AMD New Horizon

[Phate.exe]

56 minutes ago, Brian McKee said:

They used 2 different benchmarks, one with integers and one with floating point. The CPU was also intentionally crippled. The only thing that could possibly be skewed is that this is just incredibly binned silicon for the power draw...

If anything it looks like they were trying to skew things in favor of the intel chip (at least in the benches shown).

 

Really want to see how well these things clock.  I don't really see another bulldozer happening, so now we're waiting to find out if we have a merely competitive CPU, or a genuine monster on our hands.

 

Even more interested in how a lower core-count model performs.  The 6700k comparison pretty much just showed us the power of moar coars for that workload.

[Prysin]

2 hours ago, porina said:

My investigation was more testing the test, rather than testing hardware. By working out how the benchmark behaves, we can start to understand where the demo result applies, and where it might not. Because Ryzen is claimed to perform well in this task, doesn't mean it will do so for all types of tasks.

If you mean, does it use specific features of CPUs for performance boost, then I can't say with any certainty. From my testing I can only say it doesn't seem to be taking advantage of AVX. In other forums someone had compiled a modified version of Blender with AVX support and that gives a significant time reduction. I haven't looked at that in detail.

Vectorizing the workload will naturally speed it up unless you are a shit programmer. So it's not really that "shocking" that it increases performance noticeably. Heck, even AMD FX would be notably better at gaming if games were properly vectorized.

[porina]

2 hours ago, MageTank said:

In my Blender test, HT gave me a 49.5% boost, compared to running it without HT. I did use an entirely different file (Using the Blenchmark106 SceneV3 file)

Since there is some run to run variation, even if I pick the best of several runs, I was wondering if perhaps the best HT off result wasn't as good as it could be. However I saw this at both ram speeds tested getting 52% in one, 53% in the other. I don't intend to analyse this further but thought I should mention it anyway.

[done12many2]

2 hours ago, MageTank said:

I actually ran a slew of tests myself, with HT and various ram speeds. Blender was one of the tests I used. 

 

From 2133 to 3200, ram made no difference in the blender tests. This is 2133 C15-15-15-35-2, to 3200 C14-14-14-28-2. We are talking going from 34GB/s and 60ns, to 46GB/s and 46ns. That's a drastic difference in ram performance, for it to make zero difference in Blender. I went back to test it with my 3600 C14-14-14-28-2 profile with tight tertiary timings, and the result was still the same.

 

In my Blender test, HT gave me a 49.5% boost, compared to running it without HT. I did use an entirely different file (Using the Blenchmark106 SceneV3 file)

 

Any cache testing?

 

That's a huge difference in the latency and availability of bandwidth between your 2133 and 3200 configurations.

 

I'm also interested in what you would have notice between single channel and dual.  Seems like single channel would definitely lead to some wasted CPU cycles.

 

26 minutes ago, Phate.exe said:

If anything it looks like they were trying to skew things in favor of the intel chip (at least in the benches shown).

 

 

What makes you think this?

 

[S w a t s o n]

CHILDREN, I thought we settled this. No one was skewing anything towards anyone. Plsopls just let this thread die. You can't really extrapolate any Ryzen perf info from 2 benchmark runs (1 each) and now people are picking apart blender as if it hasnt existed for years.

Everyone just be patient and wait for the final fucking product to come out.

[done12many2]

1 minute ago, Swatson said:

CHILDREN, I thought we settled this. No one was skewing anything towards anyone. Plsopls just let this thread die. You can't really extrapolate any Ryzen perf info from 2 benchmark runs (1 each) and now people are picking apart blender as if it hasnt existed for years.

Everyone just be patient and wait for the final fucking product to come out.

 

Very effective post.  I completely understand why you think everyone else is a child.

[Brian McKee]

8 minutes ago, done12many2 said:

 

What makes you think this?

 

The fact that turbo boost was enabled on the intel chip and the AMD chip was restricted to base clock only.

[done12many2]

Just now, Brian McKee said:

The fact that turbo boost was enabled on the intel chip and the AMD chip was restricted to base clock only.

 

So you consider the fact that Turbo was enabled on the 6900k to mean that it had a clock speed advantage over the Ryzen at 3.4 GHz?

[Brian McKee]

1 minute ago, done12many2 said:

 

So you consider the fact that Turbo was enabled on the 6900k to mean that it had a clock speed advantage over the Ryzen at 3.4 GHz?

Um... yes? Because 3.4 GHz is the BASE clock of Ryzen, meaning it goes higher. At least I hope it would.

[done12many2]

1 minute ago, Brian McKee said:

Um... yes? Because 3.4 GHz is the BASE clock of Ryzen, meaning it goes higher. At least I hope it would.

 

You do realize that with Intel Turbo, clock speed scales down as core utilization increases, right?  So in a test like Blender or Handbrake, a 6900k with Turbo enabled drops from 3.7 GHz to 3.4 GHz when all cores are loaded.  It was definitely a smart move for AMD to repeatedly mention (7 times during the event) the fact that the 6900k was running with Turbo "up to 3.7", despite the fact that it wasn't actually running at that during their test comparison.

 

Regardless, they've peaked my interest in the Ryzen chip.  The Extended Range Frequency stuff sounds pretty cool if it's not gimped. 

[Brian McKee]

2 minutes ago, done12many2 said:

 

You do realize that with Intel Turbo, clock speed scales down as core utilization increases, right?  So in a test like Blender or Handbrake, a 6900k with Turbo enabled drops from 3.7 GHz to 3.4 GHz when all cores are loaded.  It was definitely a smart move for AMD to repeatedly mention (7 times during the event) the fact that the 6900k was running with Turbo "up to 3.7", despite the fact that it wasn't actually running at that during their test comparison.

 

Regardless, they've peaked my interest in the Ryzen chip.  The Extended Range Frequency stuff sounds pretty cool if it's not gimped. 

Ok yeah then let's run the tests at the 6900k's base clock of 3.2 ghz. I don't see your point.

[done12many2]

1 minute ago, Brian McKee said:

Ok yeah then let's run the tests at the 6900k's base clock of 3.2 ghz. I don't see your point.

 

During the test, they were running the same clock speed.  I understand that you didn't realize that at first, but there's no need to now say that they should have run the Intel slower than the AMD during the testing. That just doesn't make sense.

[Brian McKee]

Just now, done12many2 said:

 

During the test, they were running the same clock speed.  I understand that you didn't realize that at first, but there's no need to now say that they should have run the Intel slower than the AMD during the testing. That just doesn't make sense.

It's testing base vs base, they're not the same architecture so same clock testing really doesn't mean anything either. Face it, AMD was crippled because it was base clock only vs complete out of box performance. 

[done12many2]

 

29 minutes ago, Brian McKee said:

The fact that turbo boost was enabled on the intel chip and the AMD chip was restricted to base clock only.

 

2 minutes ago, Brian McKee said:

It's testing base vs base, they're not the same architecture so same clock testing really doesn't mean anything either. 

 

I like how you can pretend that you didn't just say something and switch positions mid way through a conversation.  You distinctly said that the Intel chip had an advantage due to "turbo boost".  I pointed out that it didn't and now you're talking about something completely different.  I'll leave you to this conversation.

[Brian McKee]

5 minutes ago, done12many2 said:

 

 

 

I like how you can pretend that you didn't just say something and switch positions mid way through a conversation.  You distinctly said that the Intel chip had an advantage due to "turbo boost".  I pointed out that it didn't and now you're talking about something completely different.  I'll leave you to this conversation.

It IS an advantage, holy shit you're thick headed. If both the products were out of the box performance than by using your BRAIN wouldn't you assume that Ryzen would have the performance edge? If you're implying comparing clocks between entirely different architectures means literally anything at all then you don't know very much. 

 

What we know from the benchmarks right now is that Ryzen at the very least matches the 6900k with lower TDP and without boosting up AT ALL. 

[Dabombinable]

9 minutes ago, Brian McKee said:

It IS an advantage, holy shit you're thick headed. If both the products were out of the box performance than by using your BRAIN wouldn't you assume that Ryzen would have the performance edge? If you're implying comparing clocks between entirely different architectures means literally anything at all then you don't know very much. 

 

What we know from the benchmarks right now is that Ryzen at the very least matches the 6900k with lower TDP and without boosting up AT ALL. 

No it isn't, as the more cores that come under load, the lower the boost speed, all the way down to 3.4GHz, which is what the ryzen sample was running at. That is unless AMD manually tweaked the 6900K to run all cores at their full boost speed.

[porina]

3 minutes ago, Brian McKee said:

What we know from the benchmarks right now is that Ryzen at the very least matches the 6900k with lower TDP and without boosting up AT ALL. 

What we know from the benchmark right now is that Ryzen at 3.4 GHz matches a stock 6900k in that single Blender test, with lower rated TDP.

 

Absent better info, I'd speculate that the 6900k wouldn't need it's full TDP in that test as it doesn't seem that heavy a load compared to other stuff I run. Actually, that could be tested relatively easily. Proposal: Run that Blender benchmark and observe both reported CPU power usage, and actual total system power usage. For a point of comparison, something like Prime95 could be used as a heavy load case as from memory that is close to running at rated TDP. I'm willing to bet Blender will come out much lower in power consumption. I could try that on the weekend if it is desired on Broadwell (to better match 6900k) as my only system isn't easily accessible at the moment. I could do Haswell or Skylake test more easily before that.

 

There are videos of a press demo they gave separately where they displayed total system power on screen. Ryzen system power was slightly lower than the 6900k system, certainly not the difference in rated TDP.

[done12many2]

 

2 hours ago, Phate.exe said:

If anything it looks like they were trying to skew things in favor of the intel chip (at least in the benches shown).

 

2 hours ago, done12many2 said:

What makes you think this?

 

1 hour ago, Brian McKee said:

The fact that turbo boost was enabled on the intel chip and the AMD chip was restricted to base clock only.

 

1 hour ago, done12many2 said:

You do realize that with Intel Turbo, clock speed scales down as core utilization increases, right?  So in a test like Blender or Handbrake, a 6900k with Turbo enabled drops from 3.7 GHz to 3.4 GHz when all cores are loaded.  It was definitely a smart move for AMD to repeatedly mention (7 times during the event) the fact that the 6900k was running with Turbo "up to 3.7", despite the fact that it wasn't actually running at that during their test comparison.

 

Regardless, they've peaked my interest in the Ryzen chip.  The Extended Range Frequency stuff sounds pretty cool if it's not gimped. 

 

1 hour ago, Brian McKee said:

Ok yeah then let's run the tests at the 6900k's base clock of 3.2 ghz. I don't see your point.

 

1 hour ago, done12many2 said:

 

During the test, they were running the same clock speed.  I understand that you didn't realize that at first, but there's no need to now say that they should have run the Intel slower than the AMD during the testing. That just doesn't make sense.

 

1 hour ago, Brian McKee said:

It's testing base vs base, they're not the same architecture so same clock testing really doesn't mean anything either. Face it, AMD was crippled because it was base clock only vs complete out of box performance. 

 

1 hour ago, done12many2 said:

I like how you can pretend that you didn't just say something and switch positions mid way through a conversation.  You distinctly said that the Intel chip had an advantage due to "turbo boost".  I pointed out that it didn't and now you're talking about something completely different.  I'll leave you to this conversation.

 

1 hour ago, Brian McKee said:

It IS an advantage, holy shit you're thick headed. If both the products were out of the box performance than by using your BRAIN wouldn't you assume that Ryzen would have the performance edge? If you're implying comparing clocks between entirely different architectures means literally anything at all then you don't know very much. 

 

What we know from the benchmarks right now is that Ryzen at the very least matches the 6900k with lower TDP and without boosting up AT ALL. 

 

 

I understand how you are confused and at this point you're all over the place with shifting focus on the topic.  You said that the 6900k had a clock speed advantage due to turbo boost.  I pointed out that they were actually running the test at the same clock speed due to the way Intel Turbo works.  You then switched to a conversation about base clocks, threw out some insults, and then further switched to talking about TDP and whatever comes next as you push away from your initial statement.  

 

I've been wrong plenty of times.  There's just a very big difference in the way you and I handle handle it when it happens.  No biggie.

 

[S w a t s o n]

This is why we can't have nice things

[done12many2]

2 minutes ago, Swatson said:

This is why we can't have nice things

 

The smell of that beard burning! 

[XenosTech]

9 minutes ago, Swatson said:

This is why we can't have nice things

Are you sure you aren't my subconscious ? You have my initials in all