Discussion AMD Ryzen MegaThread! FAQ and Resources

Page 27 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Ryzen definitely has some improvements that need to be made. Whatever the underlying reason for a loss of performance in games they have to find it and correct it. However, I can remember 8 months, 6 months ago we were all saying that Ryzen would be a huge success if they could equal Haswell upon launch. Most thought that was all but impossible. AMD has more than delivered and in many areas is actually trading blows with Kaby Lake. That is astounding, and no matter what issues Ryzen has at this stage it is a huge success and AMD is definitely back in high end processors.
 
I personally set the bar lower: Sandy. I always said they needed to beat Sandy hands down to be back in the game. Reaching parity was a great thing for them and I still believe that as well. Well, they did for the most part.

Even Intel has trouble beating the Sandy gen across the board consistently in a lot of day-to-day tasks; or beating Sandy well enough to make it worth to justify a platform upgrade.

I still think they can improve in the upcoming months in the gaming area with some patches, but that won't make a huge difference. Their biggest deficiency is not software nor uArch, but good-old Hertz. Ryzen, so far, seems to be held back badly by process and not being able to get near the clocks of the 7700K is going to hurt them in the mid-range arena. I do believe in the low end (i3, i3K and Pentium-HT) they will have a good showing though; that is until Intel decides to kill a few SKUs and move models up the tiers.

All in all, competition is back across the board and I will keep on saying: Zen is a good foundation. I hope AMD takes this opportunity to get some well deserved cash and put it to good use. They still have Vega, the something-something Ridge APU and the 4C and 6C siblings coming. And in the AMA they already confirmed Zen2 and Zen3 even.

Cheers!
 

Nope 1151

Commendable
Feb 8, 2017
70
0
1,630
I think I set the bar lower than all of you guys. I set it anything past my Phenom II x6. Not sure what has happened since, I've been stuck with borrring old pentiums or I3s. :p
 

thegentlewoman

Prominent
Feb 21, 2017
44
0
530
ITX: From reddit:

Quote ---
[–]assovertitstbhfam 78punti 8 ore fa

There were reports that AMD specifically told motherboard manufacturers to delay mini-ITX launch, can you confirm/deny/explain that? Also, are Bristol Ridge APUs ever going to be released to the public?

permalinkembedsalvasegnaladona goldrispondi

[–]AMD_jamesProduct Manager 160punti 8 ore fa

AMD did not ask for any delays on mini-ITX, rather the partners are free to release when their product line is ready.

We will introduce 7th Gen A-series APUs for socket AM4 in the channel later this year.

----- End quote---

Other info: Noctua (austria) has introduced already a specific Heat Sink for Am4 and Intel for reduced ITX solution.
So Austria has already got (on amazon you can see it) NH l9 x65 ... It's the Heat sink for tiny solutions. Well ... I find it interesting.
 
Everyone was disappointed that Kaby Lake didn't offer much besides a modest clock bump, but that decision makes a lot of sense now. I'm not sure how relevant the gaming discussion is for the 8 core Ryzens . If you want gaming performance, you're comparing everything against the 7700k, not the 6900k. Pricing and performance on the 4 and 6 core Ryzens still leaves the door wide open for cost undercutting. If SMT really is a detriment to gaming, clockspeed and then cores are all that matters. As long as it hits the same clocks as the 1800x, it should have almost identical performance in everything except the most highly threaded games. That would give the 6 core Ryzen a lot of room to position itself.
 

salgado18

Distinguished
Feb 12, 2007
966
426
19,370
Question not answered on the reviews: can the turbo and/or XFR be increased independent of base clocks? I mean, like on Piledriver, where we could make a good overall overclock, but a much higher single-core turbo?
 

MaDDD

Reputable
Jan 13, 2016
268
0
4,860
People comparing a high clock speed 4 core that most games are pretty optimised for vs some completely new tech on a lower clock 8 core. It's quite obvious that the higher clock speed CPU will be better in games. Though there does seem to be some issue at the moment with Gaming at resolutions up to 1080P. As you will see in many of the reviews, the gap is closed massively at 4k resolutions especially (Sometimes the Ryzen even supersedes the 4 core Intel chips). As patches come out and more optimisations in drivers come, that gap wil likely be closed even further. If you consider the performance per dollar as well. AMD has definitely done a good job and certainly has something to work with to be competitive against Intel. This is a very good sign as hopefully that will drive better development from both Manufacturers and actually get Intel making proper steps up with their next generations of CPU.
 
I think the *real* issue here is the hype train left the station early, and people ended up blindsided by the gaming benchmarks.

Ryzen is solid for the cost; no one is going to seriously argue otherwise. But we need a few days to dig down and determine what it does well, and what it doesn't.
 

thegentlewoman

Prominent
Feb 21, 2017
44
0
530
I told above how to dig down. We need to see how in 2009 2011\12 until today the new tech from intel has improved its performance without of course counting into the game the effect of better videocards and rams lower voltage and bla bla bla.
Only who will do this kind of objective analysis will understand the potential of Zen 1 ... not considering that in 4 years we have zen 2 .. but really this second thought doesn't even really have to be take into account. The fact is that who will do such an analysis ? who has kept the data of the benchmarks of 9 years ago 8 year ago 7 years ago and has done them every time a new card was coming out a new high requirements videogame was coming out?

Only like that you can not only forecast what is the adaptation of the SW of gaming to this Ryzen tech but much more.
It has also to be said that never we had VR in 2009 2010 .. it was there yes, but just like testing demos. So how much will ryzen will play on that territory (ever developing). I think ... just like in Archaeology you understand: we wanna talk of things we have no words to really describe and most, beside fan boys end up to just create anacronysms with things that they can't even grab or describe, because nobody so far has clearly said here: we don't know. (or few did it).

The rest of the analysis of other users are way far too clear in last 2 days.

I have a really noob question: what does it mean Single channel double rank, 1300mhz rams
what does it mean double channel single rank, double channel double rank and single channel single rank. Reviewers said that special packs of Rams will come out just for AMD (this also need to be taken into account) but besides this, what does it mean? I really tried to read stuff but can't understand what Single channel Single rank 2666 mhz SUPPORT means? I read it supports higher... but whatever...what is the meaning of that lower supporting and specific specifications ram qualities?

or better: How can I see if a 3000mhz or 3200 or 2666 couple or 4 Rams I will buy will be perfectly supported? because if I read specification of Rams there is nothing like single channel or rank etc written.
 


Agreed. Plus, Ryzen just came out, once all the bugs get fixed and software get's optimized for Ryzen, it will start performing very well.
 

NickatNight8320

Commendable
Jan 12, 2017
13
0
1,510

Look at a board that has 4 ram slots. Label them ABCD. If you place 1 strip in slot A, that is single channel single bank. The memory controller has to only keep track of data going down 1 path - so it can do that quickly.

Place a strip in bank B, and that is on the second channel. Now you have dual channel. double the data, more complexity and it slows the speed to handle the extra bandwidth.

Now, you have to pair up the ram. C will be on the same channel as A, and D will be on the same channel as B. Now you have dual channel Dual bank. The memory controller has to keep track of data going to 4 addresses. More difficulty, slower, but now you have 4 times as much memory per transfer.
 

8350rocks

Distinguished


It is a 2 fold problem, the CCX issue with cross threading is supposedly being worked on for a WinX patch.

The second problem is not all MBs were running the newest revision of BIOS.

https://www.youtube.com/user/jerzybakes420

That guy's review has Gigabyte Gaming 5 running newest revision of BIOS, and his RAM ran at 3000 MHz XMP no problem, and his results were also much closer to what I think would be indicative. He also tested a regular 1700 in lots of games as well, and it ran neck and neck with 7700K and 6800K in games like Witcher 3, etc.
 

jdwii

Splendid


Then why does the Intel chip score better?
 

wh3resmycar

Distinguished
it is good that amd is admitting where the fault lies and has an action plan for it. now we have been holding intel liable for the annual 5% generational increase and yet we dont hear intel discussing that. the question now is when will we see the patched results?
 

jdwii

Splendid




"t's not really clear to me what does it need to have a 100 fps in a game or even 150" Simple people who want the best CPU for their video card so it does not bottleneck it by 33%. If your 1080Ti is being used at 60% usage then what the hell was the point of buying a 1080Ti?

I can say straight up that jumping to 120fps is a great experince from 60 and even my 4790K is having a hard time keeping up with some titles. Why on earth would someone buy a expensive GPU and pair it with a CPU that is bottlenecking it and not allowing it to run at 95%+ usage.

More reviews i read about Ryzen the more i think gamernexus put it best I5 gaming performance(I5 no OC) with I7 productivity performance. A non fanboy would easily be able to tell a I5 is the superior choice for gamers and the I7 is an even better choice price/FPS

If one needs the extra cores like bulldozer Amd looks decent for the money if not no

Zen+ needs a better memory controller a better more redefined process node to reach higher frequency's and possibly some minor changes to the core i doubt Intel will have anything major when it comes to improving their performance in the coming years(3+)
 

jdwii

Splendid



That's why its important to see IPC comparisons something i think we really didn't get to see with all the review sites. All processors locked at the same frequency and then test it.

Also optimized for 8 cores wasn't that what people said with bulldozer? I also gave comparisons in dolphin benchmark 1800X clock at 3.6Ghz 4.1Ghz turbo 454 score
6900 Clock at 3.2Ghz 3.6Ghz turbo 384 score

Kind of sad since it scored 18.2% lower despite having a 12.5% clock speed advantage. Watch dogs 2 DOES use 8 cores well and IS optimized for 8 cores and the 1800X loses to a 7700K. Even with the 7700K OC to 5Ghz it loses to the 8 core 6900K meaning yeah the game loves cores.

Trying to be super positive below

Watching Level1techs i can say that perhaps we might see updates results a few weeks from now as there has been constant bios updates that keeps improving performance. High speed ram support might improve with further bios updates this can be the reason why some benchmarks look weak. Also as others have said SMT can be fixed with a hotfix all these things could lead to some gains.

Reality
A decent amount of these games do scale beyond 4 cores and that is why i expect the 6 and 4 core parts will not look any better and might look worse for gaming i once again doubt we will see higher frequency with these upcoming processors.
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


Gamerk is right. This problem has been known since Canard PC got an engineering sample and tested it. Canard found Ryzen to perform like a i5 in games.

http://www.guru3d.com/news-story/french-magazing-posts-engineering-sample-amd-ryzen-processor-benchmarks.html

The problem is not in the engineering sample. Before launch Canard tested a qualification sample for the 1800X and got the same conclusion: bad performance in games. They also claim the problem is not SMT, but the memory controller

https://twitter.com/CPCHardware/status/836346777267761155?p=p

And reviews find the same gaming problem in retail chips, with SMT not being the problem, because disabling it gives little benefit or even reduces gaming performance

http://www.gamersnexus.net/hwreviews/2822-amd-ryzen-r7-1800x-review-premiere-blender-fps-benchmarks/page-7

Gamerk is also right on the relation between AMD and reviewers. From gamersnexus review:

When we approached AMD with these results pre-publication, the company defended its product by suggesting that intentionally creating a GPU bottleneck (read: no longer benchmarking the CPU’s performance) would serve as a great equalizer. AMD asked that we consider 4K benchmarks to more heavily load the GPU, thus reducing workload on the CPU and leveling the playing field. While we fundamentally disagree with this approach to testing, we decided to entertain a mid-step: 1440p, just out of respect for additional numbers driven by potentially realistic use cases. Of course, in some regard, benchmarking CPUs at 4K would be analogous to benchmarking GPUs at 720p: The conclusion would be that every GPU is “the same,” since they’d all choke on the CPU. Same idea here, just the inverse.

The (non-official) information I heard is that solving the memory controller latency problem is in the to-do list for Zen+. AMD tell us a different history in public. AMD says that the problem is the lack of optimization and that software patches will solve the game problem. Time will say...
 

jdwii

Splendid


I think i'm more disappointed that people are saying the same things to justify Ryzen's performance as they did with bulldozer like moar cores for games in the future or patch this or that. It's just the SAME thing all over again like no one learned their lesson from bulldozer.

I guess we can once again wait and see if Gamer,Juan, Jimmy, or anyone else who has been with toms for years and years can reply to this what do you think about hearing this same old crap again and what do you think about Amd telling reviews to review this in a GPU bound situation(4K or 1440P)

Edit that last part is VERY important why was Amd telling people to review these CPU's in gaming benchmarks when they will be GPU bound?

Why is Amd downplaying high frame rate gaming?

These things Intel would never get away with and i sure would be speaking up if they tried like i did with Intel's claims that devil cannon could easily get 5Ghz overclocks on air
 

thegentlewoman

Prominent
Feb 21, 2017
44
0
530
Thanks jdwii for the clear answer, I learn more from you too and from juanrga and all here. Please keep posting your assumptions but let's keep out from definitive tones, we really need a distended atmosphere in these times... especially in this business. Don't you think?::
It's unsettling (not disgusting) that we are bombarded by this tone from the media from the infovortex (which most of us still after 17 years don't grab or had no time to read through: yes it's matrix) and many people do not find, even if they are wise persons and deeply mature and experienced, to reproduce the same tone, sensationalistic, advertorialistic, Breaking news style ... please avoid it.

@MrRobot tv serie: "...maybe it's about Finding good questions to good answers and good answers to good questions..." (last 2 season episode ending).

About divulgation? Is there a web site that tells how much bottlnecking the BUS of Rams, or of a MOBO and or with a video card is happening in a precise configuration? Is there anything like that?
I am referring to this :

" Simple people who want the best CPU for their video card so it does not bottleneck it by 33%. If your 1080Ti is being used at 60% usage then what the hell was the point of buying a 1080Ti? "
...

because I am aware of this and besides the very clear objective we use our MACHINES for or we are used from them for, we must know what is their efficiency and our purchase\expense efficiency I DO Agree but who does this? Just techies. Can we divulgate this knowledge so that ANYBODY can make better choices? Or we wanna keep the fog of war so that it is difficult to understand. Is there some web site that tell this precisely over computer configurations and how the components run together efficiently or no, we have to rely on personal "opinions" and to be so lucky to have people like you as friends?

NickatNight8320 THANK YOU.
So now what does it mean? I need to put just one strip of Ram if I have 2666? So it's just for a 16 gb one stripram that I should look for? What about 2 strip of 3200 mhz am4 won't support them very well ? My mind is so stubborn.
Now that screen of the 1300 1800 2400 2667 is clear or kinda clear. I still can't orientate my choice. Super thanks again for explaining.
 

jdwii

Splendid
thegentlewoman i see you are new to Toms i say welcome to the forums you will see that people here are more based on facts and not so much what people claim is possible with the future which could be wrong or not, many here are from bulldozer or older and are used to hearing such statements like "wait for this or that".

If anything i'd wait one whole month just to see if things improve if not forget it. Many of these games tested with Ryzen use 8 cores and it still loses.
 


That right there really does explain why turning off SMT enables higher performance in many games. In the deep dive reviews which are available, it was mentioned that physical cores and SMT cores when paired together competitively share resources. In addition, a SMT core is not equivalent in processing power to a physical core, especially if that physical core is already in use. So - if the game/os/thread-scheduler cannot tell if a core is a physical core or a logical core, if it assigns a thread to a physical core and an already-in-use logical core, the game performance will certainly suffer.

If turning off SMT disables the logical cores, engaging all of the physical cores - any threads assigned will get the full use of the physical core with no logical core contention for resources. It -is- very much the same thing with the Bulldozer series CPUs, where Windows would assign threads to an already in use half of a module, thus increasing contention. The nice thing was that this was fixable via patch/hotfix, and increased performance for 'free'.