CPU Ryzen benchmark

For me it comes down to real world performance and cores. If two processors cost the same and one of them has twice the cores, I'm buying that one because in the future, software, games etc will be geared more toward that processor. In the real world you see a few fps difference here and there in games. Does that few fps really matter when you're getting 150+fps anyway? Sure if you're only getting 30fps then it's like a 10% difference...

 

If your already on a Haswell or above cpu ( 4 cores k models) and strictly play games, you will always be better putting the cash towards the gpu as a gamer.

The general perception is cpu performance is at or near limits for many end users, so spending cash on a 8 core when you already have good enough is a tough ask and that's what AMD is asking.

Even the much fabled 2600k gets obliterated by a stock clocked 6700k or 7700k in games and general performance, gains can be as much as 50% on some titles. Problem is if you overclock the 2600k it gets very similar minimums and not to much worse averages. It's only when it's paired to a 1080 or titan that the differences would be bigger.

And that is only sub 1440p, as the resolution increases so does the gpu demands, put everything at 4k and most titles will offer the same fps on any modern cpu in the latest games.

I personally feel Ryzen 3 and 5 will offer better value for money as a gamer that's my feelings on it.

 

No, it just requires the right measurement.
Prior to frame time capture, the best that the frame delivery issues with Crossfire could be quantified as was "it feels juttery" or "it doesn't feel as fast as the FPS counter says". With the availability of frame-rating tools, and the ability to capture frame render times at multiple points in the timeline (frame submit/present time with APItrace or FRAPS, frame render exit time with FCAT or similar) it is possible to track the path of individual frames through the graphics pipeline. You can then quantify the variance in frame delivery and the source of that variance. Add on top of this loop latency measurement (e.g. input-timing with high-speed cameras and controllers modded to show input LEDs, or with photodiode triggers and dedicated timing equipment for applications that support it) you can get an excellent objective measurement of total latency, latency variance over time, and thus smoothness of gameplay.

For CPUs in particular, variance in submission time is likely going to be the major effect, with variance in render times being down to the GPU.

 

Except, that is patently wrong as per above.

Get a pair of GTXs 580s - whack them in SLI and max out Battlefield 3 at 1440p. You'll get 60fps but it'll be juddery as all hell (see Damien C's issues with Star Citizen in that thread).

Smoothness is all about frametime delivery.

 

Precisely.

You can Average 60 frames per second with variances anywhere between 10 frames to 100, as long as the GPU manages to drive sixty frames in that second; it'll look like 60 FPS, it'll just feel like a train wreck.

Vsync alleviates this by forcing the GPU to deliver at precisely 16.6ms intervals for 60 FPS, but that introduces input lag due to pre-rendered frames and other issues.

If Ryzen can properly delivery on a much tighter set of Frametimes; it'll be a very reasonable chip for gaming, especially once the R5 series releases with the Hex and Quad core parts to fight with Intel. The Current R7s aren't really regarded as competing with things like the 7700K by AMD (Because they lose), they're inteded to be competition for the HEDT parts on 2011 for a much lower price.

They've just got to fix all the little issues with Ryzen, currently there's a whole host of issues that need resolving before I'll be willing to put my money where my mouth is.

This is why Frametimes are also very interesting to look at, since that gives you an insight on how long each frame took to render, and you can see some hilarious variances while still managing 60 FPS, since it can seesaw between 30 & 8 ms render times to give 60 FPS while feeling like it's running as well as a bicycle with square wheels.

 

Hah! Speak of the devil, and he will draw you a graph: Techreport have supplemented their existing frame-time graphs with instantaneous framerate histograms. No sign of Ryzen having smoother frame deliver, and some minor weirdness in GTA:V with a 'lumpy' distribution, still above 60FPS instantaneous so if your refresh rate is 60Hz this will not be visible when using VSYNC (no frames delivered below 60FPS so no reason not to run with VSYNC on).

 

Whilst is doesn't say Ryzen delivers a smoother experience what it does show is that despite a >20% deficit in average frames and not reaching the heady heights the smoothness is much the same as the choice gaming CPU the 7700K, so you wouldn't feel that deficit and older chips like the 3770k etc. don't match the newer ones.

I’m up for Ryzen, I’ll give it a while for mobos to stabilize etc, but there’s no way I’d be buying less cores, certainly won’t be picking up a 6 core, I’ve gone from a 3770k at 4.5Ghz to an 3570k @4.8Ghz and that has definitely been a gaming down grade in a multiGPU setup playing at 4k. The internet will tell you that you don’t need more and average frame rates and benchmarks similarly but from my experience, more cores trumps the clockspeed advantage to a certain extent.

Ideally though I’d like to be running Ryzen with at least 4.5Ghz for those games that do rely on brute force clockspeed and >3Ghz RAM, I doubt it will be long before that is possible.

 

And just like FCAT was needed , likely NEW software is needed.

now that hundreds of users are saying the same thing - the correlation is the reviewers might not know everything at all....

 

Looks like the bug with Windows 10 scheduler has been confirmed.

Does that mean all benchmarks done on Ryzen will have to be redone once the bug is fixed?

 

Can't say as that surprises me, Win 10 is a sh*te OS (which is why I've stuck with Win 7 Ultimate x64).

EDIT - And if that is the case, then I think it's fair to say it invalidates pretty much every review and benchmark thus far

 

Exactly, bouncing around between 30 and 90 feels nothing like bouncing around between 55 and 65.

For me the problem is that the screen tearing is so painfully obvious if you go beyond what the screen can handle that leaving it off isn't much of an option.
Luckily some of the downsides of V-sync are addressed with Adaptive V-Sync.

 

Sounds like it if they were performed in Windows 10

 

They will need to be redone, for sure. Win 10 is at fault and it can only be rectified with a new testing round once the patch is released.

 

Would that be When a patch is released? Microsoft is always slow to fix these issues. Took 4 patches and around a year to solve Intel hyperthreading issues in the passed.

The post even confirms it won't be the quickest fix.

Microsoft always seems to screw stuff up. Then takes ages to fix out, w10 meant to be the future but it seems grossly inadequate for certain things, now even cpus are struggling with compatibility

 

Win 10 has been one of the worst operating systems for me to use in the past decade. Constant issues. I'd even go so far as to say that Vista was better than this. Vista had its issues, for sure, but Win 10 takes it to another level.

 

And then Microsoft wonder why people buy Apple products for day to day non gaming use

Their phones sucked owing to the OS and lets face it Windows 10 sucks

 

I think my main bugbear with Windows Vista was mainly the fact that you needed a pretty hefty PC to get the most out of it. Oh, and the fact it switched from hardware sound. That was the worst of its crimes, IMO.

I installed it a couple of years back and it was fine. Looked really nice too.

 

That's true. I remember a friend trying to run Vista Ultimate on 512MB of RAM...

 

To your average laymen equipped with an iPhone and iPad...

It's quite true Gareth

And Windows 10 does suck