the US was the chain of thought there.

Really if you make a whole GPU with more transistors, ram, power circuits, plugs, board and sell it cheaper then Intel's processors by a long ways. Even a Titan X Pascal is cheaper then some of Intel's CPU's and it is at a very profitable price. In other words AMD is maybe getting back to reality of pricing and hopefully making a decent profit to move on and continue in business. We will just have to wait for official AMD MSRP pricing for Zen vice expecting something that maybe ridiculous. Well maybe by March of next year will be my upgrade point for one of my rigs, was expecting June but if launched in January that will be great.

They can't do that with these kinds of products, they are not rip off sun glasses or knock off jeans lol.

Titan X pascal is cheaper than Intel chips and end results are it gets less margins then those more expensive Intel chips That is why Intel's margins are higher than nV's.

 

[Shintai]

Be careful comparing chips like that. Else you end up with stupid examples of the same wafer sized dies selling for 5$ while another sells for 25000$. Simply because they cant be compared like that across product groups.

 

[noko]

Be careful comparing chips like that. Else you end up with stupid examples of the same wafer sized dies selling for 5$ while another sells for 25000$. Simply because they cant be compared like that across product groups.

Well you can say the Tesla GPU's are high dollar chips per die size etc.

 

[Shintai]

https://www.heise.de/newsticker/mel...enCompute-mit-ROCm-aber-kein-Zen-3491050.html

- No samples were presented/showcased at SC16 conference, not even behind the public
- In NDA sessions only slides were presented
- In SPECint2006 AMD replaced Intel with gcc compiler and the performance for 32 core Naples was slightly below Haswell
- Results are without Hyper-Threading because of somes issues
- When all is working AMD claims they will be on par with Skylake EP (Xeon E5 2698v5), but there is no info on what this is based on or if this is even based on some real comparisons rather than predictions, also AMD remained silent about some other important HPC performance metrics (e.g. SPECfp, Linpack)

So 32 core/64 thread Napples is around the same performance in best case as a 16-18 core/32-36 thread Haswell.

 

[razor1]

Hmm well they are expecting a 32 core Zen to compete Skylake EP, Zeon E5 2698v5, that seems likely, if its got Ivy bridge to Haswell IPC.

That intel chip is not the highest Intel in Skylake EP line at all, its err 3 or 4th down if we don't consider the e7 line at all, if we consider the e7 line its about 6 or 7th on the stack.

And most likely that specific Skylake chip won't have all cores activated and going up against the top Zen, seems about right. (26 or 24 cores max on that chip).

So it looks like AMD is expecting a 30% difference in performance on a per core with similar clocks against Sky Lake.

 

[noko]

Hmm well they are expecting a 32 core Zen to compete Skylake EP, Zeon E5 2698v5, that seems likely, if its got Ivy bridge to Haswell IPC.

That intel chip is not the highest Intel in Skylake EP line at all, its err 3 or 4th down if we don't consider the e7 line at all, if we consider the e7 line its about 6 or 7th on the stack.

And most likely that specific Skylake chip won't have all cores activated and going up against the top Zen, seems about right. (26 or 24 cores max on that chip).

So it looks like AMD is expecting a 30% difference in performance on a per core with similar clocks against Sky Lake.

I would not believe anything until real chips are available for testing, also if AMD was clearly ahead I would expect samples would have been provided or shown for testing even if for only the event. So at best just conjecture.

 

[razor1]

if AMD is expecting something, that means best case lol. That is the way its written, now my German reading isn't the greatest, and the translation could be better too lol.

 

[Shintai]

3.2Ghz base, 3.5Ghz turbo.

 

[Zarathustra[H]]

I think for me it's going to come down to how well it overclocks.

Rumors are now suggesting (though I have absolutely no idea how reliable they are, could just be fabrication, so take it with a truckload of salt) that 4.2 Ghz likely is the top speed part. Same rumors said nothing about whether this is base clock or max turbo clock.

This sounds optimistic to me, but for shits and giggles lets assume that it is true, and paint the most positive picture we can for Zen based on what we know, just to kind of grasp at what the upper limits of our expectations should be. (note there are going to be rounding errors throughout this process, following best practice I am only rounding final results)

Looking at Anandtech's bench page for the only real single threaded bench we have (Cinebench R11.5) we see that the Steamroller based A10-7870k scores 1.06 at its 4.1Ghz max turbo clock. So let's divide score by max turbo clock speed to get at our IPC. 1.06/4.1 = 0.2585

Lets add ~5% to approximate Excavator speeds: 0.2585*1.05 = 0.2715

Now lets add 40% based on AMD's statements that Zen will have 40% higher IPC than Excavator: 0.2715*1.4 = 0.3800

Lets make a complete bullshit guess and assume that the top Zen part at launch will have a 4.2Ghz base clock and a 4.4Ghz turbo clock. So since single threaded loads get the max turbo clock we can estimate the final benchmark score at this clock to be: 0.3800 * 4.4 = 1.6722

Now lets look at what else scores ~1.67 from the linked Anandtech page above:

Intel Core i5 4670K:                  1.7
Intel Core i3 4360:                   1.7
Intel Core i5 4690:                   1.7
Intel Core i5 4690K:                  1.7
Intel Core i7 4960X:                  1.69
Intel Core i7 6950X:                  1.69
My BS Optimistic Zen Prediction:      1.67
Intel Core i7 5930K:                  1.66
Intel Core i7 3770K:                  1.66
Intel Core i7 4820K:                  1.64
Intel Core i7 4930K:                  1.64

This is not bad company to be in, and a huge leap forward for AMD (though of course, this is a highly BS prediction based on very optimistic assumptions)

Now on the flip side, my 5 year old i7-3930k overclocked to 4.8Ghz scores 1.92 in this same benchmark.

So, based on these assumptions, in order to match my 5 year old Sandy-E chip overclocked to overclocked, Zen will have to be able to hit 5.05Ghz on my custom water loop.

Possible? Maybe. Likely? I'm leaning towards no.

I'm still hoping for the best, but expecting the worst.

 

[Nirad9er]

3.2Ghz base, 3.5Ghz turbo.

That would be on par with with the 6900k. Hopefully it overclocks to mid 4ghz otherewise not so enticing for me. Most people care about max over clock. I sure as hell wouldn't get 3.2ghz CPU for gaming that can't go above 4ghz. It would be a downgrade from my 4790k. Regardless, if it matches a 6900k at base clock then intel WILL have to drop prices.

 

[Zarathustra[H]]

I think for me it's going to come down to how well it overclocks.

Rumors are now suggesting (though I have absolutely no idea how reliable they are, could just be fabrication, so take it with a truckload of salt) that 4.2 Ghz likely is the top speed part. Same rumors said nothing about whether this is base clock or max turbo clock.

This sounds optimistic to me, but for shits and giggles lets assume that it is true, and paint the most positive picture we can for Zen based on what we know, just to kind of grasp at what the upper limits of our expectations should be. (note there are going to be rounding errors throughout this process, following best practice I am only rounding final results)

Looking at Anandtech's bench page for the only real single threaded bench we have (Cinebench R11.5) we see that the Steamroller based A10-7870k scores 1.06 at its 4.1Ghz max turbo clock. So let's divide score by max turbo clock speed to get at our IPC. 1.06/4.1 = 0.2585

Lets add ~5% to approximate Excavator speeds: 0.2585*1.05 = 0.2715

Now lets add 40% based on AMD's statements that Zen will have 40% higher IPC than Excavator: 0.2715*1.4 = 0.3800

Lets make a complete bullshit guess and assume that the top Zen part at launch will have a 4.2Ghz base clock and a 4.4Ghz turbo clock. So since single threaded loads get themax turbo clock we can estimate the final benchmark score at this clock to be: 0.3800 * 4.4 = 1.6722

Now lets look at what else scores ~1.67 from the linked Anandtech page above:

Intel Core i5 4670K:                  1.7
Intel Core i3 4360:                   1.7
Intel Core i5 4690:                   1.7
Intel Core i5 4690K:                  1.7
Intel Core i7 4960X:                  1.69
Intel Core i7 6950X:                  1.69
My BS Optimistic Zen Prediction:      1.67
Intel Core i7 5930K:                  1.66
Intel Core i7 3770K:                  1.66
Intel Core i7 4820K:                  1.64
Intel Core i7 4930K:                  1.64

This is not bad company to be in, and a huge leap forward for AMD (though of course, this is a highly BS prediction based on very optimistic assumptions)

Now on the flip side, my 5 year old i7-3930k overclocked to 4.8Ghz scores 1.92 in this same benchmark.

So, based on these assumptions, in order to match my 5 year old Sandy-E chip overclocked to overclocked, Zen will have to be able to hit 5.05Ghz on my custom water loop.

Possible? Maybe. Likely? I'm leaning towards no.

I'm still hoping for the best, but expecting the worst.

3.2Ghz base, 3.5Ghz turbo.

Using my same analysis as above, that results in a 1.33 score and puts us in the following territory:

Intel Xeon v2 - E5 2697:          1.39
Intel Core i3 3220:               1.37
Intel Core i5 2400:               1.37
Intel Core i7 990X:               1.33
My BS Zen Calculation:            1.33
Intel Pentium G2130:              1.33
Intel Xeon v3 - E5 2650 v3:       1.33
Intel Core i3 4130T:              1.32
Intel Core i3 6100TE:             1.31
Intel Pentium G2030:              1.26
Intel Xeon-D 1540:                1.25

If this is where it falls, I am almost definitely going to have to pass. it exactly ties the 990X, a chip which at launch will be 6 years and 5 generations old. Huge leap forward for AMD or not, this will not be sufficient for me.

 

[Boggy101]

I'm hoping that their SMT matures well. Intel's first attempt at hyperthreading was scrapped when it was shown to be not very useful or efficient.

 

[gamerk2]

I'm hoping that their SMT matures well. Intel's first attempt at hyperthreading was scrapped when it was shown to be not very useful or efficient.

That was largely because the Windows scheduler didn't have a clue how to handle SMT. It treated the HTT core as a fully independent CPU, and this naturally resulted in stalls as the HTT core blocked the physical core from executing. This is specifically why Intel added a HTT CPUID field that lets the OS know the chip is HTT capable, so it can schedule properly.

It's ironic that when BD launched it ran into the SAME EXACT ISSUES the Pentium 4 did, and Microsoft eventually had to patch the kernel.

 

[Shadohh]

That was largely because the Windows scheduler didn't have a clue how to handle SMT. It treated the HTT core as a fully independent CPU, and this naturally resulted in stalls as the HTT core blocked the physical core from executing. This is specifically why Intel added a HTT CPUID field that lets the OS know the chip is HTT capable, so it can schedule properly.

It's ironic that when BD launched it ran into the SAME EXACT ISSUES the Pentium 4 did, and Microsoft eventually had to patch the kernel.

And HTT still sucks today.

I have a much better computing experience turning off HTT on my 6 core.

 

[SvenBent]

before dissing HTT try to disable core parking it makes up huge amount of the lost performance.
Also no need to turn of HTT. Affinity control solves it (After disabling core parking).

SMT will per design have issues in certain combination as long as thread scheduler keeps using them as regular cores.

If just we had some nice free software that could disable core parking and put our games automatically with affinity to avoid HT.... oh wait... we do.

 

[Araxie]

before dissing HTT try to disable core parking it makes up huge amount of the lost performance.
Also no need to turn of HTT. Affinity control solves it (After disabling core parking).

SMT will per design have issues in certain combination as long as thread scheduler keeps using them as regular cores.

If just we had some nice free software that could disable core parking and put our games automatically with affinity to avoid HT.... oh wait... we do.

can you test if core parking have a performance impact under Win10? as by default on performance mode Core parking is disabled on Win10.. normally under Win7 I do a registry modification on HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Power\PowerSettings\54533251-82be-4824-96c1-47b60b740d00\0cc5b647-c1df-4637-891a-dec35c318583 and just change the attributes to (0), that allow to select in power option - advanced power option - processor power management - processor performance core parking min cores. [where 0% is fully enabled an 100% fully disabled].. 100% is by default.. so disabling core parking have 0 effect on performance (at least on my test) but i haven't had enough time to do some proper test..

same registry modification apply for all windows from win7 to win 10. in win8.1 is also disabled by default.

 

[SvenBent]

can you test if core parking have a performance impact under Win10? as by default on performance mode Core parking is disabled on Win10.. normally under Win7 I do a registry modification on HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Power\PowerSettings\54533251-82be-4824-96c1-47b60b740d00\0cc5b647-c1df-4637-891a-dec35c318583 and just change the attributes to (0), that allow to select in power option - advanced power option - processor power management - processor performance core parking min cores. [where 0% is fully enabled an 100% fully disabled].. 100% is by default.. so disabling core parking have 0 effect on performance (at least on my test) but i haven't had enough time to test..

same registry modification apply for all windows from win7 to win 10. in win8.1 is also disabled by default.

I already did and showed pictures of this before .... TO YOU
I don't do it by registry it an annoying way to do it in my opinion. I disable it directly. much easier and don't mess with the registry at all


here is the performance test
https://hardforum.com/threads/3570k...-it-to-oc-to-5-0-ghz.1916083/#post-1042647402

Her is where you can see them being parked
https://hardforum.com/threads/3570k...-it-to-oc-to-5-0-ghz.1916083/#post-1042647743


HOWEVER SKYLAKE MIGHT BE DIFFERENT
and i honestly things skylake is the reason people mistankenly things its not happening in windows 10. people have a tendency to run with the first explanation that pops into their heads instead of testing it.
Intel changes some of the powermangement in skylake to be conntrolled by hardware instead of OS.
I don't have skylake to test with so i don't know anything about it specifically. but it sounds reasonable plausible that that is why people get it wrong.

If skylake does not have the core parking issue, you can safely use affinity to "disable" hyper threading from a program/game

 

[Araxie]

I already did and showed pictures of this before .... TO YOU
I don't do it by registry it an annoying way to do it in my opinion. I disable it directly. much easier and don't mess with the registry at all


here is the performance test
https://hardforum.com/threads/3570k...-it-to-oc-to-5-0-ghz.1916083/#post-1042647402

Her is where you can see them being parked
https://hardforum.com/threads/3570k...-it-to-oc-to-5-0-ghz.1916083/#post-1042647743


HOWEVER SKYLAKE MIGHT BE DIFFERENT
and i honestly things skylake is the reason people mistankenly things its not happening in windows 10. people have a tendency to run with the first explanation that pops into their heads instead of testing it.
Intel changes some of the powermangement in skylake to be conntrolled by hardware instead of OS.
I don't have skylake to test with so i don't know anything about it specifically. but it sounds reasonable plausible that that is why people get it wrong.

If skylake does not have the core parking issue, you can safely use affinity to "disable" hyper threading from a program/game

yes I saw that pic but that are on Win7.. based on that same conversation the other day, I have an active i7 3770K machine with win10 and Core Parking by disabled by default, I decided to re-enable that machine to do that kind of test... Win10 which infact im using right now have core parking disabled without any modification win10 power plan is just set to high performance mode. that's all.

about the registry, isn't disabled in the registry, that modification in the registry is just to allow in the windows power management enable/disable core parking, in the same tab where you change processor minimum state, it enable another one that say "processor performance core parking min cores".. on win7 by default is 10%. win10 100% so core parking is disabled, that increased performance you have may be due the affinity management of ProjectMercury?. this is why im saying this to you, how project mercury affect a win10 machine which already have disabled core parking on high performance mode.

 

[Shadohh]

before dissing HTT try to disable core parking it makes up huge amount of the lost performance.
Also no need to turn of HTT. Affinity control solves it (After disabling core parking).

SMT will per design have issues in certain combination as long as thread scheduler keeps using them as regular cores.

If just we had some nice free software that could disable core parking and put our games automatically with affinity to avoid HT.... oh wait... we do.

the name of that software being? I assume its the mercury thing in your tag?

I will have to check it out at home, its blocked from work.

 

[SvenBent]

alto of stuff that gets seperated

yes I saw that pic but that are on Win7..

If you look at the _ [] X of the pictures you can clearly see its not win7 but win10



based on that same conversation the other day, I have an active i7 3770K machine with win10 and Core Parking by disabled by default, I decided to re-enable that machine to do that kind of test... Win10 which infact im using right now have core parking disabled without any modification win10 power plan is just set to high performance mode. that's all

Its kinda hard to argue that coreparking is disabled by default if you had to change the powerplan. its a minor change indeed but not really default


about the registry, isn't disabled in the registry, that modification in the registry is just to allow in the windows power management enable/disable core parking, in the same tab where you change processor minimum state, it enable another one that say "processor performance core parking min cores".. on win7 by default is 10%.

Ok the Windows registry edit enabled it in power manangement. But you can in registry set a max and min that is allowed to be selected in power managment.
some people set those to the same value to lock it in and thereby disable core parking.
That would be the ValueMin and ValueMax in the same registry path. (can you perhaps check what they are at?)

That also mean it possible to have it enabled by the registry some places but the one you are using does not affect it. Because you are just messing what this settings is shown or not, not what it is.
So you might have it overridden in registry from some other tool or prior registry hack.
Which would explain the symptoms you are seeing

This is the very reason I stay away from messing with disabling core by the registry because there is several path to do it and it can override each other.
Also some of the registry change only works after a reboot. ( did you remember to reboot?) where is my method of disabling it directly work live.



win10 100% so core parking is disabled, that increased performance you have may be due the affinity management of ProjectMercury?. this is why im saying this to you, how project mercury affect a win10 machine which already have disabled core parking on high performance mode.

The test results giving was without affinity adjustments. I know this because i test with and without it.
20% boost came from disabling core parking
30% boost from disabling core parking AND affinity adjusting to avoid thread conflicts
but since you were only asking about coreparking in the other thread i only provide the results from that

the original test i did back in the days can be seen here but its in a shitting ugly language
http://www.dailyrush.dk/debat/emne/disable-core-parking-blaastemplet-tweak-d/

translation of 1st test
7-zip compression 4 threads 32mb
with core parking 14.4xxkbytes/s (avergae over 6 runs)
disabled core parking 16.6xxkbytes/s (average over 6 runs)


end test
8thread HTon parking (Default)
797
797
796

8thread HTon no parking
799
798
799

4thread HTon parking (Default
473
477
474

4thread HTon No parking
541
542
536

4thread HTon No parking No HT conflicts
619
617
616

4thread HTon Parking No HT conflicts
316
315
312


As you can see it was tested individually in regards to coreparking and affinity control ( No HT conflicts)
This is not even all of the results cause i did the same test with HToff ( disabled HT in bios)
but i did report those as they where unimportant for the purpose
Trust me i always test very thoroughly

 

[SvenBent]

the name of that software being? I assume its the mercury thing in your tag?

I will have to check it out at home, its blocked from work.

Yes it is.

Its portable. You just start it and go int osettings
Put a mark in the "disable core Parking" to disable coreparking.
Put a mark in "no Ht conflicts" to have you main application ( game) have affinity set to avoid HTT.

Both of these are none permanent soo simply closing the program or rebooting will remove the effect

 

[SvenBent]

Araxie

OH and in case peopel are worrying about the increase power draw from diabling Core parking
I had this draw from the wall during idle

with core parking 123.5w to 124w
without core parking 123.5w to 124w with 6 out of 8 logical cores in parked mode

 

[Shadohh]

Yes it is.

Its portable. You just start it and go int osettings
Put a mark in the "disable core Parking" to disable coreparking.
Put a mark in "no Ht conflicts" to have you main application ( game) have affinity set to avoid HTT.

Both of these are none permanent soo simply closing the program or rebooting will remove the effect

Thanks, WOW hates HT, but BF1 likes it.

 

[SvenBent]

Offtopic sorry

Thanks, WOW hates HT, but BF1 likes it.

I'm going to make a hotkey for enabling/disabling "No HT conflicts" that way you dont have to go into the settings menu and adjust it all the time. I'm hoping that will make it more useable

b

 

[JustReason]

lol, for a min there I had to check and make sure I was in the AMD section.

 

[pendragon1]

yeah just a little ot...

 

[drescherjm]

In a few months I expect there will be a lot of discussion about HT and core parking here..

 

[Dreamerbydesign]

We are all hoping, we'll just have to see.

It doesn't scare me all that much that they down clocked an intel chip for a comparison. It ran faster on the same terms >.< Proves their architecture is superior... now it will come down to manufacturing. This far out from a launch it should be obvious they aren't running at final clock, if they where... the launch would be now. Intel has done a good job of spending tons of money on Fab... that will be the hurdle AMD has to jump.

The question will be;
Can AMDs > architecture make up for < fab tech.

This far out? If they launch q1 of 2017, these chips have to be in production now. Unlike their video cards, they can't change much on the cpu via software.

 

[Araxie]

lol, for a min there I had to check and make sure I was in the AMD section.

yeah just a little ot...

Not off topic at all, both should be reading and documenting yourself about HyperThreading because that's what SMT is.. sandy bridge/ivy bridge Hyperthreading tech as it seems to feature 4FP micro ops and 6INT micro ops per core Cycle which should provide similar scalability. well that's if everything posted {here} is correct.. haven't followed zen lately..

 

[Pieter3dnow]

In a few months I expect there will be a lot of discussion about HT and core parking here..

It really depends on the hardware scheduler .

 

[ChadD]

This far out? If they launch q1 of 2017, these chips have to be in production now. Unlike their video cards, they can't change much on the cpu via software.

You are replying to something I posted on Aug 24, 2016. Three months ago, no they where not in production for chips they where planning to sell a year out. The first zen parts that ship in Q1 are not likely to be the bulk of the line.

So yes for sure they should be in production now for chips they are planning to sell in bulk in early 2017. At the very least close to a final spec to run anyway.

What I said back in Aug is still true. Clock for clock they have a better chip... at least in so far as all the leaks to this point. Yes the intel parts are still edging them out a little bit... just not at = clocks. Fab tech is still going to decide things... if AMD is able to fab with a small enough process and still have a low enough rate of failure to keep the cost down. They are well poised to make it a race again. Intel will likely still have the fastest chips at the high end... running at higher clocks, but in the mid range. If the leaks are mostly correct AMD is going to make it a very interesting race.

 

[doz]

You are replying to something I posted on Aug 24, 2016. Three months ago, no they where not in production for chips they where planning to sell a year out. The first zen parts that ship in Q1 are not likely to be the bulk of the line.

So yes for sure they should be in production now for chips they are planning to sell in bulk in early 2017. At the very least close to a final spec to run anyway.

What I said back in Aug is still true. Clock for clock they have a better chip... at least in so far as all the leaks to this point. Yes the intel parts are still edging them out a little bit... just not at = clocks. Fab tech is still going to decide things... if AMD is able to fab with a small enough process and still have a low enough rate of failure to keep the cost down. They are well poised to make it a race again. Intel will likely still have the fastest chips at the high end... running at higher clocks, but in the mid range. If the leaks are mostly correct AMD is going to make it a very interesting race.

Keep chasing the rainbow.

 

[Spirit_Retro]

Keep chasing the rainbow.

How disrespectful.

 

[doz]

How disrespectful.

Why, because people keep believing Zen will actually be faster than the 7700k (let alone 6700k) clock for clock?

Pipe dreams are real. Maybe I will be wrong though. Again, keep chasing something you will never get to (rainbows). Sorry if I offended your SJW attitude.

 

[Wascrash]

Why, because people keep believing Zen will actually be faster than the 7700k (let alone 6700k) clock for clock?

Pipe dreams are real. Maybe I will be wrong though. Again, keep chasing something you will never get to (rainbows). Sorry if I offended your SJW attitude.

I agree with you. We cant expect this CPU to perform the same as Intel that has had years to tweak .. even at clock for clock. This is a new CPU and needs some time to mature. I hope it does offer some mid range value though. If not my next CPU will be Intel. Ill wait a year if I can lol.

 

[DedEmbryonicCe11]

If they get within 12% of the IPC of Kaby Lake it would be a major success given the massive difference in R&D spending. I don't expect it to beat Intel at <= 4 threads at anything, just to be competitive in terms of pricing per core/thread.

 

[Spirit_Retro]

Why, because people keep believing Zen will actually be faster than the 7700k (let alone 6700k) clock for clock?

Pipe dreams are real. Maybe I will be wrong though. Again, keep chasing something you will never get to (rainbows). Sorry if I offended your SJW attitude.

WTF is "SJW"?

My attitude is one of respect.