AMD CES 2019 Keynote Live Coverage

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

none12345

Distinguished
Apr 27, 2013
431
2
18,785
"Anandtech showed off a CPU shot of an 8 core with the IO Chip and some numbers showing it just under a 9900K in performance but not much was said about if the 8 core is the top end mainstream or if its just the start."

The 8 core zen2 chip wasnt slower, it was faster by a hair, while using ~30% less power.

Using that much less power should allow them to get ~10% faster then a 9900k core to core. That means either higher clocks or higher ipc, likely higher ipc and similar or lower clocks. And thats only an 8 core part, there is clearly room for 16 cores in am4....that would be almost twice as fast as a 9900k in threaded workloads and roughly matching it in single thread.

Cant wait to see the final product. If the 12 core part turns out as good as this looks.....ill be a day one buy.
 

joeblowsmynose

Distinguished


Well that's funny, because my previous quad core on FSX was always at 100% CPU when I played it. Always - it never dropped below that - that means it used all 4 at 100% load, 100% of the time. That's my real world testing.

Ever since FSX SP1 or "Acceleration" updates (like 11 years ago), FSX has been a multi-threaded program. Microsoft even sponsored articles to make this claim ... From Phil Taylor - MS FSX spokesperson back in 2008 ...

"As far as practical limits on number of usable cores, currently SetThreadAffinityMask only allows explicit scheduling of threads on 32 cores (the mask is a dword) on Win32. So that's our effective limit on number of cores. But as soon as there is a way to explicitly schedule them, we can handle 256 cores.""


Check this article with a screenshot of FSX using all 16 threads on an 8 core Ryzen with the user delighted how much "smoother" it was ... https://www.reddit.com/r/Amd/comments/6xujfe/fired_up_fsx_for_the_first_time_since_i_upgraded/

This conforms to my experience of full usage on four cores. (I haven't installed it since upgrading)


X-Plane - also fully multi-threaded and can "utilize as many cores as you throw at it" - according to the devs.


Do you have any other fake news for me to clarify for you? :)
 

InvalidError

Titan
Moderator

If we had 10 years of even competition between AMD and Intel instead of ~10 years of absolute Intel domination, things would likely have been drastically different. ARM can't move up aggressively on core count since it is power-limited by thermals and battery, can't push more clocks and cores than necessary on a 1-3W budget.

As for why AMD would be more aggressive, that's simple: AMD needs the market share and it needs to build it before Intel gets back into a more competitive shape. If AMD fails to acquire market share before that happens, AMD won't have enough sway over developers to get them to optimize for their architecture and Intel will retain its de-facto optimization advantage across most of the board. Pushing Ryzen hard while it can is is the only chance AMD is going to get to dig itself out of that hole.
 

joeblowsmynose

Distinguished


100% correct - it is recommended to turn off core 0 and Intel's HT for best performance in FSX (no idea on AMDs SMT). It certainly does not run nearly as well on one core over say 8 or 16 threads - that is for absolute certainty. Also note that the Multi-core implementation in FSX doesn't increase frame rates per se, but is designed to allow you to crank up the object quality and detail with far less of a performance hit on FPS and to reduce stuttering.


EDIT to add: I guess I did have FSX installed - just ran a little test and core 0 is loaded to 100% with all other threads lightly loaded. Turned off core 0 (left SMT on) and the performance on all other threads improved - with a few also running near 100%.

My conclusion is that high clock is desired for the base FPS on that first core, and more threads are needed to ensure smooth gameplay and transitions with detail settings turned up - I run with most of my details near max. DX11 Steam version. I didn't have my HOTAS set up so my test was very limited as I tried to fiddle with keyboard and mouse ... lol.
 

Specter0420

Distinguished
Apr 8, 2010
111
28
18,710


X-Plane 11 is top notch, laminar is working hard to take advantage of multiple threads too, but it really only uses two so far. People report that their OCed Ryzen 2 chips run fine in 2D but struggle in VR. I know, from a VR Aviation group on FB, that low settings with a 7700K OCed to 4.7Ghz will yield 48 CPU FPS in the Rift. The 1070ti in this system was providing 120ish GPU FPS. X-Plane will give you a metric for CPU frames and GPU frames, divide 1 by this number and you get the FPS coming from each component, it is super useful for finding your bottleneck (and putting AMD fanboys in their place). AMD can't provide acceptable FPS here (45FPS with ASW is the minimum), it is impossible for them even at minimum quality settings with a maximum overclock running the vanilla sim with no add-ons.

If you ever plan on going VR I'd advise you to wait and check VR flight sim specific benchmarks for your desired hardware. Ryzen struggles extra in many flight sims, and it struggles extra in VR. In other words, that 7700K downclocked to 4.3Ghz will stomp all over a Ryzen 2@4.3Ghz in flight sims, especially VR flight sims. I think it has to do with Intel's superior memory bandwidth but I am not sure (the IPCs are too close to explain it). With Gen 2 VR headsets right around the corner, why buy into a platform with a proven history of failing with Gen 1 VR?

I wish we could get some professional review sites to bench flight sims too. I got "forgamers" on YouTube to do some DCS VR stuff for a while by nagging him in the comments but he dropped it from the suite pretty quickly.

I've also been disappointed by the CPU advancements over the last decade. I like to do heavy research and upgrade when there is a performance jump and great OCing potential. I went from the 1st Gen i7-920 to the i7-8086K. Both are ahead of their times and achieve an extra 1.2Ghz OC over base.

VR is the future of flight simming.
 

XMEN_2012

Distinguished
Oct 13, 2012
107
0
18,690


I wish you people invest valuable time in better discussion then proving your point. It won't change anybody's mind nor Intel or AMD going to pay you any sums for that..
 

joeblowsmynose

Distinguished


Now that was decent post, as you didn't try to make any ridiculous claims. I'll give you that most FS (as do most games) do appreciate high clocks to form that base FPS - as I did mention just above, but I never argued against that at all.

I think we're almost on common ground here finally ... ;)

So here's this ... if AMD can exceed 4.5ghz clocks with Zen 2, say maybe up to 5ghz, with an affordable 6 or 8 core product (2/3 the price of an Intel performance equiv), that matched or beat Intel in bottlenecked gaming benchmarks (we all know they are roughly equiv without that artificial bottleneck), and exceeded it elsewhere would that make your arguments against AMD go away ... or would you just find new ones?
 

joeblowsmynose

Distinguished


2 things ...
1) He has yet to prove a point, so far its just been me ...
2) this discussion is apparently, currently about the muti-threadedness of flight sims and not an AMD vs Intel debate ;)
 

XMEN_2012

Distinguished
Oct 13, 2012
107
0
18,690


I really have no idea on that part but I followed both of your comments and points and really see potential to help people on their issues.. I don't see many people using flight sims not so many thread here as well.
 

junglist724

Honorable
Apr 10, 2017
126
38
10,640
xcoy64g.jpg

You can even see the traces reflecting in the light that are exactly the size and shape for another 8 core chiplet.
 
Ryzen 3000 series seems VERY interesting. How brown would Intel's pants be if Lisa turned to the crowd and said "We're clocked lower too."? Hehehe, just a fanboy's fantasy, not suggesting it would actually happen at this point.

But really, core for core, thread for thread, and beating Intel at ANYTHING is pretty darn impressive. Even if it is clocked a little higher that in and of itself is seriously impressive for Ryzen. Honestly, if AMD released that CPU today with that performance we'd finally see AMD match Intel again, after more than a decade of waiting. If they are taking a little bit longer to make it better, we are in for some real magic in the middle of the year. Especially if the rumors are true and this is a Ryzen 5.

To heck with it... this is worth getting on the hype train in ANY way it turns out. CHOO CHOO!
 

joeblowsmynose

Distinguished


Yup -- it wouldn't be offset like that if they didn't have any plans to fill that empty spot that just "happens" to be the exact right size for another 8 cores ... I was pretty doubtful on some of the leaks, but now that I've seen that, I'm starting to get a little skeptical about my skepticism ...
 

XMEN_2012

Distinguished
Oct 13, 2012
107
0
18,690
To all those people still has any doubt on AMD's capabilities should know that -

AMD earns 8% of what Intel does (AMD US$5.33 Bn - Intel US$ 62.76)
They operate at 1% of what Intel does (AMD US$204 Mn - Intel US$ 17.93 Bn)
and workforce is what 8% of what Intel has (AMD 8900 - Intel 107,100)

You should appreciate their effort to put a giant like Intel to run for money and forcing them to disable iGpu so that they can compete with Ryzen.. That's more like David-Goliath story.
 

Specter0420

Distinguished
Apr 8, 2010
111
28
18,710




I am comparing the base clock to the max OC of the high end parts of both lineups. One is the highest clocks guaranteed by the factory for all cores all the time, the other is the highest clocks achieved on all cores via tuning. This is called apples to apples. You should already know this but manufactures will often take the same chip, lower the base clock, and sell it as a cheaper model. It will still, generally, OC just as far as the model it is based on (it may be a lower binning though). This is why I compare the high end. It is like the car salesmen telling you he will give you an even better deal, then you watch as he slaps a new "WAS" price on it, while the "SALE" price stays the same. Not the best analogy, but it works.

You tried to refute this using 2 core max turbo boost vs max all core OC. What kind of deceptive BS is this? All core max OC vs 2 core max boost? Yep, this is apple to apples, IF you are a moron. Can you explain your deception here or is it just ignorance?

All core boost isn't accurate either. Your CPU will run at that IF you have a good enough motherboard, if you have good enough cooling, if running a load with the right instruction sets, if, if, if, if.... But even if we play this stupid game, Intel still overclocks farther from max all core boost, not as much in the newest generation but I am still correct here no matter how you look at it. Not only does Intel go farther from the all core boost, that final number is an entire Ghz faster on Intel AND Intel has the better IPC.

You claimed 99% of people care about overall performance, performance/watt, and price/performance. Then you said people like me, that need IPC and Ghz for their uses, are only 1% and that we are either fanboys or stupid (ignorant). I thoroughly proved you wrong. You just admitted yourself that gaming wins on Intel. So, do you claim that gamers are only 1% of high end PC users and they are stupid and fanboyish? Because that would make YOU very stupid. This false attack by you is the entire reason we are here, besides your blind fanboyism.

You said,"1 ghz OC hasn't applied to Intel for the last two generations"
My i7-8086K... What generation do you think that is? What is 5.2-4.0? How does your foot taste?

You are wrong about x-plane 11, every developer wants to boast their work is good (threaded). Go Google "xplane 11 cores or Ghz" notice how all 10 of the first 10 google results agree with me, using real benchmarks and tests? Honestly, I haven't touched FSX and P3D for years. They are not true simulators in the sense that they use lookup tables instead of blade element theory and have garbage flight models (some add-ons are decent). This probably makes threading easier but, landing in a gusty 30 knot crosswind should require some massive corrections, it doesn't in FSX/P3D.

I've read flight sims are very difficult to thread because everything must happen in order, the current equation requires the answer to the previous and the next equation requires the answer to the current one. You don't need a bunch of smart buddies, you need one fast math genius. If you use lookup tables it is much easier to preload the possible outcomes, there are only so many in the table. Real simulations are harder to thread, IIRC Laminar made some new hires just to work on threading a while back. They say it will be a lot of work for little gain, they even also said it won't change the balance of power, it will just improve performance across the board (between Intel and AMD).

My point all along has been that Intel has more overclocking headroom and a higher IPC. AMD has lower overclocking headroom, lower IPC, and more threads/$. Both pieces of hardware have their own performance advantages, only most of AMD's are trumped by a decent video card. This is just a fact, deal with it.

I've always bought the best chip for my purposes. My first build was an Athlon 64, my second was an Opteron 185. Then I went Intel and they haven't lost the lead since. I am glad to see competition again and hope AMD stomps all over Intel for less money, it will only mean good things. I'll go AMD if they can take the VR Flight Sim and gaming crown when my next build it due. They have a lot of work to do because they can't even do Gen 1 VR Flight Sims today in the sims I care about most, DCS and X-Plane 11.

I like seeing that AMD has Intel huffing and puffing from the factory with the i7-9900K! Hopefully AMD blows us away but I am skeptical. The Ryzen 2 chips, after overclocking, compete with the OCed i5-4690K in X-Plane and DCS World. That Intel chip is from 5 years ago right?
 

InvalidError

Titan
Moderator

The 8086k doesn't run at 4GHz stock, all-core boost at 4.3GHz and 5GHz stock max turbo boost. It will spend a large chunk of its time above 4GHz depending on available power, cooling and turbo settings even without explicit overclocking.

I don't count anything below stock turbo as overclocking since the CPU is basically guaranteed to go at least that far, albeit at the expense of unreasonable power and cooling.
 

Specter0420

Distinguished
Apr 8, 2010
111
28
18,710


Fine, about 15% of 8086K's can reach 5.3 Ghz all core. There is your 1Ghz over-all-core-boost-clock-with-non-AVX-load-even-though-your-overclock-has-no-AVX-offset-overclock.
AKA: 1.3Ghz overclock.
I've been overclocking for about 15 years, you can't redefine the terms for everyone now. Feel free to count whatever you want for yourself though.

VR uses AVX.
JoeBlowsMyNose claimed people that need IPC and Ghz over more slow threads are only 1% and we are ignorant fanboys. He is wrong, I'd argue the great majority need IPC and Ghz over slightly more slow threads.
 

joeblowsmynose

Distinguished


I never said that. What I said was 99% of what people care about was "performance, performance per watt, and performance per dollar, and the rest are either fanboys or ignorant."

Do you think that I meant that IPC and GHZ don't contribute to the "performance" parts of that equation? I don't think I was ignoring the "performance" aspect by only consdering it in all three considerations ... Here you go with your creative interpretations again ... My posts don't need interpretation - what I write is what I mean. Try to read posts more carefully and maybe you won't have to get so worked up. You were talking about overclocking at that point in time ...
 
RYZEN 3rd Gen 8C/16T pitted against i9-9900K 8C/16T and performed on par with it at base clock. OC performance and OC capability of Ryxen Gen 3 is still unknown but if there is that much improvement in IPC and if the CPU is priced at half of what i9-9900K is available for then even without high OC capability it will be great for gaming crowd.

Radeon 7 performance wise is fine with me if offered at a lower price around $500-550 but at $700 it kills off its purpose of competition. Bad move from AMD.
 

Specter0420

Distinguished
Apr 8, 2010
111
28
18,710


Why lie about what you said? You said "overall performance" not "performance" and here is the your line directly in front of that one;
"If all you want is an OC ghz number then buy a damn pentium, lol." It sounded like you are touting overall performance over single core performance to me. Now lying about what you said in a way that hides that, it looks extra suspicious. Maybe I give you too much credit?

Regardless, we appear to agree on the strengths and weakness of each side. You just don't like that Intel can OC farther than Ryzen and Ryzen's advantages are largely nullified by a video card. Again, I'd advise you to deal with it.

If you don't like how OCing is measured take it up with someone else. Your way is... Well maybe if you see it, you'll get my point.

Your way;
Tuned All Core Clock Rate - Two Core Max Boost = OC?
Nope, sorry. Maybe if you are pathetically trying to win an argument you've already lost...

Another way mentioned here;
Tuned All Core Clock Rate - All Core Max Boost = OC?
Nope, this is much better than that pathetic first one, I mean it actually kinda makes sense, but it isn't correct either. You don't get the all core max boost all the time, especially in AVX loads. My OC doesn't use an AVX offset so it runs at 5.0Ghz regardless of the workload. This isn't apples to apples. Overclocks have NOT been measured and compared this way for the last 15 years.

The correct way;
Tuned All Core Clock Rate - Base Clock = OC!
The fastest the CPU can run tuned, no matter what - the fastest the CPU can run stock, no matter what.
You see, that IS apples to apples. See how both sides of the minus sign show a fair comparison? For example, there isn't an AVX offset on one side but not the other... You aren't foolishly comparing 2 cores' temporary clock rate to all cores' permanent clock rate... This is how overclocks have been measured and compared for the last 15 years.

I guess we should just agree to disagree. I really do hope AMD does great, I'd love to see Intel engineers scramble for their brown pants. It will only mean better things for my next AMD or Intel build, whoever suits my needs better.
 

PapaCrazy

Distinguished
Dec 28, 2011
311
95
18,890


Interesting that AMD chips respond to affinity masks the same as Intel chips do, but that would make sense since it's up to Windows scheduler and FSX coding. Thanks for checking. I'm feeling pretty excited about the upcoming Ryzen chips, and unless they are lying about Cinebench scores, it looks to stand close to the i9 in IPC/clock performance which Cinebench prefers. It would be nice to be able to keep Cores 0+1 at 5.0ghz for both OS and dedicated FSX core, but honestly anything close to this performance level would stomp on my 2600k, regardless of a few mHz here of there.

I believe you are correct about the thread loading, 2 and 3 are for textures and scenery (which is why splitting those loads into virtual threads with higher latency creates stutter even without an obvious FPS drop).
 

PapaCrazy

Distinguished
Dec 28, 2011
311
95
18,890


I'm extremely excited about X Plane. I am attached to FSX because of PMDG, weather, an orgy of addons, etc but so very tired of the performance, bugs, and ridiculous costs of P3D which looks to only improve OOMs and graphics, so I won't bite. Community is getting kind of bitter too. What I have seen from X-Plane looks amazing, absolutely amazing, and I think the addon community will develop a bigger following around it. I think I heard Active Sky already made a weather engine. It just lacks textures. Anyway, I'm very excited to delve into a new sim, and shake the cobwebs of FSX off. It's been... too long.

I may get into VR (that'd be a long shot because I get motion sick easily) but plan on waiting for verifiable 3700x benchmarks either way before purchase. I'm no pre-order fool, no matter what. But if AMD delivers, I'm going to reward them. May even throw in more cores for the hell of it, as long as they provide a legacy mode across Ryzen again, I could use extra cores for rendering as long as I can turn them off later.

Maybe I'm wrong about it being a small crowd in flight sim. I've been playing them since the IBM XT and it's amazing where they've gone. Either way, I think the unique case flight sim engines represent could stand as an interesting benchmark indicator.