Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [Guru3D] AMD Readies 14nm Zen - up-to 40 percent faster IPC performance
New Posts  All Forums:Forum Nav:

[Guru3D] AMD Readies 14nm Zen - up-to 40 percent faster IPC performance - Page 21

post #201 of 334
Quote:
Originally Posted by EniGma1987 View Post

Quote:
Originally Posted by epic1337 View Post

well yeah, suddenly having a magical leap in performance is quite unlikely, unless it was a breakthrough in architectural design.
but it still doesn't justify doing a less cost efficient chip, well while FX-9590 did go like that, it still isn't practical.
how long have they been designing zen anyway? knowing how poor their R&D department is, i doubt they could get a breakthrough in such a short notice.

will all the hype over zen being this toxic imho its making me want to see it be all lies, just to see disappointment flood these sort of threads.

Aren't you just a bundle of joy.

Zen has been in design since 2012 or maybe a little earlier. That was when the bulk of the team was hired for the new design and we started seeing interviews of the team's ideas for Zen.

What makes you think Zen is not as cost effective as Bulldozer? There is no evidence to support your thought. Also, any magical leap (either forward or backward) is possible with a whole new architecture. We saw the magical leap when AMD did the Athlon 64 and again with Bulldozer. One jumped way forward, one jumped way backward. In comparison to Bulldozer, this one will pretty much have to jump way forward, in comparison to any other decent architecture it is just minor to moderate improvements. It's all about perspective and what you are comparing against.

ask these guys, they're insisting that Zen would be absurdly more expensive than piledriver in cost efficiency.
and as for a new architecture, i don't deny making a large leap over bulldozer, even 50% over piledriver is already huge by intel's standard.
but these guys are saying that we should expect around 125% increase over piledriver instead.
Quote:
Originally Posted by Kuivamaa View Post

Quote:
Originally Posted by epic1337 View Post

i just don't understand you guy's logic.

my logic:
FX-6300 = <$100
FX-6300 + 50% performance = 6core Zen
<$100 + 50% = ~$150

is there anything wrong with this logic?

This isn't how the market works. See 5930k vs 5960x, 2 more cores, double the price. 4930k was losing in ST performance to 4770k and still it cost almost double the price. If zen is within 15-20% of skylake in ST , expect a 6C/12T chip to cost at least 50€/$ more than a standard unlocked i5 4C/4T intel quad. It will even trounce the i7 quad in MT while staying competitive in ST. It will be costly.
Quote:
Originally Posted by KyadCK View Post

Quote:
Originally Posted by epic1337 View Post

i just don't understand you guy's logic.

my logic:
FX-6300 = <$100
FX-6300 + 50% performance = 6core Zen
<$100 + 50% = ~$150

is there anything wrong with this logic?

Yes.

What is the price difference between a 5960X and a 5920k again? Price does not scale linearly with performance, it scales exponentially.

Also the 6300 is a PD chip, and as such it only uses 80% of each core under full load due to the decoder. It's a 4.8 core. The 40% is also taken from EX, not PD, which itself has fair IPC improvements. And then there's SMT. So all said and done, thats a hypothetical x2.25 performance over your 6300. Since that would easily compete with an i7, expect that to be a x3-x4 cost.

The 8-core version would add another 33% on to that, and as such it would become more like x5-x6 the price.




Quote:
Originally Posted by EniGma1987 View Post

Quote:
Originally Posted by epic1337 View Post

K12 should be a good server chip in terms of raw compute, but enterprise groups now prefer power efficiency over upfront costs.
and in which case atom server racks are now a thing, and ARM is now actually fairly viable to a broad range of usage.

K12 is supposed to be an ARM chip...? Pretty much screams power efficiency and lower costs.

well yeah K12 is ARM, and not all ARM SoCs are power efficient.
AMD seems to be targeting raw compute with their K12 rather than aiming purely for utmost efficiency.
Edited by epic1337 - 5/18/15 at 9:28am
post #202 of 334
Quote:
Originally Posted by KyadCK View Post

Because PD's IPC is lower than Ph II, and SR's contribution was fixing the decoder so that 4 cores were actually 4 cores and not 3.2 cores when fully loaded.
Steamroller increased multicore efficiency but also the singlecore performance.
post #203 of 334
Quote:
Originally Posted by Serios View Post

Steamroller increased multicore efficiency but also the singlecore performance.
actually kaveri single-thread performance at the same clock seemed lower than richland.

http://semiaccurate.com/2014/01/23/kaveri-versus-richland-performance-per-clock-comparison/
post #204 of 334
There's 100's of pages of kaveri info on this forum newer than jan '14.
SR is 10-15% faster single thread than PD
Summit Ridge
(16 items)
 
ASUS R510DP
(8 items)
 
 
CPUMotherboardGraphicsRAM
AMD Ryzen 1600X ASRock Fatal1ty AB350 Gaming K4 HIS Radeon HD7870 IceQ 2x8GB G.Skill TridentZ 3200 CL16 
Hard DriveHard DriveHard DriveHard Drive
Plextor M6S Plus 256GB SSD Toshiba X300 6TB Toshiba X300 6TB Toshiba 2TB 
CoolingOSMonitorMonitor
Deepcool BETA 400 ST Windows 8.1 Pro x64 HP S2031 20" Samsung SyncMaster 932BW 19" 
PowerCaseMouseMouse Pad
Seasonic S12G 750w Lian Li full tower Logitech MX310 SteelSeries 4HD 
CPUMotherboardGraphicsRAM
A10-5750m 3.5Ghz ASUStek A75M FCH HD 8650G + HD 8670M dual graphics 2x4GB Samsung 1600Mhz 
Hard DriveOptical DriveOSMonitor
Toshiba Q300 SSD 120GB Panasonic CD/DVD Windows 8.1 Pro 15.6" 1920x1080 
  hide details  
Reply
Summit Ridge
(16 items)
 
ASUS R510DP
(8 items)
 
 
CPUMotherboardGraphicsRAM
AMD Ryzen 1600X ASRock Fatal1ty AB350 Gaming K4 HIS Radeon HD7870 IceQ 2x8GB G.Skill TridentZ 3200 CL16 
Hard DriveHard DriveHard DriveHard Drive
Plextor M6S Plus 256GB SSD Toshiba X300 6TB Toshiba X300 6TB Toshiba 2TB 
CoolingOSMonitorMonitor
Deepcool BETA 400 ST Windows 8.1 Pro x64 HP S2031 20" Samsung SyncMaster 932BW 19" 
PowerCaseMouseMouse Pad
Seasonic S12G 750w Lian Li full tower Logitech MX310 SteelSeries 4HD 
CPUMotherboardGraphicsRAM
A10-5750m 3.5Ghz ASUStek A75M FCH HD 8650G + HD 8670M dual graphics 2x4GB Samsung 1600Mhz 
Hard DriveOptical DriveOSMonitor
Toshiba Q300 SSD 120GB Panasonic CD/DVD Windows 8.1 Pro 15.6" 1920x1080 
  hide details  
Reply
post #205 of 334
Quote:
Originally Posted by epic1337 View Post

actually kaveri single-thread performance at the same clock seemed lower than richland.

http://semiaccurate.com/2014/01/23/kaveri-versus-richland-performance-per-clock-comparison/
Man the article confirms that single thread performance is better not lower.
post #206 of 334
Quote:
Originally Posted by Serios View Post

Man the article confirms that single thread performance is better not lower.

i dunno about you but most of those benches are pretty multi-threaded to me.
and PPC is performance per clock, e.g. performance as a whole.
Edited by epic1337 - 5/20/15 at 4:42am
post #207 of 334
AMD desires to compete with intel along entire PC market segment that includes low end to high end. Message that they dont want to be cheaper option anymore is very clear message of what they are doing. That alone implies that they will have more expansive offerings.
New processors (if they deliver) will be priced accordingly.

Zen CPU offering for AM4:

Zen 4 core 4 threads priced slightly under i5 / 150 to 200$

Zen 4 core 8 threads price range between i5 and i7 / 250 to 300$

Zen 6 core 12 threads price range between low and mid end intel i7 extreme procesors / 350 to 450$

Zen 8 core 16 threads priced close to the high end intel i7 extreme processors / 700 to 999$

Of course they will have different versions of the same processors clocked higher and lower to fill in the gaps in pricing.
Edited by Stormscion - 5/20/15 at 5:28am
post #208 of 334
oh wow, now its 4cores at the $250 price point, i don't know what to think of AMD anymore.
i already miss piledriver, i wish they had decent MATX 990X board is all. sadsmiley.gif
FX-8320e 1/2 price of i5 and performs just 20% slower.
Edited by epic1337 - 5/20/15 at 6:07am
post #209 of 334
Quote:
Originally Posted by Stormscion View Post

AMD desires to compete with intel along entire PC market segment that includes low end to high end. Message that they dont want to be cheaper option anymore is very clear message of what they are doing. That alone implies that they will have more expansive offerings.
New processors (if they deliver) will be priced accordingly.

Zen CPU offering for AM4:

Zen 4 core 4 threads priced slightly under i5 / 150 to 200$

Zen 4 core 8 threads price range between i5 and i7 / 250 to 300$

Zen 6 core 12 threads price range between low and mid end intel i7 extreme procesors / 350 to 450$

Zen 8 core 16 threads priced close to the high end intel i7 extreme processors / 700 to 999$

Of course they will have different versions of the same processors clocked higher and lower to fill in the gaps in pricing.
There is no reason for them to have non-SMT products other than to handicap themselves or differentiate their product line. SMT uses so little die are that having all four SMT logic units defective is far less likely than having a single core be defective, which is the reason why all Zen-based products will have it enabled. I'd expect the 4C/8T model to be priced between $150 and $250 but based on the clock speed and TDP rather than SMT. At 3,5GHz the 4C/8T variant should be somewhat equivalent of the 3770k, so I'm guessing just under $200 for that.
Quote:
Originally Posted by epic1337 View Post

oh wow, now its 4cores at the $250 price point, i don't know what to think of AMD anymore.
i already miss piledriver, i wish they had decent MATX 990X board is all. sadsmiley.gif
FX-8120e 1/2 price of i5 and performs just 20% slower.
People think it will be nearly as fast as Haswell, but forget that Skylake is going to release before it. AM4 should have plenty of smaller boards now that the i5 equivalents aren't going to use 125W.
Edited by Tojara - 5/20/15 at 6:11am
post #210 of 334
Quote:
Originally Posted by Tojara View Post

People think it will be nearly as fast as Haswell, but forget that Skylake is going to release before it. AM4 should have plenty of smaller boards now that the i5 equivalents don't use 125W.

not worth it though, i'm aiming for a 4core skylake non-K below $200.
if 4core/4thread would be about just below haswell, then why bother when skylake is around 20% faster.
in which case, its either a 6core Zen or i'm sticking with skylake, at least i get to have a highly efficient chip.

ohh and 6core/12thread zen over $400 is overpriced, considering 6core haswell is priced below $400. thumb.gif
and on that note, just because it suddenly became competitive, doesn't mean they have to follow intel SRP, intel's pricing is overpriced already.

tbh, price/performance of piledriver chip as it is is roughly 42% better than intel's ( $135 FX-8320e 80% of $240 i5-4690K = ($250 * 0.8) / $135 = 42% )
if they price Zen cores close to Haswell, when it's barely as fast as haswell, then its technically 30% less cost efficient than piledriver.
Edited by epic1337 - 5/20/15 at 6:31am
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Rumors and Unconfirmed Articles
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [Guru3D] AMD Readies 14nm Zen - up-to 40 percent faster IPC performance