Explain this shit to me Sup Forums

Explain this shit to me Sup Forums

>GTA V, 720p test.
>Ryzen 70+ fps, threads 61-78%
>7700k 85+ fps, threads 95%+

Why isn't the Ryzen using 95%+ on each core like the 7700k?

Source: youtube.com/watch?v=nsDjx-tW_WQ

Other urls found in this thread:

youtube.com/watch?v=Vb3IMTJjzfo
twitter.com/SFWRedditImages

No game in existence today is written with Ryzen's architecture in mind.

Windows Scheduler hasn't been updated to understand Ryzen's SMT yet. So it makes less efficient use of the cpu. And the motherboards are tsill buggy anyway. And it's possible that gtav doesn't scale past 8 cores or something.

Because Ryzen is a piece of SHIT

>launch performance
>versus more mature architecture

Why is the sky blue?

A new product that hasn't been optimized like Jewtel is running with less performance while being used consumed less?
Makes you think.

In reality Ryzen is better for productivity and will be more future-proof than the current i7 line-up, plus it's cheaper for performance and multi-core games are rising due to Xbone / Sup ForumsS4.

youtube.com/watch?v=Vb3IMTJjzfo

Also this test is stupid, it's not testing nearly the same situation. Look at the gpu usage, completely different. Look at the memory usage, 5852MiB on Ryzen vs the Intel scene using 1712MiB. It's a useless comparison that says nothing.

Why isn't the 7700k using 100% constantly?

Windows is switching threads between cores basically at random on ryzen so there's performance losses

Windows also isn't scheduling properly on ryzen

Motherboards need bios updates to optimize for ryzen

Basically all software made within the past 5-10 years is optimized expressly for intel cpu's and architectures instead of AMD

In spite of all this, ryzen is only 10-15% behind on single-thread performance

That's because the 7700k is actually bottlenecking in that benchmark, while the 1700x is doing better thanks to its higher core count. Notice the CPU usage. The 1700x is also with SMT enabled which has already been show to negatively impact performance on Ryzen. But yeah, I'm sure all those other benchmarks without any sort of real time footage and less information about the setups are far more trustworthy.

Because nobody has ever had to program for AMD's SMT before and it's done slightly differently than Intel's hyperthreading. I dunno what Sup Forums was expecting but an entirely new architecture with a bunch of new features will require optimizing for, and that hasn't happened. AMD's fail wasn't the design, it was failing to work with game devs for years before to get SMT properly implemented.

In spite of the cuckery by their Lord and Saviour AyyMD, the Sup Forums edgelords still believe in Ryzen.

look like amd win, normal Sup Forums tard would have shit ton stuff on background

>Presented with a factual account of why Ryzen's launch performance was below it's capabilities
>call names
Intelbreds should be euthanized.

so far 1700 is basically only viable 8c/16t CPU on the market, there is no competition.

GTA V came before Ryzen, that's why.

The 1800x is great if you're looking for better binning 8c/16t and don't mind paying the premium for it.

Literally paying more for useless features that you will disable if you are actually going to OC it.

If you're buying the 1800x you should be using it in a workstation where those features aren't useless.

You sound like some kind of idiot who's fooled himself into thinking the only thing people do with CPU's is play games on them.

XFT only can OC 1 core
1800x will downclock at full load

Not OCing it, is fucking stupid, keeping your 8 cores at 3.8Ghz is already a massive improvement in workstation over default.

>they can fix the shit hardware! its a software issue!!!!!
zimbabwe

other than couple of games, 1700 is 5fps behind 7700k, fix what exactly? your fucking brain?

Please fuck off back to with your memes, or just use any of the other 9999 Ryzen threads.

This

I'm sure you Sup Forums kiddies are too young to remember, but this is the same story as it was with AMD Athlon when it came out first time. Comments were like that: good performance, but since it was derived from server chip (Opteron) is good for scientific calculations but not for games. Intel is better overclocker and gaming chip. Once Athlon matured it showed its power and Intel didn’t recover till Core.

Most people on this board don't remember the same shit happening with Nehalem aka the first multicore arch with HT.

>higher fps
>its bottlenecking
XD

>nearly 100% CPU usage
>not bottlenecking

>doesnt know what cpu scheduling/timesharing is
XD

>zen is shit in real world scenarios
>amd blames everyone but themselves
LEL

>720p
what's this, the switch?

AMD has no drivers.

>what's this, the switch?
No, it's Intel

People should benchmark on all "played" resolution. game act differently and dont scale linearly like the review make it seem

What is happening there is that at 4k they are gpu bottlenecked, that's why the performance is so similar. At 1080p the systems are not gpu bottlenecked, so the real performance discrepancy in low threaded workloads (ie games) shows up.

>those Doom benches
But wasn't the 1800x shown as having the same Single Core performance as the 6950k? Why is it behind a whole 30FPS/18%? And why is their such a large disparity in performance between the 1700x and 1800x?

But then GTA5 rolls around. A near 10% frame deficit to the 6950k, but has the same FPS as the 1700x.

The fuck is with these benches?

The game probably can't scale to 16 threads efficiently, so it benefits more from the faster cores. Also having more cores will have a tendency to lower individual usage. Imagine the game has a busy thread using 75% of 1 core and another lighter thread using 15%. If these threads get scheduled to different cores you'll see 75% and 15% usage, but if they're on the same core you'd see 90%. A similar scenario could be happening with the 8C/16T Ryzen too, so it's doing the same work but spread over multiple cores, as such individual usage is lower.

>Windows is switching threads between cores basically at random on ryzen so there's performance losses
>
>Windows also isn't scheduling properly on ryzen
Has this actually been confirmed or is it still speculation based on the fact that disabling SMT increases performance? At the very least Windows knows that half the cores are logical and not physical (Task Manager shows 8 cores, 16 logical CPUs in the same fashion it does with Intel HT CPUs).

>Windows knows that half the cores are logical and not physical
Supposedly it treats them the same though

Some game do use more cpu power if you increases the res. This not pentium era cpu testing anymore

>Supposedly it treats them the same though
I know the story, but is it based on any fact or is it purely speculation? Because I haven't read anything that provided some evidence, just speculation. Not that it's necessarily implausible, but still.

What's the invalid RE7 benchmark supposed to be about?

>Why isn't the Ryzen using 95%+ on each core like the 7700k?
because
>muh cores
is a meme

you literally cannot fully utilize more cores for a certain task beyond a certain point. the best you can hope for is to add more stuff like more characters and more physics objects and things like that which can be computed in parallel.

something something something muh cores something something the future of gayming something something

Maybe this is related?

>What's the invalid RE7 benchmark supposed to be about?
To show that performance isn't static and can change at the drop of a hat if there are bugs and kinks.

But the cores thing is probably based of something. Google it and look for sources.

IIRC there was an issue with Guru3D's methodology in that case (and they fixed it), so it actually doesn't make your point at all. I don't really remember, something about a setting I think it was.

I guess, but a Bulldozer issue doesn't say much about Ryzen. That's still no evidence that Windows has broken scheduling.

>IIRC
Stopped right there
Sources or bust

Read the fucking article, it's corrected and updated now. I'm not going to write a summary for you.

THEN POST THE ARTICLE
STOP ANSWERING WITH
>"yeah it might have been this but im not going to show you because i just want you to take me at face value."

And no games ever will.

1800X will almost always clock higher than the 1700.

The only difference between the 1700 and 1800X is that the 1800X's have been confirmed to hit higher clock speeds stable.

If you're too dumb to find it despite the source already being mentioned literally a few posts above there's actually no point in attempting to discuss anything with you. You posted the image, but you don't even know where it comes from? How retarded are you?

>despite the source already being mentioned literally a few posts above
Except I wasn't asking for the source
I know its from guru3d, it's right in the image
I wanted you to post the URL because you're making a claim. If you can't, then you're clearly wrong and just attempting to obfuscate that fact while wasting my time

AMD had to include a ptach in the 4.10 Linux kernel to help Linux's scheduler understand how to use their hyperthreading. Windows will almost definitely require something similar.

>So, without further decoding cpu_core_id is unique for each thread rather than the same for threads on the same core. This didn't affect systems with SMT disabled.
>This didn't affect systems with SMT disabled.

>I guess, but a Bulldozer issue doesn't say much about Ryzen. That's still no evidence that Windows has broken scheduling.
It says Zen in bold, and the patch was submitted less than a month ago. Bulldozer disn't even have real multithreading anyways, and yes, considering Windows shows a decrease in performance with SMT enabled shows there's a problem with SMT.

>I wasn't asking for the source
>I wanted you to post the URL
What URL do you want if not the article?

Yeah I know, I'm not saying it's not plausible that Windows will require an update because it most certainly is, I was just curious if there was some actual evidence for this beyond an educated guess.

>What URL do you want if not the article?
That exact one, which isn't in this thread or included in any image in this thread.

>It says Zen in bold
It does, right below that it says Bulldozer too. A problem with SMT isn't necessarily a problem with scheduling or a problem just with scheduling.

I mean even Intel with HT sometimes loses performance when it is enabled and there surely aren't scheduling issues with HT, just because AMD's SMT brings with it a performance drop in certain scenarios doesn't mean the scheduler alone is to blame.

You said you don't want the source, but you want the URL to the article, which is literally the source. What do you want exactly?

>You said you don't want the source
Guess I'll recap then
>faggot says "IIRC there was an issue with Guru3D's methodology in that case (and they fixed it), so it actually doesn't make your point at all. I don't really remember, something about a setting I think it was."
>understandably, i ask for a source
>"faggot, its Guru3d, you should know that"
>Clearly this guy has never made a Sources cited page
>I ask for URL to article since explaining why he's retarded would be a wasted effort
>"but I already gave you the source user, I told you it was Guru3d"

And here we are, you pretending to be retarded to hide the fact that you were lying, and myself without an article.

Care to elaborate on what you mean by this?
CPU's are not GPU's, you can't use the same argument "game developers design games with Nvidia drivers in mind"

If you go and read the thread, a fix they made several months earlier to not treat Bulldozer threads like hyperthreads broke SMT (which is not a feature of Bulldozer, only Zen) for Zen, which they submitted a patch to fix.

So the evidence is that, besides SMT decreasing performance when it should be increasing it, both AMD and Intel have code in the Linux kernel to help the kernel/scheduler understand how to use their cores/threads. It's pretty ridiculous to assume Windows doesn't need that kind of code.

When Intel first launched hyperthreading it had a lot of issues as well.

What he's probably talking about is the whole meme about Intel's current lineup of CPUs still being based on a CPU Architecture that's over 2 decades old. Any kinks with the CPUs themselves should already be worked out.

So basically
Old and reliable
compared to
New and unstable

See

The thing is, Ryzen's instruction set is still x86-64
What he is saying literally makes no sense. Unless there is more to it than this, in which case I'm interested tell me more.

Let me get this straight. The problem is that you know the benchmark is from Guru3D, but despite being armed with that knowledge you cannot find the article itself? Fucking hell I was right from the start, you're too dumb to be worth talking to. Here's the corrected benchmark though, you can apply yourself to find the actual article by working from the extra clue, should be good practice for you.

>It's pretty ridiculous to assume Windows doesn't need that kind of code
The assumption wouldn't be that Windows doesn't need it (it would need it, obviously), the question is if it already has it or not, all of this taken in the context that the Task Manager displays the correct number of physical cores and logical processors already.

>The problem is that you know the benchmark is from Guru3D, but despite being armed with that knowledge you cannot find the article itself?
No, If i really wanted to find it, it's only >Image search>google away

Hell, I could have done that a hundred times over with that time you've spent on deflection.

But fuck you, I'm not the one attempting to make the argument. That's you, and you have refused to do so.

If you don't want to prove your statements, then don't make them to begin with.

my theory has to do with xeons

you can get same fps as 4.5ghz i57600 now on 8 years old xeon with any videocard
drivers are multithreaded for a long time, they can't be efficient otherwise

but game logic gets slightly stuck on new arch, same way xeons were crap for gaming 10 eyars back,
meaning ryzen has more years in it than 7700K

so make your choice, also it smooths out driver overhead to non existent even now, look at frametime graphs i7 ahs jumps to 60ms at times, ryzen keeps it under 16ms

If it couldn't use more cores, there would be unused cores.

they're only ~70% utilized in OP's pic. the actual workload is probably more asymmetrical and the scheduler is moving threads back and forth to different logical cores to balance the load

i like how the picture name is 800*600 and the picture size is 680 * 480.
made me kek

Good point, didn't think of that