AMD Ryzen 1700 Emulation Testing...

[BONKERS]

*EDIT* Its MAME 0.123b 64bit . I know this particular game used multithreading for 1 core to emulate the R5000 and the other to emulate the 3DFX chip. I remember Aaron Giles (the programmer for this driver) blogging about it and it did indeed help performance on this particular driver when I tried it back in the day. 64 Bit OS also helped quite a bit over the 32bit version.

I never tried to go higher than 3.6, I didnt even have to touch the voltage to reach it. I could probably get 4.0 with some voltage tweaks, but Im all about longevity, I dont want to stress the chip more than necessary. Ill post my 2500k results later for reference.

I did try disabling SMT, and at first it looked like it gave an extra 5% or so, but more testing showed it was probably a wash. I do have the ability to shut off cores in the BIOS, I believe I can also isolate cores to a single CCX (2+0 vs 1+1). I'll also have to look into the multithreading issue with newer MAME versions to see how it compares. A lot of variables really, but I love testing this kind of stuff.

I plan on adding the Dolphin Benchmark as well as in game if possible, and also things like CPU-Z, CrystalMark and perhaps some 3D Mark tests, though Im only using an R9 285 (which I bought used and Im not so sure its not doing some wonky stuff, need to test further).

I bought the chip mainly because my mobo for my 2500 was doing some really wonky stuff and you cant really get those new anymore, so I figured an octocore would be nice! I dont really game much other than some emulation here and there , but I do a good bit of photo and video editing/encoding, so this looked like a killer chip for my purposes. Other than having hell (and giving up) on running it on Windows 7 , ( I got it working, just had driver issues), it seems pretty bad ass so far. More in the coming days....

I wouldn't worry about longevity.

I've been running a first generation Core i7 950, overclocked to a static 4Ghz (Speed step doesn't work properly on the first series. So, no dynamic clocks. 4Ghz all the time) with a voltage of 1.39/1.4 or so. For 6 years straight, haven't had a single CPU issue.

And this was a CPU that stock clocked at 3Ghz.

Mark my words, Scorpio is going to be Ryzen, and the decision to wait for Zen instead of just bumping up the clocks a bit is going to be why Scorpio so enormously faster than the PS4 Pro.

I hadn't thought about that, but I think you're right. The PS4 Pro is severely limited by its CPU at this point.

While the CPU is the big bottleneck ),I think it's far more limited by the inconsistency of support by developers and even Sony.
Personally, the PS4Pro 400$ upgrade has been a joke so far.

1080p60 is still an issue, and official patches are often shoddy, or regressive, or offer no real support at all.
Then there's the rest of the library that only gets boost mode. Which does nothing unless the game has performance problems to begin with. And in that case it's not super impressive that this 400$ purchase can only make your 30FPS game drop less frames.

If a 400$ GPU upgrade on PC performed this bad. You can only imagine the ire.

Ran the Dolphin benchmark. Results below.

Not too impressive. My i7 950 @4Ghz does it in shorter time...despite being 8 years older.

[Udderdude]

Curiosity, why are you already locked into Vega? nVidia has been pretty handily ahead of AMD on the GPU side for.. seems like forever now, and until we see actual Vega results there isn't much to indicate this is going to change. Recent case in point, the Polaris GPUs are good, but probably not as good as hoped, and certainly not a better architecture than the competing nVidia parts.

I'm irritated by how Nvidia drops support for their previous generation hardware like a hot potato as soon as the newest generation comes out. They just assume everyone is going to immediately run to upgrade to the next one. They're also more expensive. I'm willing to go for max price/performance.

[bigbadboaz]

Cool. Here's hoping Vega overdelivers vs. Polaris -

[BONKERS]

Curiosity, why are you already locked into Vega? nVidia has been pretty handily ahead of AMD on the GPU side for.. seems like forever now, and until we see actual Vega results there isn't much to indicate this is going to change. Recent case in point, the Polaris GPUs are good, but probably not as good as hoped, and certainly not a better architecture than the competing nVidia parts.

I'm irritated by how Nvidia drops support for their previous generation hardware like a hot potato as soon as the newest generation comes out. They just assume everyone is going to immediately run to upgrade to the next one. They're also more expensive. I'm willing to go for max price/performance.

Yet they haven't done this with the 10 series. The 970 is still being used as a recommended card interchangable with the 1060, even though they aren't similar in performance.
Clearly they are still supporting Maxwell in addition to Pascal. Just as they did with 700 series Kepler and 600sK. It's only when Maxwell was introduced and Kepler hardware performance started showing it's issues that people threw a fit.

You can still game just fine on older Nvidia hardware. Just don't expect to run things at max settings at 60FPS.

This is how things have been for over a decade, cards start showing their age a lot quicker on PC compared to consoles. Which have better optimization, while PC continues to run at higher resolutions and push visuals/CPU logic further and further beyond what the console is capable of with each release.
https://www.guru3d.com/articles_pages/m ... ew,12.html
A GTX 780 Ti can still perform admirably even in the latest games if you play at 1080p.
And it also shows Maxwell cards still performing just fine too. A 980 Ti, still only performs slightly worse than a 1070.


I doubt AMD is still updating drivers and trying to optimize stuff for 2xx or 7xxx cards either. Meanwhile, their DX11 driver still is mediocre and that's the primary reason why they perform so much better in DX12.

I doubt Vega will top the 1080Ti. It might get close, for a lower price. But it will still be pricey. The R9 Fury launched at 550$ , and the Fury x launched at 650$

Neither team is perfect or any kind of ethical saint.

[bobrocks95]

RE the PS4 Pro- it's not a $400 upgrade, it's a $100 upgrade- the slim now costs $300 MSRP.

If anyone rushed out to buy a PS4 Pro to replace their perfectly fine PS4, I can understand the frustration, but I also don't see why you'd be in a hurry to do that.

[Udderdude]

It's only when Maxwell was introduced and Kepler hardware performance started showing it's issues that people threw a fit.

I still own two Kepler GPUs. Can't stop me from being bitter. :p

[Josh128]

Would also be interested to see what results you got with say Mushihimesama Futari in Mame with an unlocked framerate on both of your processors as the CV1k driver is more taxing than most stuff on there.

Please don't think I'm shitting on the Ryzen range, no doubt multicore is the future for gaming and agree that they seem like a great cpu for the examples you gave like encoding etc. Will be very interesting to see how Intel responds now that they actually have some competition.......

Here's Mushihemisama Futari, Tekken 3, Solar Assault, Marvel vs Capcom. Had to run Mushi on a v0184 due to my v0123 not recognizing. So many variables, as different mame builds can have big differences in speed....

[Josh128]

Some more 3D Mark Results... Ryzen 1700 Stock, R9 285 GPU, 8G 2667MHz RAM

[Josh128]

This is really cool, had the idea to run 3 different instances of MAME and see how the speed was affected... apparently it was not affected much at all, check this out. I would have run more instances but I couldn't get the % emulation to show up in more than 3 at a time. Interesting results, wonder how lesser core chips cope with this type of stuff....

[tomwhite2004]

Yep, many variables will affect it but I ran a few quick and dirty comparisons with an i3-6320 and mame 0.171 (0.123 is ancient dude, you are missing out on lots of good stuff!), sorry for not providing any screenshots but I really cant be bothered to crop and upload everything right now.

Percentages taken in game, not on attract loops and ran for a good while to get a ball park average. I went for a conservative lower end of the total figure I saw but it went up by in places by an extra 20-30 points in most cases.

Mushihimesha Futai 1.0- 400%
Solar Assault- 200%
Tekken 3- 250%
Marvel Vs Capcom- 1800% (was surprised that this didn't see as much of an gain as the other games)

Also tried running those three titles in separate windows and saw around 10-20 off the total percentage decrease in all games, but still possible with a dual core.

I think the heavier Dolphin cpu bound stuff like The Last Story or a lot of the more demanding PCSX2 games like Shadow of the Colossus would be a better workout, but getting everything in a like for like situation would be difficult.

[Josh128]

All the percentages I posted were during in game moments during attract loops, not during title screens, etc-- these scenes pull the same load as in game, sometimes even more due to additional overlays on top of the in game graphics.

Will try to get some Dolphin, PCSX2, NullDC/Chankast, and maybe some Supermodel and Model 2 Stuff as I have time.

[Josh128]

Ok, this is taken a bit to the extreme, but below are 3 different screen grabs. In this I am running 7 simultaneous instances of MAME, and I tried to screen grab when all 7 were showing in game 3D models and/or sprites in Futari. I also have the Windows utilization monitor up so you can see the CPU utilization. All 7 of these games were emulating at way over 100%, each was updating over 60 fps despite at least 20 of them (Mace and Solar Assault) being 30 fps games. The Ryzen was showing about 75% utilization, and remember, this is at stock 3GHz clock. I'm fairly certain dual cores and probably most quad cores will not achieve this, but I would be interested to see how they stack up.

I stopped at 7 instances of MAME because the screen was simply too cluttered. The point here is not to shit on dual and quad cores, but to show unique things that these 8 cores can do and also to point out that they still run MAME quite well despite the apparently widespread notion that they are not good choices for emulation. None of these titles dropped to under 100% at any time that I saw, despite 7 simultaneous sessions running. I'll be moving on from MAME to other more recent hardware emus now unless there are any other specific requests.

Games used were:
Mace: The Dark Age (150MHz CPU + Voodoo Graphics)
2 Instances of Mushi Futari (100+MHz CPU, SH-3)
Tekken 3 (Dual CPU, 100MHz and 15MHz)
Soul Calibur (Dual CPU, 100MHz and 17MHz)
Solar Assault (Tri CPU, 32MHz Power PC, 36MHz, 16MHz)
Killer Instinct 2 (100MHz CPU)

[tomwhite2004]

to point out that they still run MAME quite well despite the apparently widespread notion that they are not good choices for emulation.

Its not that they are not a good choice for emulation but the point is that for half the price you can get better results on non multi core supported emulators. And as has been previously discussed those emulators are in their infancy. And I'm not going to set up seven instances of Mame as I'm not really interested in going to such a extreme with a non real world scenario, but sure those extra cores help.

This discrepancy used to be ok as historically AMD have been much cheaper than Intel but the cpu you are testing is £300 which puts it into another bracket of performance altogether and just wouldn't be good value on an emulation focused build right now.

In fact I would say that we are at the stage with desktop processors that you can buy any new cpu from either AMD or Intel and have no problems running pretty much anything full speed except for a few notorious examples like Shadow of the colossus which everything struggles with. The real challenge now is for mobile and portable cpu's to catch up...

[Josh128]

Some Dolphin Benchmarks. Run with 1,2,3, and 4 simultaneous sessions. Really interesting occurence is, I got the highest framerate on a single session when I had 3 other sessions also running. Very strange, but when you have 4 sessions open, whichever one you hover the mouse over gains fps, robbing from the other 3. If you click on some other program, such as the utilization monitor, all 4 sessions even out. The strange thing about this is, I can consistently get the highest fps in a single session, by having 3 others open!

If I open just a single instance and go to the exact same scene (practice, no opponents, first level, firestingray), I dont quite reach the same fps as when I highlight one with 4 others running. Very strange, and speaks to how Windows, multicore processors, and emulators function when attempting to delegate a large workload. Perhaps this is causing 1 or 2 cores to turbo to 3.7GHz for the active window, but I cant say for sure because Win 10 doesn't really show you that. More later...

[Josh128]

Ryzen 1700 [email protected] DX12 vs DX11 Dolphin Comparison - DX12 gains 100fps over DX11! That's a gigantic difference, I presume the perf increase is similar in other CPUs, can anyone else verify? Also included Dolphin Benchmark @3.9GHz. Looks like clock for clock at least it beats Bonkers 8 year old 950 right. In all seriousness, optimizations for the Ryzen uarch for Dolphin would probably give some gain, just like what happened with the recent AOTS patch. Lets face it, most software in the last 5 years has been built on and optimized for Intels newest CPUs. There are obvious differences with the Ryzen architecture, programs have not been optimized for it yet, but the potential is there.

Also, playing around with the multithreading on the Dolphin bench, I found that disabling it only knocks down the speed 10% or so. Its not very far from a single threaded benchmark.

[tomwhite2004]

Ryzen 1700 [email protected] DX12 vs DX11 Dolphin Comparison - DX12 gains 100fps over DX11! That's a gigantic difference, I presume the perf increase is similar in other CPUs, can anyone else verify?

You have an AMD GPU which have very poorly optimised drivers with DX11 across the board, so its not a reflection on the CPU's ability. If you have Nvidia hardware OpenGL is the quickest and most stable.

[Josh128]

Ryzen 1700 [email protected] DX12 vs DX11 Dolphin Comparison - DX12 gains 100fps over DX11! That's a gigantic difference, I presume the perf increase is similar in other CPUs, can anyone else verify?

You have an AMD GPU which have very poorly optimised drivers with DX11 across the board, so its not a reflection on the CPU's ability. If you have Nvidia hardware OpenGL is the quickest and most stable.

Even with one session at 640x480? OGL was about as fast as DX11, but actually looked to have some real strange jerkiness, almost like hard drive loading/spooling issues, while the DX drivers were very smooth. R9 285 has been out for 2 years now, drivers should be fairly mature. Can anyone else confirm this difference only exists for AMD gpus?

*Edit-- I can test this on an i5-2500 machine-- I just put a GTX 1050 low profile in my sons Win7 machine a couple weeks back. I will run the DX11/DX12 comparison in it and see what differences show up...

[Josh128]

to point out that they still run MAME quite well despite the apparently widespread notion that they are not good choices for emulation.

And I'm not going to set up seven instances of Mame as I'm not really interested in going to such a extreme with a non real world scenario, but sure those extra cores help.

This discrepancy used to be ok as historically AMD have been much cheaper than Intel but the cpu you are testing is £300 which puts it into another bracket of performance altogether and just wouldn't be good value on an emulation focused build right now.

When the 6 and quad core Ryzens hit in a couple weeks you will be looking at stuff half the price of this processor, with likely identical performance per clock as I'm getting here, so theres the value proposition. I agree with the notion that any 8 core is overkill for a MAME-centric build, but the tests here are not to prove the 1700 is the chip to get, rather how the new arch performs in emulators, information which I'm sure you'll agree has been sorely missing /hard to come by since the release of Ryzen.

I still have a working i5-2500 rig with which to test the MAME multi-session stuff, and will do just for sake of comparison, but I would still like to know how a fast dual core (like your chip) would handle such a load. Perhaps you could do the single, dual, triple and quad Dolphin session test as I did, (much easier to setup! )

[tomwhite2004]

Even with one session at 640x480? OGL was about as fast as DX11, but actually looked to have some real strange jerkiness, almost like hard drive loading/spooling issues, while the DX drivers were very smooth. R9 285 has been out for 2 years now, drivers should be fairly mature. Can anyone else confirm this difference only exists for AMD gpus?

*Edit-- I can test this on an i5-2500 machine-- I just put a GTX 1050 low profile in my sons Win7 machine a couple weeks back. I will run the DX11/DX12 comparison in it and see what differences show up...

Dolphin in game will always use your GPU to render, even at 640x480, I'm pretty sure it doesn't have a pure CPU only mode like PCSX2 does. Even if you did not have a GPU Dolphin would revert to using integrated graphics instead.
So the better the GPU the more Dolphin can stretch its legs in an unlocked game example like F-Zero. Of course the CPU is important but like I said earlier every new CPU on the market will allow full speed in most cases and wont be the bottleneck.

Its well known that AMD has a weakness with DX11 and Open GL compared to Nvidia, its why people got so excited for the latest cards they bought out as they promised to make huge gains with upcoming DX12 titles as AMD currently have better DX12 drivers.

A better test for you would be to use the old i5 with the same GPU rather than switch it around, after all its the CPU you are looking at.

When I get the time later will run a single instance of F-Zero but it wont tell us much as we don't have the same cards. The actual benchmark I posted earlier in the thread is still the better marker of what the CPU can do with the program. Sorry, have no interest in running multiple instances, like I said its obvious that a multicore chip will perform better and serves no practical purpose.

[Josh128]

PCSX2 Tests. Stock Ryzen 1700 clocks, 8G 2667MHz RAM, standard PS2 resolution. Difference between OGL, DX9, and DX11 was negligible. Full speed on all titles tested. All games emulated from my personal discs.

Gradius V




MGS3




GOW1




Capcom Classics Collection

[orange808]

Dolphin in game will always use your GPU to render, even at 640x480, I'm pretty sure it doesn't have a pure CPU only mode like PCSX2 does. Even if you did not have a GPU Dolphin would revert to using integrated graphics instead.
So the better the GPU the more Dolphin can stretch its legs in an unlocked game example like F-Zero. Of course the CPU is important but like I said earlier every new CPU on the market will allow full speed in most cases and wont be the bottleneck.

Its well known that AMD has a weakness with DX11 and Open GL compared to Nvidia, its why people got so excited for the latest cards they bought out as they promised to make huge gains with upcoming DX12 titles as AMD currently have better DX12 drivers.

A better test for you would be to use the old i5 with the same GPU rather than switch it around, after all its the CPU you are looking at.

When I get the time later will run a single instance of F-Zero but it wont tell us much as we don't have the same cards. The actual benchmark I posted earlier in the thread is still the better marker of what the CPU can do with the program. Sorry, have no interest in running multiple instances, like I said its obvious that a multicore chip will perform better and serves no practical purpose.

Nobody uses software rendering, so testing that way is equally impractical. Nobody does that.

[Josh128]

Dolphin in game will always use your GPU to render, even at 640x480, I'm pretty sure it doesn't have a pure CPU only mode like PCSX2 does. Even if you did not have a GPU Dolphin would revert to using integrated graphics instead.
So the better the GPU the more Dolphin can stretch its legs in an unlocked game example like F-Zero. Of course the CPU is important but like I said earlier every new CPU on the market will allow full speed in most cases and wont be the bottleneck.

Its well known that AMD has a weakness with DX11 and Open GL compared to Nvidia, its why people got so excited for the latest cards they bought out as they promised to make huge gains with upcoming DX12 titles as AMD currently have better DX12 drivers.

A better test for you would be to use the old i5 with the same GPU rather than switch it around, after all its the CPU you are looking at.

When I get the time later will run a single instance of F-Zero but it wont tell us much as we don't have the same cards. The actual benchmark I posted earlier in the thread is still the better marker of what the CPU can do with the program. Sorry, have no interest in running multiple instances, like I said its obvious that a multicore chip will perform better and serves no practical purpose.

Nobody uses software rendering, so testing that way is equally impractical. Nobody does that.

Exactly. I want to know how it will run real emulators, in game. Sure its fine to see a dry CPU only run, but that doesn't show how the games actually run when you play them. DX12 gave a gigantic boost to my FPS in Dolphin, that's it. Is it due to using an R9 285 rather than an Nvidia card, as this guy claims, I don't know yet. Due to some Win7 issues, I have not been able to test DX11 vs DX12 Dolphin with an Nvidia card yet. Currently trying to get it updated to allow Dolphin 5 to run. PCSX2, on the other hand, shows no difference between OGL, DX11, and DX9.

Also, it seems AMD are prepping some new microcode that will be released in the form of a BIOS update very soon. This update is supposed to improve apps that are latency bound and resolve some other minor issues. It may or may not improve emulation performance (which I can say from my testing, is already damned good) but we shall see. You can be sure I'll try it.

[tomwhite2004]

Nobody uses software rendering, so testing that way is equally impractical. Nobody does that.

They absolutely do in PCSX2 as hardware mode introduces quite a lot of issues with many textures and effects. See the black splotches of colour on the ground on his Metal Gear Solid 3 screenshots? That's an issue with the hardware mode not rendering the light coming through the trees properly, switch to software mode and its gone as it is more accurate.

Off the top of my head Okami, Yakuza 1+2 and FF12 all have pretty bad issues with offset bloom and ghosting in hardware mode which can only be resolved by running in software. In fact hardware mucks up most instances of bloom it seems, you only have to skim through the wiki to come across games that are best played in software.

So until they improve the accuracy of the hardware mode software will still be used extensively, which is a shame as being able to render in non native resolutions is a real plus.

Anyway, my point was that software mode in PCSX2 is a great way to remove the variable of the GPU altogether so we can really see what the CPU is capable of, which after all I would have thought is the point of this exercise. Like I said earlier, we know ANY new desktop CPU will cope with all this stuff tested so far at full speed, that's why its best to find ways to actually stress it.

DX12 gave a gigantic boost to my FPS in Dolphin, that's it. Is it due to using an R9 285 rather than an Nvidia card, as this guy claims, I don't know yet.

As AMD have worse DX11 drivers (seriously just google it) the percentage gains their cards saw with DX12 are much higher than an Nvidia user would have seen as a) their DX11 performance was already good and b) the Nvidia DX12 driver isn't up to the same standard as AMD yet.

In Dolphin the DX12 backend also has the effect of reducing the CPU load too.
https://forums.dolphin-emu.org/Thread-u ... 12-backend

So its not possible for an Nvidia user to see such huge gains as AMD from a comparable GPU that supports DX12.

[bigbadboaz]

For those interested, Scorpio reveal is actually this Thursday:

https://www.forbes.com/sites/davidthier ... a2ee457194

Digital Foundry should make for a good breakdown, and implies MS does indeed have something to be proud of on hand.

[Josh128]

Like others, I call it will be some kind of Zen variant, either that or a 3+GHz FX model to have some talking point to really separate themselves from PS4 Pro... If its Ryzen, it will be the first time in history that a home console releases on the same process node as the most state-of-the-art chips on the planet. That would be huge.

[Josh128]

Checked some Dreamcast emulation, Null DC 1.0.4 Final. Ryzen rips Crazy taxi, over 220fps on average, smooth as butter, no hiccups... Can try more games later, just had this one handy. BTW DeMUL sucks, NullDC is the way to go for DC emulation.

[Josh128]

Heres how Ryzen handles both GC and DC at the same time, will try to add PS2 when I have more time...

[Josh128]

Dolphin, PCSX2, NullDC simultaneous run. No noticeable effect to speed of either...

[Josh128]

Full display of Ryzen power. DC, PS2, GC, Atari Seattle arcade hardware simultaneous run. All in game. Check those framerates/percentages.

[bigbadboaz]

Pretty fantastic. How do you find NullDC and PCSX these days as far as actual play quality?