Considering a new CPU?

[Denniss]

Read some real tests of games with proper multicore support. Be aware that many/most Ryzen test were made with nvidia GPU and the drivers do sometimes act strangely with Ryzen, especially in DX12.
7700k is only better in games with limited core support, living from its (turbo) clockspeed advantage then.
In the Ryzen series the six-core R5 1600/1600x has best price/performance ratio.

[BoilerDown]

Its your allegation, provide us some gaming benchmarks where Ryzen wins over an i7 4790k, 6700k, or 7700k.  Because every one I've seen has Intel winning.

[Bizman]

I just read a comparison review in our leading household IT magazine. The Ryzen was superior in crushing digits, but for gaming they recommended Intel for better compatibility. So the superiority depends on what games you want to play.

[Denniss]

It's not a problem of compatibility, it's a problem of scalability or how far game developers were able to think.
Most stopped at 4 core/8 threads, some may have optimized to 6/12 to cover the exceptionally horridly priced Intel sixcores.
As always with introduction of a completely new platform and CPU design there are several problems to resolve, especially in Software. Motherboards/Bios already work better than on initial release, Win10 should have received an update to properly handle or disable Core parking on Ryzen (disabled on Intel) and to its sheduler so it assigns worker threads to physical cores first + to keep threads within the same Core CompleX, etc, etc.
Time will show that no one with an open mind regarding CPU performance should buy 4-cores for higher-end gaming anymore.

[Pudgie]

It's not a problem of compatibility, it's a problem of scalability or how far game developers were able to think.
Most stopped at 4 core/8 threads, some may have optimized to 6/12 to cover the exceptionally horridly priced Intel sixcores.
As always with introduction of a completely new platform and CPU design there are several problems to resolve, especially in Software. Motherboards/Bios already work better than on initial release, Win10 should have received an update to properly handle or disable Core parking on Ryzen (disabled on Intel) and to its sheduler so it assigns worker threads to physical cores first + to keep threads within the same Core CompleX, etc, etc.
Time will show that no one with an open mind regarding CPU performance should buy 4-cores for higher-end gaming anymore.

I've been doing a lot of testing w\ my Intel I7 5820K 6-core CPU, even though it isn't an AMD Ryzen 5 1600X CPU, it is a very good comparison as to the issues that Denniss has posted concerning a Ryzen CPU as both of these CPU's (this also goes for Ryzen 7\I7 6900K as well) are gonna suffer from a lack of scalability\usability from the typical gaming software being used as most of this software was optimized to be run on a max of 4 physical CPU cores....HT\SMT aside. All HT\SMT does is ensure that the 4 physical CPU cores are FULLY utilized (this is the job of the OS) as the current batch of gaming software is concerned as most gaming software isn't very multithreaded capacity-wise as there isn't much that can be multithreaded code-wise due to the makeup\nature of most games, meaning they won't take full advantage of a CPU equipped w\ more than 4 physical CPU cores equipped w\ large L3 caches & fast, wide interconnect branches.....much less w\ HT\SMT added on top of it. AHIII is 1 of the few exceptions in which it does create multiple threads from a single process but I can show you that even AHIII as it's currently coded can't keep the 4 CPU cores busy enough under a greater than typical game load (default in-game settings) to prevent them from occasionally going into low power states (sleep mode) that I have set CPU affinity\priority in the OS on my I7 5820K CPU....that is also not using HT as well, due to lack of sufficient CPU core usage (the game is running on these 4 CPU cores exclusively w\ all else running on the other 2 CPU cores & even these 2 CPU cores do not get utilized enough to keep them from going into low power states). On these type CPU's when gaming is concerned this action causes a LOT of CPU core latency issues from having to wake CPU cores back up when threads are given to them when in this state & when you add HT\SMT into the mix you will cause some more latency thru the OS thread schedulers trying to determine which CPU cores are the physical 1's vs the logical 1's so that they're used properly AND which CPU cores to PARK due to lack of use & UNPARK when needed, because the Win OS thread schedulers use a token ring philosophy to assigning threads (round robin approach starting in numerical order....Core0, Core1, Core2) even more latency is created by the OS schedulers switching threads across the entire SMP structure instead of staying w\ the CPU cores close by (from CPU core0 to CPU core5...as the CCX issue is on Ryzen CPU's, just not as severe) which ends up making more thread duplication in the L1-L2 caches which the OS spends more time cleaning this up to prevent drawbacks, contentions, etc which causes more CPU latency. The list can go on...................

This excess CPU latency is the main issue IMHO for these types of CPU's running game software, and at this time the only remedy that I see, until the OS's are revamped to better handle these CPU's when under lighter threaded loads and game softwares are written to make more efficient use of all the "extra CPU cores past 4", is for a user to apply CPU core priority\affinity w\o using HT\SMT on these types of CPU's to "make up" for all this.....I can also show that this does work & work well w\ my Intel I7 5820K 6-core CPU & I firmly believe that the very same type results can be gained w\ AMD's Ryzen CPU's when handled in the same manner simply due to similarity in concept even though there are slight structural differences between the 2 CPU's.

Intel got away w\ this due to creating an environment in which they could inflate the price of these parts by designating these within a fully separate subculture of computing from the mainstream gaming culture knowing that as long as there wasn't a viable competitor that could actually marry these 2 computing subcultures (gaming\prosumer) together into 1 competitive package that could provide good performance in both subcultures at a competitive price\performance bracket, they didn't have to & folks would pony up the moola..............

AMD, whether you like it or not, has just busted this wide open w\ Ryzen CPU's on the AM4 socket....a single socket platform that will support the ENTIRE consumer computing spectrum....from entry level APU's to full blown prosumer usage...including enthuiast level gaming within this spectrum.....for similar\less costs than an Intel Zxxx mainstream platform costs. Intel cannot compete w\ this outside of it's vast mindshare advantage & they know it.

With the advent of the X299 platform Intel will now "fix" what they already should've fixed earlier starting w\ X79 forward......due to AMD's success w\ Ryzen in spite of the small hiccups w\ it & it wouldn't surprise me at all to find this X299 platform suddenly being equipped w\ dGPU capability in it's CPU socket to fully compete w\ Ryzen on the AM4 socket....all that is left to see is how well the AMD Raven Ridge APU's (Zen CPU\Vega GPU on single APU interposer using HBM2 mem) do on the AM4 socket...........

From what I have witnessed\learned w\ my Intel I7 5820K 6-core CPU, I definately will not go back to a 6-core CPU, much less a 4-core CPU, going forward & until Intel does better than what it is currently doing it won't be an Intel platform either when I can get a 8-core\16 thread AMD Ryzen 7 CPU that can perform in gaming on a close par w\ Intel then equal\exceed Intel in other areas of computing for less money to invest\maintain than an Intel 4-core CPU mainstream platform equivalent................... ....

Can't wait for a full, ground up, water-cooled AMD Ryzen 7 CPU\AMD Vega 10 GPU package build.....even though I'll have to swallow a Win 10 OS install on it (which I already have a legal retail version of on hand)....................

Just as soon as I get retired & receive my severance\vacation pay.............coming hopefully very soon. Along w\ a little Mopar muscle car from my youth days that I've been keeping an eye on....................

Going w\ the underdog this time around..............

My 2 cents.............

 

[Skuzzy]

It's not a problem of compatibility, it's a problem of scalability or how far game developers were able to think.
Most stopped at 4 core/8 threads, some may have optimized to 6/12 to cover the exceptionally horridly priced Intel sixcores.<snip>

It has nothing to do with game developers being able to think.  If you knew anything about game design, virtually everything in a game relies on a sequence of events, which much occur in a specific order.  Games are highly synchronous.

You just cannot go off hey-willy-nilly and create threads.  Threading works best when you have asynchronous events to deal with.  Not much in a game is actually asynchronous.  Now you go and think about it before you go off insulting game developers ability to think.

Something else to consider.  Most things which can happen asynchronously are already threaded.  Such as audio.  At any given time there can be eight, or more, threads spawned just due to audio.  Unfortunately there are no tools with small enough granularity where you could see them happen.  They spawn and die very quickly,for the most part.

Here is a simple example.  Pull the trigger, you hear the gun fire, at the same time you hear the engine running, at the same time you can hear explosions, or other vehicles in the area.  Each sound is a thread.  At the same time, the audio has to be synchronized with what you are seeing graphically, therefore some control has to be maintained with that, which prevents it all from being fired off into different threads.

I am assuming the reader has a modest understanding of how Windows threads work here.  If you have to synchronize events, you, typically, would not fire off threads for those events as it is clumsy to try and use the Windows inter-process communications facility to manage synchronous events and there is no benefit, other than adding more overhead to use threads for synchronous events.

[Bizman]

Denniss, thanks for correcting my wording. Scalability sounds much better than compatibility.

Then again, since games can't make use of threading because of being synchronous, the Ryzen is not the best choice for gaming. So would it be correct to say that most games are incompatible with the Ryzen?

[Skuzzy]

Games can use some threads, but it will all depend on the game.  We have a couple of threads, native to the game, which we use. 

Most games though, which advertise a lot of multi-threading are simply counting a number of the other threads Windows runs in DirectX.  Technically, it is accurate to claim those are yours, but it is something you get for nothing.

There are a lot of threads related to getting one frame drawn, to be able to run those parallel pipelines in the hardware.

[Drano]

Hey something silly that just popped into my head reading this thread which I'll admit much of which towers over my head. Thinking of the micro freeze thing. Could it be that because AH3 is putting so much more emphasis on the GPU vs AH2 did that the parts of the CPU the game uses are going to a dead stop in some respects even if only for a few milliseconds? Think that could be the cause? Any way to put some work back on the CPU constantly if only just enough to keep it rolling all the time? Some stupid calculation or something? Just a thought.

Sent from my XT1585 using Tapatalk

[Skuzzy]

That should not happen, unless Windows power management is shutting down the CPU or any core.  Hmmm,..possible I suppose, but unlikely.

Easy way to fix that is to use the correct power management settings (like none!).

[Pudgie]

That should not happen, unless Windows power management is shutting down the CPU or any core.  Hmmm,..possible I suppose, but unlikely.

Easy way to fix that is to use the correct power management settings (like none!).

Skuzzy,

I have set up Windows 7 Power Plan to High Performance plan, have already made all the settings changes that you've posted within it, then for good measure went into the UEFI and disabled both Intel CEIST (which the OS will hook into to control the CPU power\frequency thru any of the 3 power plans according to CPU load if this is enabled in the UEFI....all 1 really should need to do is to disable this in the UEFI (or BIOS if you prefer) to effectively kill any Windows power control of the CPU regardless of the power plan chosen within Windows.....AMD equivalent is PowerNow!) and Platform Power Management (the other power control settings that the OS will hook into to perform any other power management of this X99 platform that it can control) so there is absolutely no path available for Windows to control any power aspect of my X99 platform....both at the platform level, the CPU level & the OS level itself as if none of this is brought across within the UEFI cache for the OS to hook into once the OS boots up then the OS simply can't control it as it simply isn't present.

Correct?

Then running the game using RivaTuner Statistics Server's overlay displaying all of MSI Afterburner's GPU\CPU graphs....along w\ using HWINFO to provide the CPU clock speeds per CPU core (6 in my case) due to HWINFO ability to hook into the Intel CPU driver cache within the UEFI cache that Windows uses to configure itself during bootup AND to use the RivaTuner OSD in conjunction w\ MSI Afterburner as well, I can clearly see all 6 CPU cores of my I7 5820K 6-core CPU occasionally dropping from the TurboBoost clocks of 3.7Ghz down to the low power state clocks of 1.5Ghz....all while playing the game under Dx11 under full game load as read by both the in-game HUD & the RiveTuner overlay w\ all pertinent data present running in real time...............

The only explanation for this that I can see is that this CPU power control is actually written into the Intel CPU at the hardware level & it's driver will operate accordingly at a base level on-die when a CPU core is not being used within a certain amount of time at\below a certain CPU load threshold.......OR it could be a granularity\synchronization issue between the 2 pieces of software that their process CPU priorities are read by the OS as being "below normal" (which most monitoring softwares are designated as below normal CPU priority for good reason so they don't tend to HOG CPU core processing time from other running processes....like the OS for starters....), which will mean that these are given much less CPU processing % time AND fairly long time slices due to their lower priority designation so can be actually stalled waiting in the pool for a CPU core to come loose to process their threads.....thus can conceivably "miss" the actual CPU core operating state in the process.......

HWINFO reads the CPU core clock minimum clock frequency as reported from the UEFI, the maximum CPU core clock frequency as reported from the UEFI then reads the current CPU core frequency as reported thru the UEFI in "real time" as measured thru the polling frequency as set up in the HWINFO software....... Just so you know, I have ran this as fast as 100 ms polling rate so the "miss" window is fairly small.....................

So it is either A.) Intel CPU's actually do have hardware onboard\on-die CPU power\frequency control capability at a base level controlled from it's CPU driver running independent of any OS support or B.) the authors of both RivaTuner Statistics Server\HWINFO software haven't put out good, well written software that runs well enough on this Win 7 OS for their results to be considered "trustworthy".........................

Until some concrete evidence is provided that will solidly discredit the fidelity of these 2 programs, I won't pass judgement in this area & will instead go w\ option A as this option does make the best sense case for Intel to provide some form of ample CPU protection that a user can't mess with\mess up to maintain their CPU structural fidelity claims.........obviously for legal implications, of course........

So all I will say in closing is that I've witnessed my Intel I7 5820K 6-core CPU actually sending lightly used CPU cores into a low power state on my box in real time under a decent, actual AHIII Dx11 operating game load & can confidently say that the Windows OS had no part to play in any of it as it had no path to gain access to interfere at all...........

YMMV......................

 

PS---Almost forgot, I will record & post a running video of my box running AHIII using Dx11 w\ all this overlaid on screen being recorded in real time for all to see for yourselves then you can make your own judgements afterwards................... .........

 

[Skuzzy]

Did you disable "Intel SpeedStep Technology" in the BIOS?

That is Intel's hardware power management.  I always disable that.  Many OEM computers do not allow access to that setting.

[Toad]

Hypothetical:

If one wanted to buy an Intel CPU that supported Win7, what CPU would provide a good price/performance choice while having the horsepower to run AH3 without breaking a sweat? (Coupled with adequate RAM and a vid card in the GTX 1070 range.)

I see that _some_ of the "6th Gen" Skylake CPUs are supposed to support Win7 but not all. Apparently all the "5th Gen" CPUs will support Win7.

So where's the sweet spot?

[Bizman]

Did you disable "Intel SpeedStep Technology" in the BIOS?

That is Intel's hardware power management.  I always disable that.  Many OEM computers do not allow access to that setting.

Are you going to add this to your Hints and Tips?

[Skuzzy]

Are you going to add this to your Hints and Tips?

I tend to shy away from telling users to go into the BIOS and make any changes as they can, very quickly, render their computer a brick.  And most OEM computers do not grant access to that feature.

The overall power management scheme in Windows needs a drastic overhaul.  With hardware doing power management and software doing it, it quickly becomes a mess.