*** AMD "Zen" thread (inc AM4/APU discussion) ***

[Silent_Scone]

What's the stock speed on a 6850k?

As above, but if you're having to ask that, you're missing the point.

Under 3 seconds difference in Blender with two cores less, in an ideal world the 6850K shouldn't be getting that close...Zen needs to overclock like a gem if it's got a hope in hell's chance of beating out BWE, if prices don't reflect.

It's the perfect example if you understand how heavily threaded Blender is...when most real world applications simply aren't.

The average overclock of the Core i7 6800k is 4.2ghz based on over 2200 submissions on HWBot:

http://hwbot.org/hardware/processor/core_i7_6800k/

The Core i7 6900k is 4.35ghz based on just over a 1000 submissions:

http://hwbot.org/hardware/processor/core_i7_6900k/

These include water cooling too and other more extreme forms of cooling.

Over 75% of CPUs tested by ASUS were capable of 4.3 or higher.

 

[Harlequin]

Over 75% of CPUs tested by ASUS were capable of 4.3 or higher.

do you have a citation for your claim?

 

[Silent_Scone]

If I must, based on hundred or more samples...

http://edgeup.asus.com/2016/06/17/broadwell-e-overclocking-guide/


This is well beyond the-actual point, though. The point is that Blender is a highly threaded application, with a 3 seconds difference even there, the real world IPC difference is going to be distinguishable. It needs to overclock.

 

[CAT-THE-FIFTH]

As above, but if you're having to ask that, you're missing the point.

Under 3 seconds difference in Blender with two cores less, in an ideal world the 6850K shouldn't be getting that close...Zen needs to overclock like a gem if it's got a hope in hell's chance of beating out BWE, if prices don't reflect.

It's the perfect example if you understand how heavily threaded Blender is...when most real world applications simply aren't.




Over 75% of CPUs tested by ASUS were capable of 4.3 or higher.

I would rather trust a few 1000 submissions on HWBot - they will be done over a long time over many different motherboards and many different production batches.

Edit!!

HardOCP confirmed the AMD demo results:
http://www.hardocp.com/article/2016/12/15/amds_new_ryzen_cpu_smt_ipc#.WGToeHinzqA

They confirmed Ryzen is on par with BW-E core for core and clock for clock in Blender.

 

[Silent_Scone]

I would rather trust a few 1000 submissions on HWBot - they will be done over a long time over many different motherboards and many different production batches.

CPU-Z validations over actual testing by ASUS in HANDBRAKE of all things?

My word.

Look I'm not trying to twist your niblets but that's seriously warped, dude. Handbrake is one of the most intensive things you can do with a CPU, a CPU-Z validation takes two seconds. Digression, though...

 

[CAT-THE-FIFTH]

CPU-Z validations over actual testing by ASUS in HANDBRAKE of all things?

My word.

Look I'm not trying to twist your niblets but that's seriously warped, dude. Handbrake is one of the most intensive things you can do with a CPU, a CPU-Z validation takes two seconds. Digression, though...

Your the one is twisting things by on purpose by ignoring 2000+ submissions on HWBot.

You are so warped you on purpose want to nullify 1000s of people submitting overclock results just so you can cherry pick a higher average overclock on a few samples tested by one company.

That is an overclock on HWBot on average loads of people are getting in the wild.

That includes on air and on water.

 

[Harlequin]

If I must, based on hundred or more samples...

http://edgeup.asus.com/2016/06/17/broadwell-e-overclocking-guide/


This is well beyond the-actual point, though. The point is that Blender is a highly threaded application, with a 3 seconds difference even there, the real world IPC difference is going to be distinguishable. It needs to overclock.

yes if you make an argument in support of your position then cite a source to support it!

certainly not `if I must`.

 

[Silent_Scone]

Your the one is twisting things by on purpose by ignoring 2000+ submissions on HWBot.

You are so warped you on purpose want to nullify 1000s of people submitting overclock results just so you can cherry pick a higher average overclock on a few samples tested by one company.

That is an overclock on average loads of people are getting in the wild.

I can't help you understand why user submissions based on conditional stability results aren't valid? If anything I'd be doing you a favour by doing so, considering some of the clocks used there aren't realistic.

I thought that would be obvious to anyone with half a loaf, frankly. I take it you know what Handbrake is? I'd like to assume so considering in part, it was what Lisa Sue demonstrated Zen with.

It doesn't even matter, though. All I am saying is Zen needs to overclock...

 

[zoomee]

CPU-Z validations over actual testing by ASUS in HANDBRAKE of all things?

My word.

Look I'm not trying to twist your niblets but that's seriously warped, dude. Handbrake is one of the most intensive things you can do with a CPU, a CPU-Z validation takes two seconds. Digression, though...

Yes, I would far rather trust real-life HWBOT submissions than something Asus say. Asus are full of poopookaka with half of their claims. Products which can't do what they say on the tin, functionality being removed, boards having sneaky revision changes. Obviously they are going to inflate the figures to make it look like their product somehow offers something far superior to other manufacturers.

I'm saying this from experience on X79 platform and what Asus claimed most X79 chips could do - they were plenty out from the actual end-user reality (probably had hand picked chips supplied by Intel to make it look good).

 

[CAT-THE-FIFTH]

yes if you make an argument in support of your position then cite a source to support it!

certainly not `if I must`.

HWBot is the biggest data point we have for the overclocks - it's done over multiple motherboards and CPU batches.

I love how he wants to use Asus as a standard but then since they are a gold standard but a data point which is a few times larger he ignores it.

Plus even Asus says 4.3GHZ is the average overclock they saw - HWBot says 4.2GHZ on 2226 samples.

 

[flopper]

Ryzen is more than competitive its why people (Intels) have a hard time with its results shown.

 

[Silent_Scone]

HWBot is the biggest data point we have for the overclocks - it's done over multiple motherboards and CPU batches.

I love how he wants to use Asus as a standard but then since they are a gold standard but a data point which is a few times larger he ignores it.

Plus even Asus says 4.3GHZ is the average overclock they saw - HWBot says 4.2GHZ on 2226 samples.

I love how you're ignoring the fact we don't know if Zen can even overclock yet? Which is the point...

HWBOT results are based on users who are gauging stability on a multitude of tests (not to mention multiple results from the same user/CPU) that don't constitute anywhere near unconditional. If ASUS are running Handbrake to test stability on ALL samples, rather than a multitude of tests, then this is more conclusive than Joey running a CPU-Z Validation for a few seconds.

Clearly.


It's easy to throw big numbers like two thousand around when you don't understand what they consist of.

Yes, I would far rather trust real-life HWBOT submissions than something Asus say. Asus are full of poopookaka with half of their claims. Products which can't do what they say on the tin, functionality being removed, boards having sneaky revision changes. Obviously they are going to inflate the figures to make it look like their product somehow offers something far superior to other manufacturers.

I'm saying this from experience on X79 platform and what Asus claimed most X79 chips could do - they were plenty out from the actual end-user reality (probably had hand picked chips supplied by Intel to make it look good).

I don't care for tinfoil, all I know is their boards do the numbers in all the places they should, plus for an overclocking overview there's no reason to lie, and it's consistent with the several CPU's I've had here.

 

[Silent_Scone]

Also for what it's worth, if you use the same runtime version as AMD did which is 2.77, I can render the scene quicker, at 35.22

http://i.imgur.com/zLkAcQq.jpg

 

[FoxEye]

This was posted last month on Hexus:

http://hexus.net/tech/news/cpu/99400-amd-zen-chips-debut-17th-january-chinese-mobo-maker/

MAXSUN is one of the motherboard partners AMD has in China. Having said that,they said 4.2GHZ for "normal" overclock for their ES.

Interestingly they mention the CPU they have has the same frequencies as the ES Canard PC tested. OTH,Lisa Su said they were launching at 3.4GHZ+ for Ryzen.

It makes me wonder whether the delay is to boost clockspeeds higher.

Hmm, AMD's equivalent of the Intel "K" variant CPUs to cost (up to) twice as much as non-K?!

I hope they haven't been studying Intel's play book too closely after all.

 

[r7slayer]

I love how you're ignoring the fact we don't know if Zen can even overclock yet? Which is the point...

HWBOT results are based on users who are gauging stability on a multitude of tests (not to mention multiple results from the same user/CPU) that don't constitute anywhere near unconditional. If ASUS are running Handbrake to test stability on ALL samples, rather than a multitude of tests, then this is more conclusive than Joey running a CPU-Z Validation for a few seconds.

Clearly.


It's easy to throw big numbers like two thousand around when you don't understand what they consist of.





I don't care for tinfoil, all I know is their boards do the numbers in all the places they should, plus for an overclocking overview there's no reason to lie, and it's consistent with the several CPU's I've had here.

Zen can overclock we already know this. It's called boost and based on thermal and power etc it can boost to certain speeds just like how nvidia cards do. Motherboard vendors will also try their best to provide options and methods to boost your CPU speeds as its usually a strong selling point for motherboard vendors on higher ends otherwise might as just get a bog standard motherboard.

 

[r7slayer]

Hmm, AMD's equivalent of the Intel "K" variant CPUs to cost (up to) twice as much as non-K?!

I hope they haven't been studying Intel's play book too closely after all.

Thats for the SR7 which will be the 8 core 16 thread variant which is completley different to the mainstream 6700k variant. It's more likley going to compete with the 6900k which is what £1000 roughly? This means the SR7 OC top end chip is going to be significantly cheaper. For 6700k your going to be looking at SR5 which is likley to be the 4core 8threads and this will be priced around the 6700k area.

 

[Silent_Scone]

Zen can overclock we already know this. It's called boost and based on thermal and power etc it can boost to certain speeds just like how nvidia cards do. Motherboard vendors will also try their best to provide options and methods to boost your CPU speeds as its usually a strong selling point for motherboard vendors on higher ends otherwise might as just get a bog standard motherboard.

If we're being pedantic, yes lol . I'm talking about it overclocking well. Enough to gain it some traction against BWE parts. Again, though, price will put this into perspective also.

 

[FoxEye]

Thats for the SR7 which will be the 8 core 16 thread variant which is completley different to the mainstream 6700k variant. It's more likley going to compete with the 6900k which is what £1000 roughly? This means the SR7 OC top end chip is going to be significantly cheaper. For 6700k your going to be looking at SR5 which is likley to be the 4core 8threads and this will be priced around the 6700k area.

Well the article is clearly nonsense anyhow, as it says "most SR7 chips will be between $250 and $300".

Which leaves SR5 and SR3 crammed into a very small price range.

And they'd never sell SR7 that cheaply if it walks all over a £350 Intel 7700k.

So, file that one under "wccf".

 

[Davedree]

Well the article is clearly nonsense anyhow, as it says "most SR7 chips will be between $250 and $300".

Which leaves SR5 and SR3 crammed into a very small price range.

And they'd never sell SR7 that cheaply if it walks all over a £350 Intel 7700k.

So, file that one under "wccf".

Which is why amd need to get their arse in gear, write off bristol ridge and get raven ridge out.

 

[~>Dg<~]

Ryzen is more than competitive its why people (Intels) have a hard time with its results shown.

no as pointed out ten times its because we have seen nothing of note we seen half a benchmark for blender that is it.

we dont know the clockspeed.the actual speed in other programs.what it oc to.how much its going to be.

so call me picky but ... nothing #!

amd have released before a cpu at 1000 quid ! it wasnt competitive . quickly dropped.if the cpu is roughly what it might be has to be priced under £400.

5820k ocd and a 6800k ocd can get the same results.or very similar.if its quicker in blender than the above two chips and slower at the rest its still a fail even if priced the same.if its slightly cheaper but same results well we had this power for a good while already.

so pricing is key.