Thread replies: 37
Thread images: 16
Thread images: 16
File: Screenshot from 07-02-2017 16:36:22.png (190KB, 502x506px) Image search:
[Google]
190KB, 502x506px
>Zen 6 core matches Intel 6 core clock for clock
>But it has shit IPC unsuited for gaymss!
Really fires up the brain juice.
>>
>>
>>59015737
???
>>
>>59015722
>matches clock for clock
>doesn't have matching instructions per cycle
>>
>>59015790
You might be retarded, anon. Best get a check on that.
>>
File: Intel-8th-Generation-Core-i7-8000-Series.png (402KB, 1683x948px) Image search:
[Google]
402KB, 1683x948px
>>59015722
4 year old Intel performance, Kaby lake is 15% faster than Skylake and much faster than Broadwell not to mention Haswell, AMD once again doesn't even have 60% of Intel's IPC and there's nothing you can do about it
>>
>>59015790
If it has the same performance at the same clocks then it has the same instructions per clock, holy shit people performance is a measure of clockspeed*instructions, if it has the same performance, same clockspeed, how can it have different instruction execution? HOLY SHIT PEOPLE IT'S LITERALLY FIND 2 + X = 4 PLEASE FIND X
>>
>>59015833
>sysmark
The gpu makes the difference you shill faggot
>>
>>59015880
It actually doesn't and Kabylake GPU is barely faster than a Skylake one.
>>
>>59015722
If you subtract the L3 and L2 from each you get
Zen: 5.5mm2
Skylake: 6.575mm2
The two caveats here are that intel's process is nominally 15% denser than 14nm LPP, and Skylake has 256bit datapaths in a monstrously big FPU. So Zen isn't quite as big as intel's core arch, but it doesn't necessarily need to be. A lot of transistors and silicon in current intel core arch are dedicated to processing ops which are rarely ever used outside of HPC, and in HPC large vector workloads don't need to be handled by the main processor on a node. Intel created the Xeon Phi specifically for this. In fact large vectors don't necessarily need to be used at all, hence why GPUs are still a staple.
In other areas of the core Zen isn't lacking. The throughput for integer bound ops is very high, and even 128bit FPU ops have high throughput due to the individual add and multiply pipelines.
Though as a fun comparison of die size Excavator is 14.48mm2. With perfect area scaling being ported to 14nm LPP it would be 4.344mm2. To date Excavator is the biggest design AMD has ever made in terms of transistor count and relative size. The Zen core here is over 25% larger than the Excavator module.
>>59015833
Kaby Lake is only faster than Skylake if you compare differently clocked SKUs. They have nearly the exact same performance per clock because internally Kaby Lake is just Skylake with a few hardware bugs fixed. There were no arch changes, just fixes. Perf/clock is nearly exactly the same, margin of error is less than 1%.
>>
>>59015833
Kabylake is around 100% faster than Sandybridge
footnote: KBL at 4.5 and SB at 2.5GHz
>>
>>59015833
>Kaby lake is 15% faster than Skylake
I like that there are people out there who actually believe this.
>>
>>59015909
Might as well post all the available ISSCC slides
>>
>>59016019
I can't believe I missed the >40% IPC
>
>>
>>59016038
>>59016038
Its somewhere 50-55% higher than Excavator
>>
>>59015722
>But it has shit IPC unsuited for gaymss!
what do you mean?
>>
>>59015892
The difference is somewhere around 10-15%. Couple that with few hundred mhz higher cpu clocks and you get the 15%. Kaby lake is 0% faster than Skylake.
>>
>>59016134
Last one I have.
The full presentation hasn't been leaked yet, these are just whats available from Goto-san. Everything else won't be available until NDA is up.
>>
>>59016058
i wonder if they are using the same branch as excavator, would be interesting especially after the more efficient SMT that HT claims.
>>
>>59016669
The perceptron style branch predictor is an evolution of what is utilized in the cat core family, not related to whats in Excavator.
>>
noice... I want to buy one of these, how much will I have to wait?
>>
>phoneposter filename
>pol meme spam
poetry
>>
>>59017241
>>phoneposter filename
>Screenshot from 07-02-201(...).png
I'm pretty sure single Linux/UNIX screenshot tool uses similar naming
>>
File: suicidewatch.png (17KB, 1004x378px) Image search:
[Google]
17KB, 1004x378px
>>59015833
Literally damage control.
http://www.hardocp.com/article/2016/12/09/intel_kaby_lake_core_i77700k_ipc_review/2
>>
File: 1487524017762.png (234KB, 800x612px) Image search:
[Google]
234KB, 800x612px
>>59015737
>>
>>59015737
how many rupees do you get per post?
>>
>>59015866
I mean technically its 2x = 4 but same exact result anyway, i dont get why these intel shills are claiming ipc doesnt work that way
>>
>>59018616
>DVD Rip to 720p
>>
>>59019571
http://www.anandtech.com/show/10959/intel-launches-7th-generation-kaby-lake-i7-7700k-i5-7600k-i3-7350k/8
http://www.hardocp.com/article/2016/12/09/intel_kaby_lake_core_i77700k_ipc_review/4
>>
>>59019625
Why are you posting those links?
Thread posts: 37
Thread images: 16
Thread images: 16