AMD Ryzen - Games in 1080p can improve more than 35% !

[PCGuy_5960]

Just now, N3v3r3nding_N3wb said:

Well, it's a single die, just with two CCX on it.

Yep. I will fix my post now

[PCGuy_5960]

2 minutes ago, zMeul said:

as I said in the other thread, that's crap - it doesn't say anything

I'm searching something like this:

  Reveal hidden contents

^ that's Bulldozer btw

You can't find a diagram like this. Ryzen is too new for this

[N3v3r3nding_N3wb]

2 minutes ago, zMeul said:

as I said in the other thread, that's crap - it doesn't say anything

I'm searching something like this:

  Hide contents

^ that's Bulldozer btw

I should have made this more clear, but the article I linked was an article on the infinity fabric, not a source for the picture.  The picture is from GN, I believe.  That article should help you understand what Infinity Fabric is.

[Hunter259]

48 minutes ago, zMeul said:

mno!

HPET and power management (high performance) have absolutely nothing in common - nothing

don't understand what HPET does? do a google search - turning HPET off is a very very stupid idea

Actually no it's not. It's only a bad idea on anything older than Windows 8 as a new standard has taken it's place.

[N3v3r3nding_N3wb]

8 minutes ago, zMeul said:

as I said in the other thread, that's crap - it doesn't say anything

I'm searching something like this:

  Reveal hidden contents

^ that's Bulldozer btw

How about this?  Anandtech's article is very good and detailed.  I'd recommend reading it for architecture things (http://www.anandtech.com/show/11170/the-amd-zen-and-ryzen-7-review-a-deep-dive-on-1800x-1700x-and-1700/6).  Does this picture work?

[Droidbot]

29 minutes ago, Coaxialgamer said:

 

 

35% IMPROVEMENT!

(by reinstalling your OS, overclocking, and voiding your warranty!)

[LAwLz]

 

3 hours ago, imreloadin said:

What I found the most interesting was this:

 

<image>

 

You get an 18.5% increase by turning off HPET with everything else being the same.

By just changing the power plan to High Performance and leaving HPET turned ON you get an increase of 19.8% from the baseline.

 

Now I know this kind of removes the dead horse that zMeul keeps liking to beat by saying "But AMD helped create HPET!" everywhere but now users can just leave it on and switch their power plan to High Performance and they get more performance. This is with the same RAM speeds and NO overclock. By literally just going into the control panel and switching the power plan you can get an additional 19.8%

You're reading the chart incorrectly.

What it says is this:

Doing an OS reinstall, as well as reinstalling the game, and turning HPET off will give you an increases of 18.5%.

Turning on HPET and changing the plan power plan to high performance gives you an additional 1.07% performance.

 

You don't get an 18.5% increase by turning off HPET. We have no idea what benefit (if any) that has.

 

 

2 hours ago, Hunter259 said:

It's turned off due to a sleep bug more or less. The bug causes Windows timing to slow down which artificially inflates benchmark scores. HPET also can cause a very small hit to performance in certain cases. AMD just said to do that due to the bug and trying to gain every last drop of performance they can.

It's the other way around. It's when you have HPET turned off that the bug happens. If you have HPET on then you will have accurate time and your benchmarks will not be inflated. AMD however, recommended turning it off. AMD recommended reviewers running the system in the configuration where the bug occurs.

 

 

[TechGod]

13 hours ago, zMeul said:

what is this shit!?!? overclocking increases your performance!?!? who knew .... 

jeez H christ AMD are getting desperate

 

and the company that helped create HPET to recommend disabling it .. what the actual fuck!?

not even considering disabling HPET creates it's own problems

OMG BIG FAN, HAI!!!!

[Space Reptile]

14 hours ago, huilun02 said:

Inb4 someone comes in saying its a shit game and isolated case. 

well this is the case when the game ran PREVIOUSLY on a quad core and thus the config file said "4 cores" a fresh install wont see much of an impact at all 

[The Sloth]

every week someone post a similar thread and we began, nothing new to contribute just bashing.

7700k for gaming

ryzen for productivity and streaming. 

[MageTank]

11 minutes ago, nerdslayer1 said:

every week someone post a similar thread and we began, nothing new to contribute just bashing.

7700k for gaming

ryzen for productivity and streaming. 

And here I am, just showing up to bask in the glory of faster memory actually showing an improvement in gaming. SEE MOM? MY MEMORY OVERCLOCKING ISN'T A WASTE OF TIME.

 

Too bad Ryzen's IMC is still awful. 

[The Sloth]

Just now, MageTank said:

And here I am, just showing up to bask in the glory of faster memory actually showing an improvement in gaming. SEE MOM? MY MEMORY OVERCLOCKING ISN'T A WASTE OF TIME.

 

Too bad Ryzen's IMC is still awful. 

 

want a cookie for your effort? ryzen is fine ( hopefully, things keep improving with updates),  it doesn't "suck". buying an 8 core cpu ryzen/6900k just for gaming is not something people should do. 

[MageTank]

1 minute ago, nerdslayer1 said:

want a cookie for your effort? ryzen is fine ( hopefully, things keep improving with updates),  it doesn't "suck". buying an 8 core cpu ryzen/6900k just for gaming is not something people should do. 

Where in my post did I say ryzen wasn't fine? Where did I say it sucked? When did I recommend an 8 core CPU for gaming? I think you got the wrong person buddy.

[The Sloth]

1 minute ago, MageTank said:

Where in my post did I say ryzen wasn't fine? Where did I say it sucked? When did I recommend an 8 core CPU for gaming? I think you got the wrong person buddy.

 

i was generalizing the thread as a whole, its a pointless discussion that is going nowhere. 

[MageTank]

Just now, nerdslayer1 said:

i was generalizing the thread as a whole, its a pointless discussion that is going nowhere. 

Oh. I thought you thought that I thought zen was bad....

 

Thought.

[Ryujin2003]

15 hours ago, zMeul said:

I actually read it, did you?!

 

 

Really? 133.5 is 2.89% increase from 129.75 FPS on the previous  setting. The percentages underneath each FPS is the percentage increased from the previous setting. In the end, 133.5 FPS is 35.55% increase over the initial 98.5 FPS.

[Doobeedoo]

Urm, well we'll see more in line performance once Windows gets patch for HPET and power plat or whatever, alongside better BIOS and RAM speeds issue ironed out.

Once all that is so called done then we can only look for difference in future optimizations for games I guess.

[MageTank]

4 minutes ago, Doobeedoo said:

Urm, well we'll see more in line performance once Windows gets patch for HPET and power plat or whatever, alongside better BIOS and RAM speeds issue ironed out.

Once all that is so called done then we can only look for difference in future optimizations for games I guess.

A lot of us memory enthusiasts are starting to lose hope that the memory issue can be fixed. The fact that tertiary timings are completely locked out on both the C6H and Taichi (the flagships of the manufacturers that excel in memory overclocking) tells us that either A: they are hiding something, or B: the IMC is so strict, that touching any of the tertiary timings will result in instability. We even went back to Carrizo to see if it's IMC was locked down, and it wasn't. Even on cheaper boards. 

 

It might be possible that it was an oversight, something they had forgotten to include due to rushing the product out, but it just seems unlikely. The fact that 2DPC is absolutely choking the IMC, scares me. After all, making 2DPC stable is very simple work. Would literally take me seconds to make a 2DPC kit work on a board that isn't validated for it. I've done it before, as have many others. If they are preventing us from doing this, for this long a time period (quoted to be a few months, if ever) then there has to be another underlying issue.

 

So far, the only way to touch tertiary timings on Ryzen, is to do so via the bus/strap combo's. Even then, it's a total shot in the dark. You can't specifically choose which timings will change, and you cannot chose if they will get tighter, or loosen up. No enthusiast that I know of, has had any luck in this regard. 

 

I would buy Ryzen and test this, but I was just tasked with overclocking laptop memory on a Kaby Lake 7700HQ laptop. Sadly, this laptop has far more memory options available to us than Ryzen does on it's best boards, lol. 

[SpaceGhostC2C]

8 minutes ago, MageTank said:

A lot of us memory enthusiasts are starting to lose hope that the memory issue can be fixed.

It depends  on which issue you want to fix, though. Having the degree of memory OC control you seek is one thing, getting the sticks to work at their rated frequencies with one click or two is a different thing. If you pool reviews together, it seems everyone is having trouble to get the Asus boards to work, not managing more than 2933, but on Gigabyte and ASRock baords they got 3000-3200 in at most one BIOS update. Don't ask me which timings accompanied those frequencies, but I'm assuming whatever presets came with the sticks

 

I guess what I'm saying is: I don't know about the IMC characteristics, or whether you'll get anywhere near your target when OCing RAM in AM4. But when it comes to the average consumer just entering bios -> enable xmp -> profit!, I wouldn't be very pessimistic either.

[MageTank]

Just now, SpaceGhostC2C said:

It depends  on which issue you want to fix, though. Having the degree of memory OC control you seek is one thing, getting the sticks to work at their rated frequencies with one click or two is a different thing. If you pool reviews together, it seems everyone is having trouble to get the Asus boards to work, not managing more than 2933, but on Gigabyte and ASRock baords they got 3000-3200 in at most one BIOS update. Don't ask me which timings accompanied those frequencies, but I'm assuming whatever presets came with the sticks

 

I guess what I'm saying is: I don't know about the IMC characteristics, or whether you'll get anywhere near your target when OCing RAM in AM4. But when it comes to the average consumer just entering bios -> enable xmp -> profit!, I wouldn't be very pessimistic either.

I understand that, but it's also deceptive in a way to just look at frequencies as the be all, end all when it comes to "memory speed". A tight 3000 will always beat a loose 3200. In fact, my tight 3600 is currently smashing any 4000 XMP kit on the market. I am not asking AMD to give us faster memory XMP's. I am only asking that they let us tighten the kits we currently have available. As it stands, we are forced to settle with whatever the IMC thinks it needs when loading an XMP. From what I've seen, it's extremely inefficient. In dual-channel, a finely tweaked overclock can achieve 95-98% write efficiency with relative ease. Zen, under these XMP's, have been as bad as 80% write bandwidth efficiency. This is all ignoring the god-awful latency that could be remedied by tRFC, tREFI and some RTL/IO-L tweaking. 

 

I completely expect XMP profiles to improve with BIOS updates, as specific board vendors tweak their BIOS to better cooperate with the various IC's we have available on these DDR4 kits. I am however, very skeptical that AMD will give us the fine-grain control we get with previous AMD CPU's (and current Intel CPU's) because for all intents and purposes, there shouldn't be any reason as to why they were not included by default. I could very well be wrong, and I certainly hope I am, but it does look fairly grim.

[juri-han]

looks like intel better watch their back (i mean polis their stack of gold bars and look at the refection behind them)

 

3 months from now once all of the issues get sorted out ryzen will be a legit gaming CPU to boot, just in time for the 1600x and 1500X

[SpaceGhostC2C]

3 minutes ago, MageTank said:

(...)

I'm not denying any of it. I just got the impression that when you talk about RAM issues, and @Doobeedoo talks about RAM issues, you are not necessarily referring to the same issues  

[xAcid9]

4 hours ago, Dabombinable said:

Those going on about HPET, I might disabling this myself and see what happens, as its currently enabled in my rig (4790K+GTX 970+16GB DDR3 1333):

https://forums.anandtech.com/threads/do-you-have-hpet-enabled-or-disabled.2362109/

Got mine disabled since 2015, still waiting for something to blow up.

[MageTank]

6 minutes ago, juri-han said:

looks like intel better watch their back (i mean polis their stack of gold bars and look at the refection behind them)

 

3 months from now once all of the issues get sorted out ryzen will be a legit gaming CPU to boot, just in time for the 1600x and 1500X

You came to this conclusion after seeing a positive result in just one title, using tweaks that should be considered industry standard (aside from the HPET part)?

 

I don't know of a single friend or family member that uses the same exact OS after swapping platforms. They are also very aware of memory's impact on gaming performance (Mostly because I never shut up about it. If you people think I am bad about harping on this on the forums, meet me in real life, lol) and overclocking tends to show a positive improvement, even if it's not always a linear scale. 

 

I would wait for more titles to benefit from these changes before calling it concrete. 

[zMeul]

1 hour ago, Ryujin2003 said:

Really? 133.5 is 2.89% increase from 129.75 FPS on the previous  setting. The percentages underneath each FPS is the percentage increased from the previous setting. In the end, 133.5 FPS is 35.55% increase over the initial 98.5 FPS.

that was not I was refering to, but please . be free to ramble