Planning a new system

Oh, I would stick with it, if the hardware would not start to signal me that it is seven years old already. It started to get hickups and occasional problems, and the mainboard has become touchy, kind of. PSU also may have hickups. I would stick with it, it is Windows 7 (and Linux dual boot). But the lifespan of the hardware may be against giving me another seven years.

Also, I am curious to try out VR.

Going W10 still makes me howling like a wolf. The new rig will serve exclusively as a game launcher, with no other stuff on it. For all "other stuff" I have Linux: better security, better privacy, better stability, better speed.

 

"First, almost none of the games/sims I am interested in, seem to support HT, most do not even benefit from more than 1-2 cores."
I agree. Currently two cores are the most common computing setup on Steam. Depending on what games you play, that makes total sense. As time passes users keep upgrading from 2-4 cores, and now from 2/4 threads to 8 or 12. That will continue. Intel's *bottom* product is going to be 4 cores from now on, and they will have a well-priced 12 thread product. AMD's highest selling product (and the highest selling CPU in the market) is 12 threads. The mountain is moving.
As that happens, game designers will code for higher core/thread counts, but that takes time. It's very interesting that Witcher 3 performs 15% better (when not bottle-necked) on 8 threads HT over 4, not because that 15% is key, but because of what it points to. At a point where only 1-2% of the market has 8 threads, the game designers have bothered to code in the basis of a pretty sizeable upgrade in performance. That points to potential.
So, yes, AC1 may only get a small kick from a high thread count, but I would really expect AC2 to respond to it, and the vast majority of >2018 games to do likewise.

Intuitively, people have made a fuss about how games could never cope with more than one core, or more than two, and now more than four threads. The hurdle gets passed whenever the technology is plentiful, and we are on the edge of that now. I don't see any problem that >4 cores that wasn't already passed in the move from 2 to 4 (which the vast majority of current games have already crossed).
"AC seems to like multi-CPUs, obviously. But most what I googled said that people exprience a small loss of frames when not switching of HT. Several videos at youtube also illustrate that a quadcore with HT may show lets say 90 fps, whereas the same quadcore with HT off shows only 86 or 87 fps. Just an example."
I don't know where those numbers are from. I'd expect maybe a 30-40% boost in a fully optimised task from HT. In game that is well optimised and not limited by GPU, then maybe 20%. If the limit is GPU (which it sounds like it would be at only 90FPS), then it just becomes a matter of tiny differences in input lag. Whether you have a Ferrari or a Mazda doesn't matter if you are timing them both in peak-hour.

Enjoying the conversation.

 

"In fully HT optimized tasks", you said, well, that a HT-optimized title benefits from having a CPU with HT on, is clear. My point is to what low level only games and sims indeed ARE optimised for HT. Its not as if this technology showed up just last year, its actually quite some years old now. Since many of the Pentium-4s, and Windows supports it since XP.

You said that the market and people just are about to adapt to quadcores (4 CPUs) as a standard, leaving dualcores out form now on. But how long are quadcores around already ? Also since quite some years. If increasing the phcical core number already takes this long time, I wonder how long it will take to have full use of all available HTR become a standard, too. Also, HT is not just doiubling the CPUs there are. Its one and the same worker (CPU) now distrubuting hiw workign time not amongst just one but two assembly lines, so to speak. The beneift is noit a doubling of work being done. He can focus just on one assembly line at a time, its just that when this line stops for a short pause, he can turn and work on the other. What you gain is no doubling of work done. You only save the pauses that now get filled with work.

Therefore I really wonder whether compromising single CPU speed for HT really is worth it for the coming couple of years. Will ther ebe a game anytime soon that becomes unp0layable if the CPU had not a minimum of lets say 8 or 12 or even 16 threads? I do not see that coming in the next 5 years. And if we speak of this demand becoming a new standard all new rleases bease upon, then it will be even more time, I think.

My current CPU is seven years old, and it still can handle all stuff I throw at it. It miught not be the latest, the newest, the most maximised stuff, but still. And it seems even VR is not depending currently on HT, makes no use of it at all, even loses frames with HT on, I read many people on the web reporting.

So when comparing Ryzen and Coffeelake, my focus is set to be on what performance I get from both in single or dual processor use, without HT. And that means Ryzen is loosing quite a big share of the argument in its favour.

Thank God we are already on performance levels anyway these days wehre we tlak luxury problems, and not essential ones. I recall the times when the question was not "80 or 90 fps?", but "17 or 22 frames?", which was a far more decisive difference and had far greater consequences.

I only want reserves for VR, thats why I consider the more expensive CPUs as well, but it seems HT must not be a concern for that consideration. I better save 100 bucks there and on the cooler, and invest it into an even beefier GFX board, me thinks. Maybe even the real big things out there: 1080Ti. Heck, but that price tag hurts. This part of the story is not yet decided...

What I wonder is: if the 1080TI, even more so the OC verison, is known to become extrenely hot in certain small points of the board, why these must be covered nevertehless by a backplate only for haviong a LED graffiti on it. It hinders the air flow, only for cosmetical impression. Very stupid, imo. I would like to see the Asus board without that stupoid backplate. The three vents do not reach these points. In these points, the board gets almost 100° hot, that is madness.

 

I would wait for 8700k and in fact this is my current plan before upgrading CPU. Based on the leaks it should have 7700k level single core performance and R7 1700 level of multi core performance and no CCX induced bottlenecks in gaming.

 

What means CCX bottlenecks ?

 

Ryzen CPUs doesn't seem to scale as well in heavily multithreaded games as Intel counterparts e.g. 6900k (performance wise not price) despite having very similar raw multithreaded performance in synthetic benchmarks.

 

"In fully HT optimized tasks", you said, well, that a HT-optimized title benefits from having a CPU with HT on, is clear. My point is to what low level only games and sims indeed ARE optimised for HT."

I tried to express that with my three tiers of performance.

"You said that the market and people just are about to adapt to quad-cores (4 CPUs) as a standard, leaving dual-cores out form now on. But how long are quad-cores around already ?"

Why have games not pushed CPU's harder over the last five years? Why have we stalled in implementation of HT and so on? Why hasn't Intel rushed to six-cores? All good questions, and I'm sure I only have a small part of the answer. Consoles have played a big part. Most large games need to also run well on XBOX1 & PS4, and these have very weak CPU's (in contrast to the PS3, which had such a powerful CPU that some developers pushed graphics onto it). If you look at most AAA games, CPU compute on Console is the limiting factor. Nobody is going to push the envelope in simulation complexity if the consoles can't keep up. The extraordinary power of a 7700K is thus not relevant in almost all top AAA games, and I can understand why some developers have not spend the time needed to fully utilise high core counts or HT. As I saw, 50% of the market is still on 2 cores, and 75% of the market is on console. Why bother.

That said, if you play Crysis 3, or Witcher 3, or parts of Tomb Raider, you can see some scenes running 60% faster on the Ryzen 1600X over the Intel 7600K. These complex scenes will be the new normal, when the consoles can support it, when developers can increase complexity and still sell.

My argument is predicated on 1) this shifting with the next console generation (which as I said, may well be a 6 core 12 thread Ryzen, or at least a 4/8), and which will normalise that technology. Also 2) longevity for CPU's has increased. CPU's have gone from being 3yr investments (2006, when annual IPC growth was 60%). Then they became 6yr investments (roughly speaking) by 2012, when the 4 core model became a strong incumbent, and when Moore's Law broke down (annual IPC growth fell to 15%, and the 4ghz max was reached). With IPC growth now around 5%, I think of CPU as a decade long investment. Why bother replacing when CPU's improve so slowly?

I do expect the number of cores to grow quickly. The thing that kept us at 4 cores has largely been solved by AMD [infinity fabric]. Both are going to fight it out to provide the most cores at the cheapest level, and I wouldn't be surprised to see 10 core 20 thread i5's before 2030.

And so, yes, I take your point about games from 2012-2015, and smaller games, and console based games, and most simulators etc, being not that crash hot at using HT or high core-counts. It shouldn't be surprising given the market dynamics and the complexity of utilising these things. I suggest that a rational consideration would be trying to look forward, not backward, and that these features will very quickly be normalised, fully-utilised, and key to longevity. In the short term, a 12 thread CPU may be only 20% better in a 2017 AAA game. In the long term having 3 times the thread should mean you have 2.5 times the potential output, and that will be used eventually.

Timing is the issue.

 

6900K doesn't scale due to it not being designed for gaming. This is intentional from Intel. 6900K is a poor guide to high core count performance going forward.

Worth watching for a better guide -

 

As for Ryzen not scaling well. That was true in April. It doesn't seem to be true now. Have you seen Digital Foundry's August set of benchmarks?

 

What I find stunning is that Prepar3d version 3 had HT recommended to be switched off, it did not make use of it, and the lacking optimization for HT caused additional "digital administration" work to push down frames a little bit, so most people reported frames were slightly higher (and temps were cooler) when they had switched off HT. With P3D v4 however, now HT seesm to be suppoprted, the reodcx mmendaiton to switch it off is nill and void for v4. But: what I read in feedback does not convionce me that P3D v4 benefits much from it, it seems to me that most people stay within the framerate window they were in without HT. But temps are higher.

In the end, while a drop of frames from 100 to 60 might be noticable, a difference between 85 and 80 frames is not noticcable at all, at least not to the eye. 20 and 25 frames: very opbvious, 80 and 85: unnoticable. However, temps and ventilators may disagree, and this mo9rning I read quite a bit about tests with the i7 and how their power consummation and temperatures change. There is also a microcode issue with the chipsets of 1151 mainboards, making a difference of over 40W between Z170 and Z270 boards, and spiking temps.

With a CPU like this, becoming so easily so hot, it is a consideration worth to spend some time on, whether HT really is giviong me a performance boost I actually< could ntpoice, while needing to buy it with an additonal spike in temperarzure raise. The hotter, the worse it is for the longevity of the CPU, and other components, the emergence throttling is at around 100°, but the temps should not go above 70-75° with the CPU, under full load. If you already balance at 80° while running a good cooler installation already, you do not want another 5-10° added to that just for HT, even less so if you cannot see a difference.

What you said about the single CPU game market and conjsoles, makes sense. however, there are genres that are not exiostent on consoles, for example realistic flight sims, namely DCS as the obvious exmaple. They never planned to do their stuff for consoles, and I cannot image how their interfaces could ever work on consoles reaosnably. Still, they optimised their titles for single and dual cores only, until today, and without HT.

To me, all this is not a need of performances for games, the games you mentioned löike Witcher and so forth simply do not interest me that much anymore. The only reason why I stray into the high-priced CPU segment, is VR for which you obviously want to have reserves in CPU and GPU. Game sin the future I wait for, are any open world RPG by Bethesda (also loikely to go VR), and maybe Kingdom Come: Deliverance, on which I have set my eyes since years ands silently monitoring it. Then there may be future racing sims (no need to hurry, I am perfectly happy with AC and RR), and maybe flightismming again one day (FSX heavily modded, or P3D), and then there is Steel Beasts Pro 4.0, which has taken a serious blow in frames when upgrading from 3 to 4, and priomises to become even more demandign with their new terrain engine. But that all are sims, not wide-audience mainstream games, and I really wonder to what degree they will bother to seriously push HT to degree where having HT or not makes the difefrence between playabiulity and non-playability, to exagerate it a bit. I do not expect to see this materialising.

P.S. If there would be no interest in VR, I would immediatately drop my aims to something in the cheaper i5 or Ryzen 1600X range, and a GTX1060 card only. Certainly saving lots of coins. I would wish that VR never had come!

I only wish they have gotten back some control over the temps in the Coffee Lake CPUs. That and a competitive price aimed to attack Ryzen, would make it easier to choose. Lower temps and still 30% performance boost? i5 8800! I think I still value single and dual core speed over # of HT, for my needs.

 

Can you provide a link?

I think there was some misunderstanding.
6900k scaled much better in heavily multithreaded games compared to 1800x (at least few months ago).
https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx11-multiplayer-fps
http://www.eurogamer.net/articles/digitalfoundry-2017-amd-ryzen-7-1800x-review

 

I did misread you. Sorry, that's my fault.

I can't find anything that re-benches those toe CPU's against each other.

Richard Leadbetter did the March review of the 1700X for Eurogamer that you linked. He hasn't re-reviewed the same product, but he did wait till August to review the 1600X. In that review he included some 1700 OC benches, that are ahead of 1600X & I5. It's difficult to figure out how that would compare against an I9 now.

Maybe less clear than I had in mind.

The 1700, in its market (against the I7), in gaming, is outmatched. The 1600, against the I5's, for gaming, is superior. I'm not sure that I've done a clear job of representing that.

 

Hey Skybird.

Yeah I can see where you are coming from. Everyone has their own uses, and I can see why high IPC is important to you.

Just wanted to defend the 1600 a bit. I think it's an important development in modern CPU history, and bodes well for future gaming.

I have a Nvidia/Intel combination by the way

 

Thanks for the exchange, Stephen, and forgive my many typos. I am terrible at typing (and too fast), and impatient when typing, but you do not teach an old dog new tricks. If my current hardware allows me the time, I wait for the Coffee Lake and see what it brings. Ryzen is not off the list already, I just question the need to have these many threads.

 

In gaming the CPU isnt that important anyway because in 9/10 modern games the gpu is the limiting factor

In TW3 for example it doesnt matter which cpu you use as you can only get as much fps as the graphics card can produce

 

By changing the number of AI's in a racing game, or units in Total War (if that's what your acronym refers to) you can easily reach CPU limits in both.

A fast CPU raises that limit.

 

Motherboard wise go with whatever board you can afford that still gives you a five year warranty. They are usually imo the better quality boards and it ensures your system warranty wise will be good for 5 years. RAM these days is lifetime, so that's covered if anything goes wrong. CPU's don't die unless you overclock em or do something to damage them. The only thing you really need to worry about is motherboard and video card. There isn't anymore video card guys I can find that offer five years but you can get three where some only offer one. It's just a pre-requisite for me.
I have the i7700K and it's a great CPU for AC if you do end up going with it I think you will be very happy.

 

What i want to say is that spending lots of money on cpus is not useful. Intel's new CPUs have like 15% more performance at the same clock than their 5 year old products for 300% $$$

 

It depends on what games one is planning to play. AMD should work well with most upcomming AAA games.
While niche, indie games or specific cases will run worse due to lack of optimisation (ARMA3, Squad, PUBG). Various older and unoptimised games. If one is not onto these then AMD will be a better choice due to perf/price.
Another note, if one is planning to run 1080p/144op 144hz+ screen with any of such games when AMD might be even worse choice.

However if one is not hurrying to upgrade I would still wait wait for Coffeelake, based on the leaks it has a chance to outclass AMD in single threaded and multithtead games for the price and probably will be very close in raw multithreaded performance. Only problem is that CPUs/mobos might not be available on launch.

 

Thanks for the latest inputs. Not much to answer there.

It is the main reason, maybe, why "by instinct" I am biased against AMD, still. The question is to what degree this old, long-lasting problem still is a valid argument with these new Ryzens?

On a sidenote, I learned last night that P3D accepts onyl credit card payment. Too bad, stupid plastic-debt-making. I am 50 years old and have given up my old Visacard years ago already, it costed me yearly fees and I needed it for some obscure online transactions once in two years. Prepaid cards also cost absurd yearly fees. Too bad, but it seems I have to pass on P3D v4. The benefit: I used to run a heavily modded FSX years ago on my last system, and stable so, with those settings it should be possible to get a stable installation free of problems once again - and being able to still use all those old addons that else would be needed to be upgraded for P3D v4. Tears in my eyes, smile on my lips: on average I should be unmoved then. Well, am I? I am not sure. LOL