Ryzen: Strictly technical

[The Stilt]

My issue isn't with Zen vs other CPUs, it's what the graphs show, and that graph shows a 52% margin yet has a smaller apparent margin that the other graph I showed.

Yeah, this is because of the autorange function in Excel.
I can redo the charts so that both the y-axises start from zero, no problem.
But deceiptful & misleading... You got some set of balls, friend.

I could somewhat understand the statement you made, if the absolute results and percentages wouldn't be displayed for each chart. But since that's not the case for any of them

 

[tamz_msc]

But it is deceitful and misleading. It may not be the intention, but that doesn't make it any less so. And the argument "one ought to take the time" can just as easily be applied to any scam - if you don't take the time, you deserve to be scammed.

I'm not saying that Stilt is trying to slant his article in one direction or another, and I'm not saying that he's trying to deceive and mislead the readers, but that does not make the graphs any less deceitful and misleading, and given the amount of effort Stilt clearly put into this, why go out of your way to undermine your credibility by using graphs that exaggerate data (which is deceitful and misleading)?

Again, you are making it out to be worse than it really is.

I'm in academia - this problem crops up all the time. At the end of the day you discuss the data - what importance you give on the on the presentation depends on the context. This thread is free-flowing and informal - a far cry from NVIDIA's presentations at media briefings. If there is scope for confusion you point it out and things usually get revised and the presentation is made clearer - it is a process of continuous refinement.

Also, software does weird things - if you ever had to draw graphs for closely spaced data on a graph-paper, then you'd know most of the time is spent in choosing the scale than drawing the graph itself.

 

[itsmydamnation]

Yeah, this is because of the autorange function in Excel.
I can redo the charts so that both the y-axises start from zero, no problem.
But deceiptful & misleading... You got some set of balls, friend.

I could somewhat understand the statement you made, if the absolute results and percentages wouldn't be displayed for each chart. But since that's not the case for any of them

if your going to redo your charts, can you also add a separate aggregate graph without the tests that have big advantage from 256bit ops

 

[Martin Schou]

Yeah, this is because of the autorange function in Excel.
I can redo the charts so that both the y-axises start from zero, no problem.
But deceiptful & misleading... You got some set of balls, friend.

I could somewhat understand the statement you made, if the absolute results and percentages wouldn't be displayed for each chart. But since that's not the case for any of them

The graph is misleading. It's not that YOU, Stilt, are actively trying to deceive or mislead anyone, but the graphs in question still do it.

It's a bit ... well, imagine if you benchmarked a game including grabbing frame times and you ended up with only four values: 1 ms (one instance), 2 ms (50%), 15 ms (50%) and 100 ms (1 instance). If you simply plotted the min, max and average (10 fps, 1000 fps and 117 fps) and called it a day, that'd be misleading and deceitful, even if that wasn't your intention. It'd be misleading and deceitful because it hides a very important bit of information, namely that the game is massively unplayable with huge amounts of stutter (basically every other frame) unless you enable vsync, even though the frame rates look nice (everything is above 60 fps for typical monitor refresh rate). A graph with just min, max and average would be misleading and deceitful in an even worse manner, because you can't see the data that has been hidden. Again, it doesn't mean that the person presenting the data wants to be deceitful.

First impressions matter. If you show someone a cropped graph they will walk away with a different impression than if you show them a non-cropped graph, even if they spend time looking at all the data available.

Now, I realise that the words I chose are somewhat combative or aggressive. It is a pet peeve when knowledgeable people use (what I at least perceive as) bad presentation tools, and this particular article was guilty of having both an apparently very knowledgeable and respected author (certainly beyond my knowledge level to try to refute anything presented) combined with cropped graphs. It's a bit like trying to watch a great movie while someone is scraping a fork across a plate - it's extremely distracting and takes away from the pleasure.

 

[SpaceBeer]

The graph is misleading ONLY if you don't read the values. Maybe it doesn't look nice, but it is accurate. It shows correct test results.

Great work @The Stilt , this is the best Ryzen review I've seen so far

 

[dnavas]

Boot up, mouse only, and if you don't have a password it goes direct to the desktop. Now with the mouse you can navigate the driver disk menu and get the drivers installed. After installing and a reboot or two you are set. You can then put the keyboard and mouse both in how you want and they will work.

Thanks very much for the instructions -- alas, this failed to (completely) work for me. I did get the mouse working through the USB / PS/2 port adapter, and installed drivers (so I now have a network), but no USB3 ports (and the front-header USB2 don't work -- presumably they are through a USB3 hub for the gigabyte gaming5 board). I can only get the mouse to work about once every six or seven reboots, which is ... frustrating. I guess I need to find the USB3 drivers and convince Windows to use them :|

I also haven't had much luck getting my RAM recognized. I go ahead and select the 2933 CL14 settings, and much flashing and long reboot later, I'm back to slow-ram mode I'd look for BIOS updates, but I kind of want a working OS first

Update: Did a brand new install of Windows 7 -- that works. Must be old drivers on my old installation. Not looking forward to a full reinstall. Bleah. Might bite the bullet and install Win10 given I've got a pair of 4k monitors, and Win10 has marginally better scaling than Win7

 

[imported_jjj]

Yeah, this is because of the autorange function in Excel.
I can redo the charts so that both the y-axises start from zero, no problem.
But deceiptful & misleading... You got some set of balls, friend.

I could somewhat understand the statement you made, if the absolute results and percentages wouldn't be displayed for each chart. But since that's not the case for any of them

If you redo them, just for precision, you could also add "higher is better" and "lower is better" since folks don't always spot that difference.

 

[PotatoWithEarsOnSide]

If its Excel that has scaled it like that, then Excel has a design flaw. At school, we were taught that a graph that doesn't start at 0 on an axis should have a squiggly (is that a word?) bottom to the axis to emphasize that it doesn't start at 0.

 

[Atari2600]

If its Excel that has scaled it like that, then Excel has a design flaw.

No shit. Excel has always had this flaw in it.

If you could set in preferences for all graphs to default originate at 0,0 then it'd be OK, but as far as I know, you can't*. So, yes, its a pretty big oversight.


*If anyone can point me in a direction to change the excel default setting, I will profusely apologise to M$.

 

[dacostafilipe]

So, "Strictly technical" relates to Excel and not Ryzen ... my bad ... /s

 

[lopri]

WRT arguments surrounding the graphs: Both parties are aware of misunderstanding/miscommunication as well as what they initially overlooked. And thankfully it appears the both parties are mature and intelligent enough to take the steps forward, not backward, to find a solution to a shared inquiry. Me likey.

 

[Atari2600]

So, "Strictly technical" relates to Excel and not Ryzen ... my bad ... /s

Having been in a similar position to The Stilt, when you are already rushing along doing something, its a right pain the ____ to have to go into excel and manually zero the axes origin for each and every graph.

If anyone from M$ is reading this and considering opening a fix on it, it'd also be useful if you could group up graphs so that you can put a common axes on them - i.e. change the axis on one graph and it automatically updates the rest - useful for scientific comparisions where you want to keep consistency of presentation.

[and no, manually copying and pasting the graph format is not an effective alternative.]

 

[JimmiG]

It's contradictory because High performance is recommended by AMD too. But at the same time screws max turbo. Core Parking is hurting AMD performance but at the same time COre Parking enables max turbo, so it's a lose-lose situation. Hopefully AMD gets it fixed with MS soon.

True, without Core Parking, you're limited to 3.7 GHz max turbo. With it, you get up to 4.1 GHz, nearly 11% higher. On the other hand, Core Parking itself can cause a 10% performance hit. So with Ryzen you only get to choose which way you want to kill performance

We desperately need an update for Windows 10 that better handles parked cores, CCX's and Ryzen's SMT. There's probably at least a 10% gain to be had there, especially in games.

However, even with Core Parking enabled, XFR is very finicky. As soon as a third core gets even 1% usage, it immediately dials back the speed to 3.7 GHz. So you're only ever going to see short burst of 4+ GHz. It's not the "magical auto overclocking limited only by your cooling" that they advertised. AMD clearly thought performance with many threads was more important than XFR since they recommended High Performance mode to reviewers.

 

[Malogeek]

High starting point graphs are usually only what you see from marketing departments designed to make their competitor look bad

 

[Harney]

So what is the roughest estimate in terms of gains for the ryzen once the scheduler is fixed in win 10....

Is the scheduler working correctly in win 7 if so what's the gain there ? that should give an idea at least
or
is it a case that it will be better optimized on 10

Come M$ pull your finger out of your @@@@

 

[inolvidable]

How feasible do you see Intel pushing AVX2 patches into productivity software that can take advantage of the instruction set in the next few months (rendering, photo editing, encoding)?
If doable it could hurt the Ryzen 8c/16t CPUs where they truly shine.

Thanks in advance. I am no expert and sometimes ignorance can ask silly questions. If that is the case, also my regrets in advance.

 

[thedoginthewok]

How feasible do you see Intel pushing AVX2 patches into productivity software that can take advantage of the instruction set in the next few months (rendering, photo editing, encoding)?

I'm no expert either, but I can't see anything like this happening anytime soon. If anything I think this is something that takes longer than "the next few months".

 

[The Stilt]

if your going to redo your charts, can you also add a separate aggregate graph without the tests that have big advantage from 256bit ops

The problem with this is that where shall we draw the line?
I mean sure, Embree, Linpack and X265 are obvious but what about Blender for example?
It uses both 128-bit & 256-bit code and receives significant gains from 256-bit ops?

 

[CatMerc]

The problem with this is that where shall we draw the line?
I mean sure, Embree, Linpack and X265 are obvious but what about Blender for example?
It uses both 128-bit & 256-bit code and receives significant gains from 256-bit ops?

I say remove blender too. Anything that uses 256 bit code.

It will not be a realistic representation of Ryzen's performance, I'm just curious as to how it compares without AVX256 in the mix.

 

[Atari2600]

How feasible do you see Intel pushing AVX2 patches into productivity software that can take advantage of the instruction set in the next few months (rendering, photo editing, encoding)?

Not very.

Software developers will put the stability of an existing package above humouring Intel (who have disabled various instructions as they tier their products - meaning developers would be addressing a restricted market).

But, future releases (as opposed to updates), may benefit from AVX2.

 

[R0H1T]

Check my other thread ~ https://forums.anandtech.com/threads/rightmark-processor-power-management-panel.2500879

 

[CrazyElf]

The base-clock (BCLK)

Overclocking the base clock (BCLK) on AM4 platform is possible, however generally not recommended. This is due to its frequency relations with other interfaces, such as the PCIe. Unlike with Intel's more recent CPUs, there is no asynchronous mode (straps / gears) available, which would allow stepping down the PCIe frequency at certain intervals. The PCIe frequency relation is fixed and therefore it increases at the same rate with the BCLK. Gen. 3 operation can generally be sustained up to ~107MHz frequency and higher speeds will usually require forcing the links to either Gen. 2 or to Gen. 1 modes.

Unstable PCIe can cause various issues, such as system crashes, data corruption (M.2 SSDs), graphical artifacts and various kinds of other undefined behavior.

Can you verify this?

This OC exists with on an Closed Loop Cooler:
http://valid.x86.fr/qmfrkd

Note the base clock and multiplier.

 

[The Stilt]

Can you verify this?

This OC exists with on an Closed Loop Cooler:
http://valid.x86.fr/qmfrkd

Note the base clock and multiplier.

Verify in what respect?
Sure, you can hit those BCLKs if you set the PCIe to run in Gen. 1 mode.

 

[Atari2600]

It will not be a realistic representation of Ryzen's performance

Why not? If the existing software releases include the code?

I'm just curious as to how it compares without AVX256 in the mix.

Not better for you then to downselect and compare those that don't include AVX256 yourself?

 

[CatMerc]

Why not? If the existing software releases include the code?

I'm saying that removing AVX256 is not realistic, because AVX256 exists and is used in said software.