Ryzen Overclocking and Simultaneous Multithreading Info

[leadeater]

3 minutes ago, outercry said:

Oh well, waiting for the next hype train, possibly Vega...

Due to the amount of BS and thread spam etc when it comes round to Vega pre-launch I or we the mod team will likely look at cracking down on this. There have been too many threads, many sudo different topics but really the same thing referencing the same source(s) etc.

 

Reddit posts wouldn't normally qualify as an allowed source for news topics either, not for things like CPUs and GPUs, but as above so many damn threads made it hard to enforce this and would have been easier to stop if that was something we had to plan on doing from the start.

[outercry]

1 minute ago, leadeater said:

Due to the amount of BS and thread spam etc when it comes round to Vega pre-launch I or we the mod team will likely look at cracking down on this. There have been too many threads, many sudo different topics but really the same thing referencing the same source(s) etc.

 

Reddit posts wouldn't normally qualify as an allowed source for news topics either, not for things like CPUs and GPUs, but as above so many damn threads made it hard to enforce this and would have been easier to stop if that was something we had to plan on doing from the start.

I'm really glad to hear that. I do think the sheer amount of "different" threads about the same thing basically have made these forums a bit of a mess to be honest. So yeah, really glad to hear you will be tackling this with Vega etc., because it will be inevitable.

[XenosTech]

3 minutes ago, leadeater said:

Due to the amount of BS and thread spam etc when it comes round to Vega pre-launch I or we the mod team will likely look at cracking down on this. There have been too many threads, many sudo different topics but really the same thing referencing the same source(s) etc.

 

Reddit posts wouldn't normally qualify as an allowed source for news topics either, not for things like CPUs and GPUs, but as above so many damn threads made it hard to enforce this and would have been easier to stop if that was something we had to plan on doing from the start.

#ModRage

 

I feel you though, used to work as a mod for a game forum, I used to be tearing out my hair at the amount of threads for a game bug that poped up when there was learly one we had pinned for it already and all people had to do was read it >.>

[Techicolors]

6 minutes ago, Bananasplit_00 said:

its not fun to overclock "just a bit". thats why i find it a dissapointment. we dont have any benchmarks or real overclocks yet but i would find 4.4GHz to be a bit lacking. also havent talked to a single person who hasent been able to push there I5 4690K above 4.4GHz, most iv talked to can get 4.6GHz or so but i havent talked to 2000 people i gues

i have a 4690K overclocked to 4.6ghz. could probably go higher but i have to add a lot of voltage 

[leadeater]

2 minutes ago, Bananasplit_00 said:

its not fun to overclock "just a bit". thats why i find it a dissapointment. we dont have any benchmarks or real overclocks yet but i would find 4.4GHz to be a bit lacking. also havent talked to a single person who hasent been able to push there I5 4690K above 4.4GHz, most iv talked to can get 4.6GHz or so but i havent talked to 2000 people i gues

You shouldn't expect such high frequency from 6/8+ core CPUs, this just isn't possible due to power and heat reasons. Has nothing to do with AMD or Intel, physics/science must be obeyed . I don't run my 4930k over 4Ghz 24/7 ever, I can run 4.3 to 4.4 but it's not exactly fully stable or something I'd leave on for any longer than I'd need to which to be honest is never as no game needs it.

 

Plus using raw numbers like this is just bad, use percentage. A 4.4Ghz OC might be amazing for a Zen based CPU, if base clock was 2.8GHz. It's not but illustrates the point.

[outercry]

Oh and just to clarify, I'm not having a go at the mods or anything like that, just wishing people wouldn't re-post the same things under different names over and over again.

Like XenosTech said above, being a mod isn't easy, especially when running a huge forum like this.

[Dabombinable]

Sandybridge does 5GHz on air, Haswell does 4.8GHz despite the FIVR, and then you have the FX series which manages the same. Big woop.

[leadeater]

1 minute ago, outercry said:

Oh and just to clarify, I'm not having a go at the mods or anything like that, just wishing people wouldn't re-post the same things under different names over and over again.

Like XenosTech said above, being a mod isn't easy, especially when running a huge forum like this.

Oh we know you aren't but I do fully agree with what you said.

[Guest Darth Revan]

Does anyone know if Ryzen will have dedicated h265 decoding like Kaby Lake? You know like the whole DRM with windows 10 and Netflix, if you want to view 4K content from Netfilx you need the Kaby Lake decoder and Win10.

[MageTank]

8 minutes ago, Bananasplit_00 said:

its not fun to overclock "just a bit". thats why i find it a dissapointment. we dont have any benchmarks or real overclocks yet but i would find 4.4GHz to be a bit lacking. also havent talked to a single person who hasent been able to push there I5 4690K above 4.4GHz, most iv talked to can get 4.6GHz or so but i havent talked to 2000 people i gues

I consider myself to be a competent overclocker, and the 4690k sitting behind me was only able to hit 4.4ghz. It's voltwage wall happened at 1.25v, and even at 1.3v, couldn't hit 4.5. This is with a PH-TC14PE strapped to it. It's not limited by thermals either, it simply falls unstable after 4.4. If you'd like to count the 4690k's and get the exact average, by all means, I won't stop you, lol: http://www.overclock.net/t/1490324/the-intel-devils-canyon-owners-club

 

As for it not being fun to overclock "just a bit", what do you mean by this? Going from 3.3 to 4.4 (if this extremely vague rumor is to be believed) just as impressive as going from 3.5 to 4.7? With that logic, the 4790k is a disappointment because it only goes from 4.0 to 4.7 on average (technically 4.4 if you compare it's all-core boost). If you compare the 3.9ghz all-core boost of the 4690k to the rumored 3.7ghz boost of the 1600x, your chip and "presumably" this Ryzen chip get an extra 800mhz out of overclocking.

 

People are boasting about these new Kaby lake chips hitting 5.2ghz, but their base is 4.5ghz. Again, only an extra 700mhz. It seems as if you are only quoting what Linus said from the WAN show, so I'll say this: We have not had chips with huge overclocking windows since Sandy/Ivy. Sure, we had the G3258's and the locked Skylake chips, but those are a different story entirely, and would certainly not hold a candle to even average-clocked, higher-thread CPU's. 

[Techicolors]

5 minutes ago, Darth Revan said:

Does anyone know if Ryzen will have dedicated h265 decoding like Kaby Lake? You know like the whole DRM with windows 10 and Netflix, if you want to view 4K content from Netfilx you need the Kaby Lake decoder and Win10.

Ryzen doesn't have an iGPU, but AMD may support the DRM with their APUs 

[Bananasplit_00]

6 minutes ago, leadeater said:

You shouldn't expect such high frequency from 6/8+ core CPUs, this just isn't possible due to power and heat reasons. Has nothing to do with AMD or Intel, physics/science must be obeyed . I don't run my 4930k over 4Ghz 24/7 ever, I can run 4.3 to 4.4 but it's not exactly fully stable or something I'd leave on for any longer than I'd need to which to be honest is never as no game needs it.

 

Plus using raw numbers like this is just bad, use percentage. A 4.4Ghz OC might be amazing for a Zen based CPU, if base clock was 2.8GHz. It's not but illustrates the point.

well the thing is that we basically have the base clocks confirmed at this point, haveing multiple independant sources confirming the same info, the ones we have seen have all been running a base of about 3.5GHz and the top end chip getting closer to 4GHz with a 4GHz base turbo. ofcource an architecture with a lot of cores will clock lower but seeing as they are rated for relativly low TDPs they are also going to be efficient if they have the preformance thats promised. its not a lack of preformance really, 4.4GHz with Broadwell-E IPC and 8c/16th is great, but i just find the overclock dissapointing personaly.

 

8 minutes ago, MageTank said:

I consider myself to be a competent overclocker, and the 4690k sitting behind me was only able to hit 4.4ghz. It's voltwage wall happened at 1.25v, and even at 1.3v, couldn't hit 4.5. This is with a PH-TC14PE strapped to it. It's not limited by thermals either, it simply falls unstable after 4.4. If you'd like to count the 4690k's and get the exact average, by all means, I won't stop you, lol: http://www.overclock.net/t/1490324/the-intel-devils-canyon-owners-club

 

As for it not being fun to overclock "just a bit", what do you mean by this? Going from 3.3 to 4.4 (if this extremely vague rumor is to be believed) just as impressive as going from 3.5 to 4.7? With that logic, the 4790k is a disappointment because it only goes from 4.0 to 4.7 on average (technically 4.4 if you compare it's all-core boost). If you compare the 3.9ghz all-core boost of the 4690k to the rumored 3.7ghz boost of the 1600x, your chip and "presumably" this Ryzen chip get an extra 800mhz out of overclocking.

 

People are boasting about these new Kaby lake chips hitting 5.2ghz, but their base is 4.5ghz. Again, only an extra 700mhz. It seems as if you are only quoting what Linus said from the WAN show, so I'll say this: We have not had chips with huge overclocking windows since Sandy/Ivy. Sure, we had the G3258's and the locked Skylake chips, but those are a different story entirely, and would certainly not hold a candle to even average-clocked, higher-thread CPU's. 

seems more like 4.5GHz for the avrage I5 4690K from that source to me, anyway i was pretty sure i was about an avrage chip on my chip, thats why it seemed so dissapointing, a bit better IPC but quite a bit lower frequency feels kind of meh, still this 4.4GHz is a speculation and we will have to wait and see. also for voltage on my I5 4690K i run 1.335V, try that if you want that CPU higher, it was the magic number for me.

[leadeater]

10 minutes ago, Bananasplit_00 said:

well the thing is that we basically have the base clocks confirmed at this point, haveing multiple independant sources confirming the same info, the ones we have seen have all been running a base of about 3.5GHz and the top end chip getting closer to 4GHz with a 4GHz base turbo. ofcource an architecture with a lot of cores will clock lower but seeing as they are rated for relativly low TDPs they are also going to be efficient if they have the preformance thats promised. its not a lack of preformance really, 4.4GHz with Broadwell-E IPC and 8c/16th is great, but i just find the overclock dissapointing personaly.

Yea this was actually something Linus and Luke were talking about on the WAN show. It's good for the consumer now that CPUs are coming out closer to their maximum potential but it does take away the ability to really OC, then there are things like GPU Boost and potentially XFR that make it even more pointless.

[Bananasplit_00]

2 minutes ago, leadeater said:

Yea this was actually something Linus and Luke were talking about on the WAN show. It's good for the consumer now that CPUs are coming out closer to their maximum potential but it does take away the ability to really OC, then there are things like GPU Boost and potentially XFR that make it even more pointless.

yah im not a big fan of auto overclocking features, sure if they overclock a bit then fine but GPU Boost 3.0 takes over controll really hard from what iv heard(i dont own a pascall card so i dont know) and every time iv seen someone use a motherboard auto overclocking its pushed quite a bit more voltage then needed too.

[Guest Darth Revan]

34 minutes ago, Technicolors said:

Ryzen doesn't have an iGPU, but AMD may support the DRM with their APUs 

Yeah but doesn't kaby lake have a dedicated h256 decoder in it? Or in order to do that it has to use the igpu? So the h256 decoder is integrated in the gpu part.

[Dabombinable]

6 hours ago, Prysin said:

https://en.wikipedia.org/wiki/Simultaneous_multithreading

 

Btw, first "HT" CPU came out in 2004. But there was one planned from another company in 2001, but it didnt launch.

The first CPU to have hyperthreading actually came out in 2002.
http://www.cpu-world.com/CPUs/Pentium_4/Intel-Pentium 4 3.06 GHz - RK80532PE083512 (BX80532PE3066D).html

[Bouzoo]

4 minutes ago, Dabombinable said:

The first CPU to have hyperthreading actually came out in 2002.
http://www.cpu-world.com/CPUs/Pentium_4/Intel-Pentium 4 3.06 GHz - RK80532PE083512 (BX80532PE3066D).html

And a whole bunch of them in 2003.

 

https://en.wikipedia.org/wiki/List_of_Intel_Pentium_4_microprocessors#Pentium_4_HT

 

That 130nm though...

[leadeater]

2 minutes ago, Dabombinable said:

The first CPU to have hyperthreading actually came out in 2002.
http://www.cpu-world.com/CPUs/Pentium_4/Intel-Pentium 4 3.06 GHz - RK80532PE083512 (BX80532PE3066D).html

I had that P4, most useful thing about HT back then was when a thread went rouge and gets stuck at 100%. Previously this would pretty much lock a system up so bad it was hard to even kill the process, HT stopped that.

[Dabombinable]

 

4 minutes ago, Bouzoo said:

And a whole bunch of them in 2003.

 

https://en.wikipedia.org/wiki/List_of_Intel_Pentium_4_microprocessors#Pentium_4_HT

I've got the first LGA 774 Pentium 4, the 540J and it was manufactured in September 2003, far sooner than the August 2004 (also, it is the same speed as a 65W P4 631).

[Bouzoo]

8 minutes ago, Dabombinable said:

 

I've got the first LGA 774 Pentium 4, the 540J and it was manufactured in September 2003, far sooner than the August 2004 (also, it is the same speed as a 65W P4 631).

I do still have my old P4 2.0 GHz in the attic in my other house, god knows which version. I should check it out at some point. Same as P3 and few other goodies. 

[Humbug]

Ryzen overclock potential will get better over the next few months. Keep in mind that this is a brand new design and global foundries will definitely produce better results after experience with mass production.

 

At launch I'm sure there will be some headroom for OC to satisfy us, but do not expect intel rivaling OCs unless you are lucky and get a golden chip.

[Brooksie359]

8 hours ago, Bananasplit_00 said:

yah im not a big fan of auto overclocking features, sure if they overclock a bit then fine but GPU Boost 3.0 takes over controll really hard from what iv heard(i dont own a pascall card so i dont know) and every time iv seen someone use a motherboard auto overclocking its pushed quite a bit more voltage then needed too.

Having auto overclocking features on the pascal gpus makes overclocking them kinda pointless. My gtx 1070 hit 2000 mhz out of the box without messing with anything. It's great for people who don't want to overclock because they get a ton of extra performance but it does make overclocking it yourself less fun and rewarding. 

[Thony]

"Experts say AMD's new Simultaneous Multithreading is actually BETTER than Intel's Hyperthreading."

Im not saying its absolute truth but just a single thought of this happening gave me goosebumps !!! :-O

[Techstorm970]

4 hours ago, Thony said:

"Experts say AMD's new Simultaneous Multithreading is actually BETTER than Intel's Hyperthreading."

Im not saying its absolute truth but just a single thought of this happening gave me goosebumps !!! :-O

If I remember the video correctly, he said something about the cores not having to fight over threads all that much in the Ryzen architecture.  On the other hand, Hyperthreaded Intel cores do have to fight over threads more often than in Ryzen (although still not that much).

 

My broad conclusion: AMD has figured some stuff out over the past 5 years.

[vanished]

17 minutes ago, techstorm970 said:

If I remember the video correctly, he said something about the cores not having to fight over threads all that much in the Ryzen architecture.  On the other hand, Hyperthreaded Intel cores do have to fight over threads more often than in Ryzen (although still not that much).

 

My broad conclusion: AMD has figured some stuff out over the past 5 years.

A slightly more technical explanation of what that even means wouldn't hurt