AMD New Horizon

[DocSwag]

18 minutes ago, Space Reptile said:

while looking for the ryzen logo you saw on stream (if somebody has a hq cap pls tell me , i like that logo) 

i found this  : https://trademarks.justia.com/871/15/amd-87115584.html

 

im suprised this wast picked up earlyer , could have known the name of zen , "ryzen" 2 weeks ago  

It actually was picked up. There was a post on these forums in tech news and reviews about like 6 or 7 names AMD registered including ryzen and threadripper. We just didn't knnow at the time what they all meant.

 

EDIT: Here it is 

 

[Space Reptile]

4 minutes ago, DocSwag said:

It actually was picked up. There was a post on these forums in tech news and reviews about like 6 or 7 names AMD registered including ryzen and threadripper. We just didn't knnow at the time what they all meant.

yea i just saw that , @SpaceGhostC2C beat you to it 

[done12many2]

1 hour ago, SpaceGhostC2C said:

Matching 6900K performance and selling at $350 can't coexist. The reason is that in that case I, and others like me, would buy all their chips the moment they are out, hoard them, and resell them at a huge profit like they're tickets to a sold out concert.

 

I wish that were the case, but it's beyond obvious that AMD skewed the Blender render test.  Not even stock 6900k CPUs or stock 6950x CPUs can match the render times that AMD was able to pull off during the AMD Horizon event.  

 

I'm a fairly realistic guy and understand that companies hype things up a bit during promotions and roll outs, but what AMD did this time was different.  They showed the results of the render test and even offered to provide the files to the public so that they could run the same test on their own computers for comparison.  The problem is they used custom settings to run their bench without mentioning it.  This resulted in much higher render times for everyone who's run it on their own systems to include the 6950x, 6900k (overclocked and stock), 5960x (overclock and stock) and many other CPUs.  Guys with 6900ks overclocked and running better system configurations couldn't even match what AMD was able to do with a 3.4 GHz Ryzen chip or their "stock" 6900k.  

 

There's a running thread on overclock.net where everyone is trying to match AMD's magical render times with the exact same file provided by AMD via their website.  They make no mention of modified settings to Blender, but mention that both system configurations are run as they are straight out of the box.

 

Additionally, they stated that you'd be able to run both tests for yourself, but didn't provide the Handbrake file for comparison at all.  We're really no better off as far as a good understanding on how Zen will perform.  Like I said, feel free to check out the overclock.net thread where folks are tying to pull off the amazing feat that AMD has supposedly accomplished with that test.

 

http://www.overclock.net/t/1617227/amd-new-horizon-zen-preview-on-12-13-at-3-pm-cst/600_100#post_25710833  (Skip towards the end where the results start showing up)

 

Anyways, here's my own comparison, which convinced me that they were blatantly skewing the render test.  

 

 

 

 

 

 

 

[SpaceGhostC2C]

12 minutes ago, done12many2 said:

 

I wish that were the case, but it's beyond obvious that AMD skewed the Blender render test.  Not even stock 6900k CPUs or stock 6950x CPUs can match the render times that AMD was able to pull off during the AMD Horizon event.  

 

 

But notice I make a conditional statement: if that's the performance, that can't be the price. If that's the price, that can't be the performance. Then it all may or may not happen, I just don't think both can happen simultaneously

 

Now, independently of that, you are showing an OCed 5960x basically do the same that the 6900K and Ryzen in the video. It's hard for me to tell whether any advantage the 6900K may have over the 5960x is more or less than what you gain by overclocking to 4.8. They could have uploaded a different job to make any CPU you have pale in comparison  But that would be silly, as people with a 6900K could quickly check whether they match the mark or not, since it's supposed to be done at stock speeds. I don't know if there's any other specs (RAM?) that could matter, or whether the background tasks you mentioned can have a significant impact.

 

Do you know (or is it too much trouble for you to go back and forth) how much time would it take for your 5960x to do the same job? 

 

PS: a quick Youtube search tells me not many people has replicated yet. Then again, the quick YT search didn't return your video... 

[Clanscorpia]

24 minutes ago, themctipers said:

Please be cheap.

-Canada with our $550 i7 6700k and $400 RX 480

I got mine for $280

[themctipers]

1 hour ago, Clanscorpia said:

I got mine for $280

Motherfuc-

 

lets hope that the holidays brings me a cheap RX 480

Or a high USD -> CAD but same prices (I need that extra 4 cents / dollar for exchanging $220..)

[skyler_mertz]

6 minutes ago, done12many2 said:

 

-snip-

 

 

 

 

 

You ran your own test on a 5960X running 4.8 GHz while Zen was running 29% slower at 3.4GHz and achieved a render time of 35 seconds where Zen is at 51 Seconds making it about 31% faster than Zen. Also, Broadwell IPC and Haswell IPC are equivalent within the margin of error AFAIK. I could probably also assume you are running DDR4 in quad channel at probably a frequency higher than 2133mhz? Could also factor in with blender tests.

 

In other leaks like the one from the asian site a week or so ago showed a list of cinebench scores putting it at ~1300 right with the 6900K yet you scored a ~1900. 

 

So I'd say, quantitatively, Zen is pretty accurate at least within the margin of error to the blender tests given.

[done12many2]

1 minute ago, SpaceGhostC2C said:

Now, independently of that, you are showing an OCed 5960x basically do the same that the 6900K and Ryzen in the video. It's hard for me to tell whether any advantage the 6900K may have over the 5960x is more or less than what you gain by overclocking to 4.8.

 

I've actually owned a 6800k and 6950x and other than the fact that a BWE chip requires just over 200 MHz less to perform the same work as a HWE chip, they scale the same.

 

1 minute ago, SpaceGhostC2C said:

 

They could have uploaded a different job to make any CPU you have pale in comparison  

 

  I believe so or the setting were run differently and we weren't informed.  

 

1 minute ago, SpaceGhostC2C said:

But that would be silly, as people with a 6900K could quickly check whether they match the mark or not, since it's supposed to be done at stock speeds. I don't know if there's any other specs (RAM?) that could matter, or whether the background tasks you mentioned can have a significant impact.

 

Silly?  Please go to the link I provided for overclock.net.  Nobody with stock or overclocked 6950x or 6900k chips can match AMD's unbelievable results.  It has nothing to do with my own comparison.  Nobody can do it.  Actually one guy with a 6950x beat it at 4.4 GHz, but that was the only one out of many examples.

 

1 minute ago, SpaceGhostC2C said:

Do you know (or is it too much trouble for you to go back and forth) how much time would it take for your 5960x to do the same job? 

 

Can you clarify?

[SpaceGhostC2C]

1 minute ago, skyler_mertz said:

You ran your own test on a 5960X running 4.8 GHz while Zen was running 29% slower at 3.4GHz and achieved a render time of 35 seconds where Zen is at 51 Seconds making it about 31% faster than Zen. 

That was Handbrake. The Blender test ended in 35 seconds as well (first part of his video).

[done12many2]

8 minutes ago, skyler_mertz said:

You ran your own test on a 5960X running 4.8 GHz while Zen was running 29% slower at 3.4GHz and achieved a render time of 35 seconds where Zen is at 51 Seconds making it about 31% faster than Zen. Also, Broadwell IPC and Haswell IPC are equivalent within the margin of error AFAIK. I could probably also assume you are running DDR4 in quad channel at probably a frequency higher than 2133mhz? Could also factor in with blender tests.

 

In other leaks like the one from the asian site a week or so ago showed a list of cinebench scores putting it at ~1300 right with the 6900K yet you scored a ~1900. 

 

So I'd say, quantitatively, Zen is pretty accurate at least within the margin of error to the blender tests given.

 

You're in a different conversation bud, but thanks for your input.

 

I'd like to see anyone on LTT give it a shot.  It's not doable.  You cannot match the performance of the Ryzen or 6900k results whether you are overclocked or not.  They ran a different configuration on the test than what they provided to the public for their own comparison and now I doubt that the configurations between the Ryzen and 6900k were alike.

[SpaceGhostC2C]

2 minutes ago, done12many2 said:

Silly?  Please go to the link I provided for overclock.net.  Nobody with stock or overclocked 6950x or 6900k chips can match AMD's unbelievable results.  It has nothing to do with my own comparison.  Nobody can do it.  Actually one guy with a 6950x beat it at 4.4 GHz, but that was the only one out of many examples.

I missed the link!

 

2 minutes ago, done12many2 said:

Can you clarify?

I'm asking it returning to stock clocks and running the test again is too much hassle for you, or if you can quickly go to stock and then return to your OC without issues. But now that I've seen the link is not as important, I can look for more results there  I just wanted to get a sense of how much you gain with the extra frequency in this test, other things equal.

[done12many2]

9 minutes ago, SpaceGhostC2C said:

I'm asking it returning to stock clocks and running the test again is too much hassle for you, or if you can quickly go to stock and then return to your OC without issues. But now that I've seen the link is not as important, I can look for more results there  I just wanted to get a sense of how much you gain with the extra frequency in this test, other things equal.

 

Gotcha and thanks for clarifying.  After you come back from that thread, you'll know what I mean.    Those guys over there do nothing but overclock and compare.  They do a great job of trying to break down how they suspect AMD did it, but it had to do with the run configuration and possibly more. 

 

I should add, none of this would be out of the standard as far as marketing practices are concerned, but this time they went a little too far by telling us to compare it for ourselves with the exact same file.  BEYOND misleading.

 

19 minutes ago, SpaceGhostC2C said:

PS: a quick Youtube search tells me not many people has replicated yet. Then again, the quick YT search didn't return your video... 

 

I set all my videos to unlisted.  I'm not trying to run my own channel or anything. 

[Mooshi]

I don't need another pc, but I kinda wanna build one just to have an AMD rig. 

[SpaceGhostC2C]

4 minutes ago, done12many2 said:

 

Gotcha and thanks for clarifying.  After you come back from that thread, you'll know what I mean.    Those guys over there do nothing but overclock and compare.  They do a great job of trying to break down how they suspect AMD did it, but it had to do with the run configuration and possibly more. 

No doubt I did

It's still strange to understand, because they didn't even mentioned the time, only the comparison. So, whatever rigging they made, they had to make it so both PCs end at the same time. As long as they could show that (after all, we're taking their word for it, they could be running two Ryzens ), what is the point in not giving us the same project? No one can repeat the head to head after Ryzen is out anyway.

[vanished]

47 minutes ago, themctipers said:

Please be cheap.

-Canada with our $550 i7 6700k and $400 RX 480

Stop spreading that around, it's not even true!  

 

http://www.memoryexpress.com/Products/MX58496

Regular $500, but currently (and almost always) on sale for $470.  Yes it's too much but at least it isn't $550  

 

And most of the 480s are around $360, give or take.  Some almost as low as $300, and, yes, some over $400 but most are not.

[Prysin]

5 hours ago, Energycore said:

Yup, at least it seems like it will compete.

 

But I don't care for the 8-core, I want the 4-core! I want to look at how the $150 part does

what?! you dont want to spend close to 1k USD on the CPU alone?!?!?! You're such a cheapskate!

[done12many2]

7 minutes ago, SpaceGhostC2C said:

No doubt I did

It's still strange to understand, because they didn't even mentioned the time, only the comparison. So, whatever rigging they made, they had to make it so both PCs end at the same time. As long as they could show that (after all, we're taking their word for it, they could be running two Ryzens ), what is the point in not giving us the same project? No one can repeat the head to head after Ryzen is out anyway.

 

Well by not giving us the same project they skew the results that you'll get.  As you've already seen, if you run the test for yourself on your own computer as they suggested you do, there's a 99.9% chance that not only will your render times be much higher, but that you'll feel as if Ryzen is that much better than what you currently have.

 

It's only going to end in disappointments when actual reviews fall short of these exaggerated results. 

 

I really do want LTT 6900k and 6950x owners to run the test for themselves.  You won't come close without a very strong overclock.

[Prysin]

3 minutes ago, Ryan_Vickers said:

Stop spreading that around, it's not even true!  

 

http://www.memoryexpress.com/Products/MX58496

Regular $500, but currently (and almost always) on sale for $470.  Yes it's too much but at least isn't not $550  

 

And most of the 480s are around $360, give or take.  Some almost as low as $300, and, yes, some over $400 but most are not.

ryan, please. you speak of a difference of max 110 CAD... that's worth like 8 acorns and half a slice of bread..

[vanished]

1 minute ago, Prysin said:

ryan, please. you speak of a difference of max 110 CAD... that's worth like 8 acorns and half a slice of bread..

Well that may be true but these aren't even the best prices I've seen... it's just what it happened to be floating at for the moment.

[Prysin]

truth be told. Whilst the hype is way too high, what we've seen today/tonight, gives me hope for their APUs

[done12many2]

Just now, Prysin said:

truth be told. Whilst the hype is way too high, what we've seen today/tonight, gives me hope for their APUs

 

I agree with that and hope for competition in general.  We all gain in the end. 

 

I just don't agree with the bullshit they pulled on skewing the results and then asking us to try it for ourselves.

[That Norwegian Guy]

6 hours ago, DocSwag said:

Ya, 8 core is nice and all, and it's cool that Zen is actually on par with Intel's best offerings, but most consumers don't care about 8 cores! We want 4 cores! We want 4 cores! Now join me! We want 4 cores! We want 4 cores!

6 core or GTFO. That's the optimized count for Star Citizen.

[Dietrichw]

I am curious about the cooling they used on the test systems, I remember when they panned away I could see the glow of a stock 2011 cooler.

ex.

Spoiler

So it could be that the 6900k was thermal throttling then. 

[vanished]

1 minute ago, Dietrichw said:

I am curious about the cooling they used on the test systems, I remember when they panned away I could see the glow of a stock 2011 cooler.

ex.

So it could be that the 6900k was thermal throttling then. 

I hope not, that would be very unfair...

[TheNaitsyrk]

Agaaaain, feeling like shit for buying 6800k on x99 platform.

 

If the Ryzen is going to outperform my 6800k for a better price I might be saying good bye to my 3 month old mobo and CPU.

 

Too bad they didn't include prices.