[U-3C]

I guess I'll find out soon enough!

Always wanted one of those!

 

[Tilpo]

Is it ok if i enter into this thread even if i am not a regular Anime watcher
I have watched only few famous anime like Deathnote, Onepunch man, Attack on titans, Onepiece, Naruto, Bleach, Trigun, Fullmetal alchemist & brotherhood, Mushi shi, lelouch the rebellion, Kuroko's basketball, Shokugeki no Souma..

I haven't watched any anime in at least a year, and I'm not into hi-fi. Yet somehow people aren't chasing me away.

 

[Tiddlesworth]

AMD Ryzen seems pretty alright so far, hopefully isn't synthetic smoke and mirrors, Intel needs a wake up call.

 

[AxelCloris]

  AMD Ryzen seems pretty alright so far, hopefully isn't synthetic smoke and mirrors, Intel needs a wake up call.

Meh, I'll stick with the i7-6950X.

 

[Tiddlesworth]

 
Meh, I'll stick with the i7-6950X.

"Meh, I'll stick with my 1.7k USD cpu."

 

[AxelCloris]

  "Meh, I'll stick with my 1.7k USD cpu."

Precisely.

 

[deadlylover]

  AMD Ryzen seems pretty alright so far, hopefully isn't synthetic smoke and mirrors, Intel needs a wake up call.

Meh, I'll stick with the i7-6950X.

 
Yup it looks promising, even if AMD doesn't quite trump Intel at least they're back in the game(maybe, bench for waitmarks and all), and since all I really care about is value then they're worthy of some serious consideration.
 
Hey hey with some competition maybe they'll lower the price of the 6950X to $1600, what a steal at that price!

 

[Tiddlesworth]

   
Yup it looks promising, even if AMD doesn't quite trump Intel at least they're back in the game(maybe, bench for waitmarks and all), and since all I really care about is value then they're worthy of some serious consideration.

95w Ryzen beating 140w 6900k was pretty impressive tbh, at least we know the IPC is there, the only worrying thing is the clock speed which they locked. And pricing of course.

 

[214324]

They only locked the clockspeed for demo purposes. We can usually assume the 6900K was running at turbo speeds because, under sufficient cooling, there's no reason for the CPU to not try to run at max frequency when under full load. They didn't specify a turbo speed so as to what the maximum clock out of the box is we won't know (for Zen). They did say it was variable based on their SenseMI rundown.
 
What I don't really understand is so special is the automatic turbo headroom. It just sounds like a fancy dressup for thermal throttling; if your cooling system can remove enough heat, the CPU will run at higher turbo clocks. CPUs already do that so I don't see what's so special that makes them think they have to mention it. Some people are calling it an auto-OC but that's just what Turbo is, just that it's only meant for short term.
 
Good to see that they have an answer to Speedstep.
 
One thing I don't think is a great test is Handbrake (I stopped watching as soon as they showed the VR demo but Tang said there was a Keyshot demo). Broadwell-E is missing the full Quicksync implementation used by Skylake and Kaby Lake. IIRC Broadwell did have logic dedicated to video decoding but Skylake has H.264/265 specific logic (take this with a grain of salt, I'd have to spend some time digging into the Anand articles). It's definitely a good test but I think a more accurate comparison would be a 4 core/8 thread Zen CPU against an i7 Skylake (since Kaby Lake desktop parts aren't out yet).
 
Still it's good to see Jim Keller was able to work his magic at AMD.
 

---

 
Although why call it Ryzen? Ryze?
 

 
If we use some silly logic, every time Riot reworks Ryze he becomes busted. Jim Keller is reworking AMD's CPU architecture, so Zen confirmed busted?
 

 

[deadlylover]

 
Still it's good to see Jim Keller was able to work his magic at AMD.

 
Even at 3.4GHz base frequency that's sorta enough, the 6900k has a base of 3.2GHz and I think in the handbrake test it couldn't turbo at all due to power limits, which is to be expected running a processor at full tilt at the PL1 or PL2 short term power limits.
 
Apart from thermal throttling or dropping out of turbo, I don't think the Intel turbo boost is aware of the temperature, so maybe Ryzen can up the power limits and surpass its rated turbo clock speed temporarily if it knows there's still thermal headroom.
 
IIRC, quicksync isn't a magic bullet, there is a (sometimes significant) quality loss from using it so it's a 'shortcut or alternative' rather than a solution or boost to h264 encoding.

 

[214324]

   
Even at 3.4GHz base frequency that's sorta enough, the 6900k has a base of 3.2GHz and I think in the handbrake test it couldn't turbo at all due to power limits, which is to be expected running a processor at full tilt at the PL1 or PL2 short term power limits.
 
Apart from thermal throttling or dropping out of turbo, I don't think the Intel turbo boost is aware of the temperature, so maybe Ryzen can up the power limits and surpass its rated turbo clock speed temporarily if it knows there's still thermal headroom.
 
IIRC, quicksync isn't a magic bullet, there is a (sometimes significant) quality loss from using it so it's a 'shortcut or alternative' rather than a solution or boost to h264 encoding.

They did say it was an unadulterated chip (they seemed like they really wanted to be honest about these tests instead of trying to tarnish their name by pulling some test cheats like the entire car industry) so if it did not run at the max turbo it probably at least did go up a few steps to 3.4-3.6GHz. If it was running at the base clock it has to be in the P1 state at the very least so Speed Shift probably took some state between P0 and P1 if not P0 itself. I don't see why it shouldn't run at P0 since the whole selling point of Speed Shift was the ability to switch really quickly between states to have a more responsive system.
 
Turbo boost is linked to the temp sensor for throttling (as we see on the mobile parts a lot of the time). I think AMD is trying to differentiate Turbo from SenseMI because Turbo is only meant to be a short-term boost; generally Intel chips from what I've seen don't run at turbo all the time because it can introduce a lot of clock-based errors (WHEA_UNCORRECTABLE_ERROR which happens really often on my laptop even though it doesn't even run at max turbo) just from flip-flops or clock-gated logic not working fast enough all the time. Anyways SenseMI seems less like a Turbo and more like an automatic OC.
 
I think in the end it'll function just like Turbo though. They are adding speedstepping so it's not like the chip will run at those turbo frequencies all the time. There's no reason for it to when it can just drop to a lower state.
 
Quicksync since Skylake is full on dedicated logic for H264 encoding/decoding so there's no shortcuts there. I think Broadwell was like a more general solution for several different types of video codecs (Anand said that Broadwell was meant to be more flexible, Skylake more specific to H264 so kind of like FPGA/reprogrammable vs dedicated codec specific logic). But QuickSync on Skylake and onward shouldn't have as many (or if any) quality problems. I haven't seen any tests regarding that for Skylake and Kaby but if there still is quality loss issues let me know.

 

[deadlylover]

  They did say it was an unadulterated chip (they seemed like they really wanted to be honest about these tests instead of trying to tarnish their name by pulling some test cheats like the entire car industry) so if it did not run at the max turbo it probably at least did go up a few steps to 3.4-3.6GHz. If it was running at the base clock it has to be in the P1 state at the very least so Speed Shift probably took some state between P0 and P1 if not P0 itself. I don't see why it shouldn't run at P0 since the whole selling point of Speed Shift was the ability to switch really quickly between states to have a more responsive system.
 
Turbo boost is linked to the temp sensor for throttling (as we see on the mobile parts a lot of the time). I think AMD is trying to differentiate Turbo from SenseMI because Turbo is only meant to be a short-term boost; generally Intel chips from what I've seen don't run at turbo all the time because it can introduce a lot of clock-based errors (WHEA_UNCORRECTABLE_ERROR which happens really often on my laptop even though it doesn't even run at max turbo) just from flip-flops or clock-gated logic not working fast enough all the time. Anyways SenseMI seems less like a Turbo and more like an automatic OC.
 
I think in the end it'll function just like Turbo though. They are adding speedstepping so it's not like the chip will run at those turbo frequencies all the time. There's no reason for it to when it can just drop to a lower state.
 
Quicksync since Skylake is full on dedicated logic for H264 encoding/decoding so there's no shortcuts there. I think Broadwell was like a more general solution for several different types of video codecs (Anand said that Broadwell was meant to be more flexible, Skylake more specific to H264 so kind of like FPGA/reprogrammable vs dedicated codec specific logic). But QuickSync on Skylake and onward shouldn't have as many (or if any) quality problems. I haven't seen any tests regarding that for Skylake and Kaby but if there still is quality loss issues let me know.

 
No, I think in the video it was shown that the 6900k could not turbo past its base clock, encoding is a power hog. The whole concept of turbo is to increase frequencies during lower loads, so it makes perfect sense that a chip cannot turbo under full stress. They might have mentioned that Ryzen did not turbo either because it was still being tweaked, but even then, we don't expect it to perform past base frequencies for encoding.
To be fair, I don't expect a 8C/16T chip to be able to turbo that high anyway, and perhaps it's not toooo important going forward as more and more games become multithreaded, we're still very GPU bound and will be for a long long time, maybe even forever.
 
For mobile chips I think its a bit more confusing because it's not 'turbo' that gets us the extra boost in clock speed for sustained loads, it's actually the configurable TDP function. The i5-6300U I have will happily play at 2.9GHz all day long with a base clock of 2.4GHz (going from memory here), but I don't think it's correct to say that it's being "turbo boosted" to 2.9GHz, it is instead just able to sip that sweet sweet 25W PL2 limit indefinitely barring thermal throttling and such compared to the 15W PL1 limit. It will turbo boost to 3.0GHz in single core loads.
And I do think it's the configurable TDP function that throttles a chip on a steaming hot ultrabook, it is aware of the temperature but I don't think the turbo boost function is.
 
edit: nvm I'm dum, it seems like the 6900k can all core boost to 3.5GHz. 

 This is getting real confusing ahahahahah.
 
All hardware accelerated encoding lowers quality compared to CPU only encoding, so it is a compromise of performance vs quality. You don't see anyone serious using a Skylake chip's Quick Sync for video encoding, they usually rely on the much larger GPU encoding engines to deliver higher performance without sacrificing too much quality. It does seem that in all of the damn Quick Sync info slides and stuff, it's meant for "casuals" converting videos for their phones and such, Quick Sync is basically useless for professional applications. That's probably why it seems like no reviewer really talks about Quick Sync past a Handbrake benchmark or two. Skylake only added some extra bit depth and h265/HEVC encoding/decoding support.
 

 
Yeah it does seem like some sort of auto-OC, good for the peasants I guess. 

 

 

[Tiddlesworth]

http://www.anandtech.com/show/10907/amd-gives-more-zen-details-ryzen-34-ghz-nvme-neural-net-prediction-25-mhz-boost-steps

Seems like their boosting tech can do increments of 25mhz as long as thermal and power allows it.
 
Kinda like modern gpu boost?

 

[jgray91]

Ah yes, the occasional burst of techno babble. My favorite topic.

 

[kaixax555]

Excited for Ryzen even though I am not gonna buy it
 
I hope AMD does good