Discussion AMD Ryzen MegaThread! FAQ and Resources

[jwcrellin]

smoothness comes from frame latency and other stats not fps.

 

[jaymc]

I wasn't saying it was from fps, actually ryzen had a lower average fps in this game than the 7700k did..


I think the cpu/gpu & cpu/usb keyb an mouse have a lower latency as a result of the architecture.. an the chipset on th MB being cut out of the equation. I'm sure it's a combination of things that are causing the smoother experience..

Well here I could be wrong but in my head it just makes so much sense.. I'm surprised others don't see it from same perspective... Maybe I shouldn't of opened my mouth but look if I'm wrong I'm wrong...it just makes so much sense to me... I really think this is whats happening...I will eat my words I'm sure if I am wrong about it.. but so be it I guess. If nobody speculates we wouldn't have a need for a forum.

I'm sure this is one of the reasons (not the only one obviously it will work pretty well on servers with Vega for GPU loads also) that they are running pci express directly into the cpu.. same goes for the usb 3.1 maybe it was a fluke but I doubt it.. it's gonna result in lower latency talking to the gpu an mouse etc.. if you cut out the chipset out completely.

input/output does not need to go through the chipset.. this has to reduce latency.

You can even see that they made sure that the two pci express lanes (slots) that are used for graphics cards (sli or crossfire) go directly to the CPU on every board I've seen anyway... The pci express lanes that hang off the chip on the MB are (or should never) never be used for the graphics cards (that I'm aware of.. it just wouldn't make sense either)... correct me if I'm wrong here guy's.

I'm sure this is for this reason.. So the CPU can talk faster to the GPU..

An also when I'm plugging my mouse an keyboard into the MB, I will be making sure that it's into the USB slot that goes directly to the CPU. An not via the chipset...for tha same reason, although this I'm sure will not make as big of a difference as running the GPU direct to the CPU, I guess.. but still !!

Edit: spelling grammer an extra words etc

 

[-Fran-]

TR already graphs frame times and Toms also has a similar methodology and in both I see Ryzen in-line with Intel. Those are the numbers, but they depend on how the data is gathered.

I do remember there's a site that does subjective interpretations of numbers seeking a "playable experience" as a goal when reviewing... Was it HardOCP? Well, I know there's a site that uses a lot of subjective interpretations, but also provides a lot of good data to backup their claims.

After all, like I said a few posts away, if this is something that jumps to the eye and reviewers don't see it reflected in their numbers, they would at least make a comment on it, right? Something like "yeah, although the numbers look similar, Ryzen still felt smoother". That would at least prompt someone to seek for a methodology to measure that and put it into numbers. There's always a way!

Well, I do hope that is the case anyway. AMD needs to get back in the game; pun absolutely intended.

So, any news about the APUs? I still feel more excited about them than the R5 and R3, TBH.

Cheers!

 

[jaymc]

TR already graphs frame times and Toms also has a similar methodology and in both I see Ryzen in-line with Intel. Those are the numbers, but they depend on how the data is gathered.

I do remember there's a site that does subjective interpretations of numbers seeking a "playable experience" as a goal when reviewing... Was it HardOCP? Well, I know there's a site that uses a lot of subjective interpretations, but also provides a lot of good data to backup their claims.

After all, like I said a few posts away, if this is something that jumps to the eye and reviewers don't see it reflected in their numbers, they would at least make a comment on it, right? Something like "yeah, although the numbers look similar, Ryzen still felt smoother". That would at least prompt someone to seek for a methodology to measure that and put it into numbers. There's always a way!

Well, I do hope that is the case anyway. AMD needs to get back in the game; pun absolutely intended.

So, any news about the APUs? I still feel more excited about them than the R5 and R3, TBH.

Cheers!

There commenting on it alright mate... here's couple examples. I was wrong (dough) about direct GPU access being responsible as most intel chipset's have direct access stupid boy jay... slightly older might just have one lane.. so that aint it..

looks like it could be mouse response times though...

This is from Techspot review:
"GTA 5 for example plays really well on the Core i7-7700K, but every now and then a small stutter can be noticed, while the 1800X runs as smooth as silk, sans stuttering from what I observed."
http://www.techspot.com/review/1345-amd-ryzen-7-1800x-1700x/page7.html

This next reviewer is basically saying the mouse response in Ryzen is way better.. An may be responsible for the smooth feel.
He wants wendell to test mouse response times.. He say's almost the very same thing about the mouse response not going through the chipset... He say's he has 10k hours logged in fps shooter's an he's convinced in smoother gameplay, He does say that skylake an kabylake have direct GPU access... Anyway here he is:
https://www.youtube.com/watch?v=88IOcFE4yho

Also when it comes to a proper dx12 game that can throw 16 threads I guess SMT is better in AMD, than HT is in Intel as well...

My mate was saying to me the game I was talking about, where the 7700k had better fps but was stuttering in the video when Ryzen was perfectly smooth with lower fps, was GTA. I recall he showed the intel chip stuttering in his video (as commented on by techspot above)... and showed Ryzen smooth.. An it was as plain as day to see...Wonder was it your man Steve from Techspot actually. I don't think I'm ever gonna find the video (I've been looking)... I give up for now anyway..

Maybe it's a combination of better mouse response, better SMT and more cores than kaby lake for smoothness.

Hopefully Wendell or someone else for that matter will please check this out and soon...

Edit:
By the way Yuka.. that sounds like an excellent idea about that site..."playable experiences"...

Also another reviewer just came to mind..
AdoredTV commented on the feeling of tomb raider being really smooth with Ryzen in his video as well, He even went on to comment then about the mouse feel !

 

[randomizer]

Maybe so guy's but it makes sense that the smoothness in game would be attributed to the chipset being cut out of the middle.. the cpu is now talking directly to the gpu and the usb keyboard an mouse.. I have asked reviewer to look into it so hopefully we will find out soon if this is indeed the case.

Even if this was the case, the difference in latency would be measured in microseconds at worst. Humans struggle to detect latency measured in the low millisecond range. Most USB devices have a polling rate of no greater than 1000Hz because of this.

I can understand what u guy's are saying that maybe it's all in the head.. but everyone seem's to agree on this one point... isn't that unusual in itself.. for everyone to agree on something across the board ?

Nope, that's perfectly normal. It's why we have groupthink and pluralistic ignorance. Which is not to say that this is a case of either, but that consensus is not as reliable as you might think, and certainly not unusual.

If it exists we can probably measure it. So someone should just measure it. Or at the very least they should do a double blind test to remove confirmation bias.

Maybe it's a combination of better mouse response, better SMT and more cores than kaby lake for smoothness.

More cores certainly helps with reducing thread context switching. I don't know whether that impacts stutter or framerate though (or both).

 

[jaymc]

All fair points I guess...so you reckon it's a placebo effect.. maybe it is..

Still I'd like see some numbers on mouse response times. Maybe it is microseconds, that reviewer I linked reckons up to 20 milliseconds I doubt that myself... But say it is 10 or 20 milli seconds in an intensely fast fps like quake would be noticeable..would it not ?
Droped frames as well... there's definitely stutter going on in gta with 7700k as commented on by steve in Techspot. linked above.
An more cores of coarse maybe they only notice this when compared to the 4 core kaby lake.

I'll bet well all be figuring out which usb port it is that goes direct anyhow..

 

[-Fran-]

Even if this was the case, the difference in latency would be measured in microseconds at worst. Humans struggle to detect latency measured in the low millisecond range. Most USB devices have a polling rate of no greater than 1000Hz because of this.

There's a lot mixed in how a human can react, but moving images to the brain is stupid fast. It's in the 15ms range (Light -> electrical charge -> reach brain zone). Recognizing an image is around 100ms. Acting on that image opens a bit the range, but 250-400ms is the common range. I am proud to say I'm in the lower 250ms. In fact, I have a best measured response time of 150ms without prediction. With prediction I have done 50ms. My average is less than 250ms when I'm fully focused.

Point is? You move your arm and depending on how fast your brain is, it will expect a response and then correlate that to the image on screen. Those 2 actions, for the brain, has a *very* low delay, since no "processing" is done and it's just signal traveling, hence, when you move your arm (and mouse) your brain will expect *something* happening immediately in conjunction with your eyes when you're trained in FPS'es. So yeah, I have never ever been able to get near VSync for that reason; the delay from the mouse input, I feel it a 100% and it's so irritating.

This is a great read on the subject, although the times it exposes are different from the ones I remember; in fact, in this they put faster times! And also explains the "prediction" part.

https://www.salk.edu/news-release/we-live-in-the-past-and-our-brain-makes-up-for-it/

Nope, that's perfectly normal. It's why we have groupthink and pluralistic ignorance. Which is not to say that this is a case of either, but that consensus is not as reliable as you might think, and certainly not unusual.

If it exists we can probably measure it. So someone should just measure it. Or at the very least they should do a double blind test to remove confirmation bias.

I agree 100% there. What you can't measure is in the real of the Gods, as I say. Everything else can be measured somehow.

More cores certainly helps with reducing thread context switching. I don't know whether that impacts stutter or framerate though (or both).

I have the same intuition there. The less the thread has to bounce, the less time the CPU will spend moving resources around and be able to actually produce frames.

Cheers!

 

[Rookie_MIB]

I have the same intuition there. The less the thread has to bounce, the less time the CPU will spend moving resources around and be able to actually produce frames.

Cheers!

That's kind of what I'm thinking too. A DX12 heavy game with a bunch of threads going on an Intel quad-core with HT is going to be loaded more than the Ryzen 8-core with SMT. I think the Ryzen probably would show much fewer spikes than the Intel into the higher percentages where you might just notice a minor stutter here and there. There's no doubt that Ryzen 8-core has just more raw horsepower to plow through workloads without bogging down, and while Intel might have a higher average throughput with frame rates, it's those little spikes that you see.

Think of it much like the latency issues that AMD had with their GPUs when compared with the 900 series Nvidia GPUs. The frame rate wasn't bad, but there were the little jerks when the latency spiked for a frame or two. Nvidia's latency was very consistent - so you didn't see any twitches. Those little twitches catch your attention.

 

[jaymc]

Interesting stuff guy's.. Blind test is great idea actually.. could be done quite easily I suppose...

Good article Yuka...

So yis don't think cutting out the chipset could drop 20ms then ?... So maybe it is jus when compared to 7700k then.

Anyway drained I surrender. Tomorrow hopefully shed some more light on the situation...Hopefully some numbers too eh...

Thanks,
Jay

 

[8350rocks]

Surely someone could just capture gaming sessions on one rig, then another and show the 2 side by side running the same missions to accomplish an objective comparison. Stutter would be obvious when it happened on one side or the other.

It may be the only way to capture the data necessary to gather the burden of proof.

 

[randomizer]

It isn't objective if you already know which system is which. Even then it's not exactly a clean comparison since the two platforms are totally different and each has a range of hardware performance tiers.

 

[8350rocks]

New BIOS updates increase Ryzen gaming performance significantly (25%+): https://www.youtube.com/watch?v=HX_WEvEzR64

 

[Nope 1151]

https://smallformfactor.net/forum/threads/msi-unveils-the-b350m-mortar-arctic.1788/
Another ITX Ryzen board spotted.

 

[con635]

Interesting stuff guy's.. Blind test is great idea actually.. could be done quite easily I suppose...

Good article Yuka...

So yis don't think cutting out the chipset could drop 20ms then ?... So maybe it is jus when compared to 7700k then.

Anyway drained I surrender. Tomorrow hopefully shed some more light on the situation...Hopefully some numbers too eh...

Thanks,
Jay

Pascal had a dpc latency problem in drivers last year, I wonder is it still there and the amd platform 'hides' it? Just a thought...could be tested with and shown with numbers as well...

 

[juanrga]

The SoC design is not playing any role here. The smoothness that some reviews are noticing follows either from subjective bias or from they testing quad-core vs eight-core and then having a dozen of background apps open: three chat sessions, encoding video, playing the game, writing the report, looking for OS upgrades,...

eight-core vs eight-core we find something like this

RyZen gameplay is smooth on Fallout4, then the same or less smooth in rest of games.

 

[aldaia]

EDIT: Most sites are reporting the gains measured by the internal bench tool included in the game. Pcgameshardware has used an external tool to measure the real performance gain and got a very different picture

https://translate.google.com/translate?sl=auto&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.pcgameshardware.de%2FRyzen-7-1800X-CPU-265804%2FSpecials%2FAMD-AotS-Patch-Test-Benchmark-1224503%2F&edit-text=&act=url

This results where signifficantly flawed and they have updated their findings several times, so I wonder what are you refering to by "very diferent picture":
1) To the initial finding that Ryzen 1800X was doing much worse that i7 7700K
2) To update 1 where a 6900k was doing even worse than 1800X
3) To update 2 where a simulated 1500X was almost on par with i7 7700K and signifficantly outperforming both 1800X and 6900K

The reason for the bad performance of the 8 cores in real savegame was "AotS lowers the picture quality on CPUs with less than six physical cores in the regular game - not in the integrated benchmark - without the player's intervention. As it appears, on Quadcores (even with SMT), ergo eight threads, whole particle systems are omitted."

 

[jaymc]

Surely someone could just capture gaming sessions on one rig, then another and show the 2 side by side running the same missions to accomplish an objective comparison. Stutter would be obvious when it happened on one side or the other.

It may be the only way to capture the data necessary to gather the burden of proof.

New BIOS updates increase Ryzen gaming performance significantly (25%+): https://www.youtube.com/watch?v=HX_WEvEzR64

Excellent stuff.. be interesting to see the numbers with AMD GPU's.. and there is another BIOS update due before Ryzen 5 release on the 11th...

Nice post Juan.. looks like their comparing apples to bananas alright.

 

[juanrga]

EDIT: Most sites are reporting the gains measured by the internal bench tool included in the game. Pcgameshardware has used an external tool to measure the real performance gain and got a very different picture

https://translate.google.com/translate?sl=auto&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.pcgameshardware.de%2FRyzen-7-1800X-CPU-265804%2FSpecials%2FAMD-AotS-Patch-Test-Benchmark-1224503%2F&edit-text=&act=url

This results where signifficantly flawed and they have updated their findings several times, so I wonder what are you refering to by "very diferent picture":
1) To the initial finding that Ryzen 1800X was doing much worse that i7 7700K
2) To update 1 where a 6900k was doing even worse than 1800X
3) To update 2 where a simulated 1500X was almost on par with i7 7700K and signifficantly outperforming both 1800X and 6900K

The reason for the bad performance of the 8 cores in real savegame was "AotS lowers the picture quality on CPUs with less than six physical cores in the regular game - not in the integrated benchmark - without the player's intervention. As it appears, on Quadcores (even with SMT), ergo eight threads, whole particle systems are omitted."

I wrote my post before the updates.

 

[jaymc]

Early AMD Ryzen 5 1400 review and comparison leaks:

http://hexus.net/tech/news/cpu/104167-early-amd-ryzen-5-1400-review-comparison-leaks/

 

[8350rocks]

Differences between GPU approaches to DX11 and DX12, and why it makes the benchmarks for Ryzen off a bit: https://na2.salesforce.com/0064000000o8B9V

(Hint: it is the drivers/hardware...)

 

[jaymc]

Differences between GPU approaches to DX11 and DX12, and why it makes the benchmarks for Ryzen off a bit: https://na2.salesforce.com/0064000000o8B9V

(Hint: it is the drivers/hardware...)

Think link is dead or article taken down ?

I found something on Nvidia forum same topic.. Gonna post it in Vega thread. It looks like Pascal is in trouble indeed when it comes to Dx12.

j

 

[dgothi]

I notice some Ryzen new motherboard's DOCP is not always ready for some RAM memory. I assume they are working on it and release a new BIOS version soon.

 

[010TheMaster010]

@Juan what would lead you to believe all these given reviewers are "having a dozen of background apps open: three chat sessions, encoding video, playing the game, writing the report, looking for OS upgrades,..."
I can't think to believe all would be doing that. Furthermore if that were true would that be bad? I often have background apps open while I game and I know others who do even more than I, so if this were true and all these reviewers have all these things going on at once (a stretch) it would be another selling point for R7 processors, no? I believe this very comparison was made by AMD with a 1700(X?) Vs a 6700k at one of their earlier events.
The problem with comparing 8c vs 8c in your screenshots is price. For 30 some dollars over 7700k with cooler you get an overclocked 1700 with comparible max frames, lower average, but smoother gameplay, plus the other advantages of the 16 threads.
Thoughts/criticism welcome, cheers all

 

[8350rocks]

https://www.reddit.com/r/Amd/comments/638drc/amd_ryzen_versus_intel_kaby_lake_in_one_potato/

Interesting graph (hand drawn) of results between 7700K and 1800X over the course of 16 games showing the actual frame times plotted against frequency and actual frame time gap. (y-axis is frequency, and X-axis is frame time graph).

 

[-Fran-]

@Juan what would lead you to believe all these given reviewers are "having a dozen of background apps open: three chat sessions, encoding video, playing the game, writing the report, looking for OS upgrades,..."

I also add myself to this question. That is clearly Juan making a hyperbole, but the matter of the fact still remains: some people just likes to play games with a bazillion other stuff running in the background (yes, another hyperbole).

Some more self-reference, which I believe is not far fetched to generalize as an opinion: Steam+Overlay, Chrome (2 or 3 tabs; I know people with 20+ and multiple instances), Winamp, Discord/Teamspeak/Skype and sometimes OBS when I feel like streaming how much I suck at games (kidding, I'm a super prah playah).

EDIT: That being said, my i7 2700K does not seem to be bothered with that load.

Cheers!