AMD Zen Discussion Thread

Icyseanfitz

EoinHef wrote: »

I never thought it was on par,i just didnt expect it to be as far behind as it is. Id say thats the same for a lot of people.

Just in gaming of course,as thats the only aspect thats important for me.

yeah same, wasnt expecting it to beat out the likes of the 7700k but i was expecting it to be close like intels x99 chipsets are, pity as i would have swapped out my 6700k just to support amd if it was

Redfox25

Looking at it another way, if its as good or better than the equivalent Intel chip at a cheaper cost for commercial users , they will likely make a butt ton more money from this market as I am willing to bet its bigger than the gaming market.
As they need cash and sales, aiming it at the commercial/server market seems like a good play.

It can only lead to competition in the cpu market which will suit us at some stage and if they do sort out the overclocking headroom issues and driver issue (if it is a fault) then it will be a decent chip eventually.

Wish they were better at keeping closer to the truth with the hype train.

With Vega not too far off they should take a lessons learned from this.

K.O.Kiki

smacl wrote: »

Given the jump in their share price over the last 18 months I'd say cynical maybe in terms of hype but not stupid. From a commercial standpoint, they needed a product in the market, and while it is not everything a high end gamer might want at this point in time, it will certainly be a very competitive product across the broader market. As for the gaming market, a large part of that is still consoles where the consumer is the console manufacturer rather than the gamer. It will be interesting to see what CPUs the next generation of consoles use.

Xbox Scorpio is already scheduled for another AMD APU.

BloodBath

smacl wrote: »

Given the jump in their share price over the last 18 months I'd say cynical maybe in terms of hype but not stupid. From a commercial standpoint, they needed a product in the market, and while it is not everything a high end gamer might want at this point in time, it will certainly be a very competitive product across the broader market. As for the gaming market, a large part of that is still consoles where the consumer is the console manufacturer rather than the gamer. It will be interesting to see what CPUs the next generation of consoles use.

Scorpio is using a Ryzen + Vega 11 combo.

Looking at more of the reviews and benchmarks it's actually not that far off in gaming with the right settings. 10-15% in most cases. For the majority of people there will be 0 difference unless you have a a 144Hz monitor and a graphics card that can drive that fps in newer titles.

The 8/16 was never going to be a great gaming cpu especially when it turned out they don't overclock well at all. Hopefully they can drive the clock speeds of the 6/12's and the 4/8's a bit higher.

smacl

L wrote: »

Might actually be good for gamers anyhow if it puts pressure on Intel to adjust the prices of their lineup to compete.

Longer term I think it will be great for gamers as we've already hit the wall in terms of return on investment for performance improvements on a single thread. As we've seen with GPUs, increased performance is best realised with increased numbers of concurrent threads or fibers and is entirely scalable on that basis. If Ryzen gives those software developers looking for high performance the solid kick up the arse they need to spend time understanding how to write robust concurrent algorithms it is already a success in my book.

K.O.Kiki

Really interesting tech analysis:
https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

K.O.Kiki

Neat little graph.

deceit

The above graph gives me more confidence in the opinion that once optimizations are put in place from game dev's the cpu should perform better than the 6900k.
It would never come close to the 7700k in some titles though unless you load obs or something like that at the same time.
If its performance in games improves to a similar scale to how it beats the 6900k in other tasks, their will be a huge shift in prices from intel to compete.
The intel has better overclocking so it would never come down to the same price but heres hoping it gets close as I want both

K.O.Kiki

https://www.computerbase.de/2017-03/amd-ryzen-1800x-1700x-1700-test/4/#diagramm-performancerating-fps-speicherskalierung

Looks like DDR4-2666 is the minimum for good performance.

Samuel T. Cogley

Very disappointed in the 1800X and 1700X. OC'd 1700's look okay for a poor-mans 6900K and hopefully the lower core counts in the R5 are going to give us some higher clock speeds. I'm not holding my breath though.

smacl

Samuel T. Cogley wrote: »

Very disappointed in the 1800X and 1700X. OC'd 1700's look okay for a poor-mans 6900K and hopefully the lower core counts in the R5 are going to give us some higher clock speeds. I'm not holding my breath though.

Longer term though, overclocking on the basis of improved single core performance is ultimately a dead end as marginal gains achieved will never scale the same way as effective use of multiple cores. This hasn't been an issue for gaming in the past as even a CPU with 8 concurrent threads was expensive enough not to be a commercially viable platform for mainstream game development. If we see proliferation of CPUs with 16 thread capability a single threaded application is using just 6.25% of the available computing power. Even on 8 thread CPUs, single threaded applications use just 12.5% of the computing power available. Overclocking might buy you 10%-20% increases on a good day at significantly increased expense. Proper parallelization of a single threaded app will get you 100%-1600% on the same device. When you consider the budget for a high end game these days is in the million, this is a nut that will be cracked. The fact that many workstation apps developed on a tiny fraction of this budget shows it is very doable*.

(*I own and run a small software house doing just this)

Samuel T. Cogley

smacl wrote: »

Longer term though, overclocking on the basis of improved single core performance is ultimately a dead end as marginal gains achieved will never scale the same way as effective use of multiple cores. This hasn't been an issue for gaming in the past as even a CPU with 8 concurrent threads was expensive enough not to be a commercially viable platform for mainstream game development. If we see proliferation of CPUs with 16 thread capability a single threaded application is using just 6.25% of the available computing power. Even on 8 thread CPUs, single threaded applications use just 12.5% of the computing power available. Overclocking might buy you 10%-20% increases on a good day at significantly increased expense. Proper parallelization of a single threaded app will get you 100%-1600% on the same device. When you consider the budget for a high end game these days is in the million, this is a nut that will be cracked. The fact that many workstation apps developed on a tiny fraction of this budget shows it is very doable*.

(*I own and run a small software house doing just this)

All very well but is it effecting the frame rates of the games I play today or even this year. If no then I'll hold off and ultimately that's the issue. It also doesn't really change the fact that the 1700X and 1800X are borderline pointless IMHO. The 1700 shows promise. I'm also very disappointed with XFR, this had the potential to be a killer feature. 100Mhz pffft.

smacl

Samuel T. Cogley wrote: »

All very well but is it effecting the frame rates of the games I play today or even this year

No, but nor was it designed to. Will it affect games written this year? Maybe, maybe not. Will it affect games written next year? Probably. So the question then becomes how much money do you want to spend now on a solution that isn't likely to see significant advancements in coming years when other solutions flourish? Even on 8 threads, spending serious time and effort optimising the performance of 1 thread while 7 lie idle in order to better support legacy code which is due a rewrite seems pointless.

Cuddlesworth

Samuel T. Cogley wrote: »

All very well but is it effecting the frame rates of the games I play today or even this year. If no then I'll hold off and ultimately that's the issue. It also doesn't really change the fact that the 1700X and 1800X are borderline pointless IMHO. The 1700 shows promise. I'm also very disappointed with XFR, this had the potential to be a killer feature. 100Mhz pffft.

I'd pick up a 1700x in a heartbeat if I hadn't bought a 5820 last year. I don't have any plans to game at super high 1080p framerates and the additional threads over a quad core would make a huge difference to virtualisation.

Also, XFR gets 4.1* on chips that can't OC well in most cases past 3.9 is pretty good. Its a single core overclock for single threaded processes.

Samuel T. Cogley

Cuddlesworth wrote: »

I'd pick up a 1700x in a heartbeat if I hadn't bought a 5820 last year. I don't have any plans to game at super high 1080p framerates and the additional threads over a quad core would make a huge difference to virtualisation.

Also, XFR gets 4.1* on chips that can't OC well in most cases past 3.9 is pretty good. Its a single core overclock for single threaded processes.

It's early days but I see no real benefit of the X range. The bit of OCing I've seen saw the 1700 @ 3.9Ghz sitting at a lower temp. Personally I think AMD are just releasing SKUs for the sake of it. If the core speed is that much of an issue surely a 5Ghz 7700K makes more sense.

K.O.Kiki

Samuel T. Cogley wrote: »

It's early days but I see no real benefit of the X range. The bit of OCing I've seen saw the 1700 @ 3.9Ghz sitting at a lower temp. Personally I think AMD are just releasing SKUs for the sake of it. If the core speed is that much of an issue surely a 5Ghz 7700K makes more sense.

Pity the R5-1600/1600X isn't out yet.

deceit

I wonder when the windows scheduler is updated will this greatly improve Ryzen game performance based on this.
https://www.reddit.com/r/Amd/comments/5xkghp/confirmed_windows_10_scheduler_is_gimping_ryzen/?st=izy50749&sh=9a1defc9

Samuel T. Cogley

K.O.Kiki wrote: »

Pity the R5-1600/1600X isn't out yet.

I'm really hoping these will have a bigger frequency range and I think people out there are forgetting these are coming. If the 6900K was released with this much fanfaire against the 7700K the conclusion would have been the same - don't buy it for gaming. But then it's not meant for it!

I have to say though that either way it's good to see AMD back even if it wasn't quite what we expected, but when is it ever?

BloodBath

Samuel T. Cogley wrote: »

All very well but is it effecting the frame rates of the games I play today or even this year. If no then I'll hold off and ultimately that's the issue. It also doesn't really change the fact that the 1700X and 1800X are borderline pointless IMHO. The 1700 shows promise. I'm also very disappointed with XFR, this had the potential to be a killer feature. 100Mhz pffft.

In fairness I think people are being overly negative. Yes in certain benchmarks at certain settings it trails behind. Are you really buying a €360-560 cpu to play games at 1080p though? In most real world scenarios it's very close to the 7700k in gaming even with the likes of a 1080ti at 1440p/144Hz. If you stream and/or record I'd say it comes into it's own.

If you do any kind of work that needs the horsepower it's a no brainer. There's a good chance the windows scheduler bug is causing issues as well. AMD should have got all of the issues sorted before launch as the negative press on the gaming side is hurting their share value atm despite Ryzen selling out. It will bounce back up.

The 1700 is a real bargain. The 1600x is going to be around €300. The 1700 is only €360 and overclocks to around 3.9 - 4.0Ghz. The 1600x is going to need to overclock to at least 4.3-4.4Ghz or why bother. Just get a 1700. Although the 1500 will be only around €260 and might be the best overall value.

Samuel T. Cogley

BloodBath wrote: »

In fairness I think people are being overly negative. Yes in certain benchmarks at certain settings it trails behind. Are you really buying a €360-560 cpu to play games at 1080p though?

Nope just at the bottom end of that and you're buying a 7700K, or I will/would be. The point being made by the like of Gamersnexus hold true as well, the reason they're bottlenecking the CPU is because we're likely to see iterations of GPUs before we upgrade CPUs. By that time Ryzen will be bottle necking 1440p.

The further point made was that Ryzen is great, but in very limited circumstances. Those circumstances are where the job can't be GPU accelerated. For the average Joe like me that mostly games, runs the odd VM and might occasionally throw up a dash cam clip Intel is still the better buy, even now. If Intel decide to do a price reduction it will be a no brainier.

Agreed on the 1700 for certain work loads and indeed one would hope the 1600 will have better clock speeds, but I doubtful. If it did it would make it a great best of both worlds solution.

Samuel T. Cogley

Very well put. I don't agree with all of it (mainly his time scales).

Venom

Samuel T. Cogley wrote: »

Nope just at the bottom end of that and you're buying a 7700K, or I will/would be. The point being made by the like of Gamersnexus hold true as well, the reason they're bottlenecking the CPU is because we're likely to see iterations of GPUs before we upgrade CPUs. By that time Ryzen will be bottle necking 1440p.

The further point made was that Ryzen is great, but in very limited circumstances. Those circumstances are where the job can't be GPU accelerated. For the average Joe like me that mostly games, runs the odd VM and might occasionally throw up a dash cam clip Intel is still the better buy, even now. If Intel decide to do a price reduction it will be a no brainier.

Agreed on the 1700 for certain work loads and indeed one would hope the 1600 will have better clock speeds, but I doubtful. If it did it would make it a great best of both worlds solution.

If (big big IF) Vega and Volta are super powerful next gen GPU's, I see Ryzen trailing behind a 7700k at 1440p and 4k just like at 1080p. People need to wise up to the fact 1080p and medium settings is the best way to test a CPU in gaming due the GPU not being a bottleneck.

I'm holding off for the moment until the R5's hit but realistically I can see myself picking up an i5 7600k for my next upgrade.

BloodBath

Good informative video. I always felt these ultra high fps tests at low resolutions weren't necessarily valid. Especially in a market that is seeing a shift towards better multi thread optimization.

The battlefield 1 benchmark is an eye opener. A 7700k at 4.9Ghz is almost being maxed with current gen cards. If you're buying a high end cpu that you might use for the next 4-5 years is a quad core even with multi threading going to cut it over the next 5 years. I don't think so.

I'm still going to get a 1500, 1600x or 1700. Just gonna wait for the 1500 and 1600x reviews before deciding.

Venom

BloodBath wrote: »

If you're buying a high end cpu that you might use for the next 4-5 years is a quad core even with multi threading going to cut it over the next 5 years. I don't think so.

Except PC gamers have had quad core multi treaded CPU's for the best part of a decade and game developers have not bothered to utilize them fully. Even if Ryzen were the catalyst for developers to start making these changes in game design, it will be years before the first of these titles hit the market.

BloodBath

Venom wrote: »

Except PC gamers have had quad core multi treaded CPU's for the best part of a decade and game developers have not bothered to utilize them fully. Even if Ryzen were the catalyst for developers to start making these changes in game design, it will be years before the first of these titles hit the market.

Have you watched the above video? BF1 is being maxed on all cores/threads on a 7700k with a titan. That's great optimization and it's using all 8 threads now. It's also using all 16 on the 1800x.

Expect the big game engines like Unreal and others to follow suite if they haven't already. The shift is already happening.

Also the average gaming PC still has 4 or less cores/threads. That's why the shift has been slow. With cheap 8+ being common this year I'd expect a big shift to 8+ and with Vulkan and DX12 seeing more titles support I don't think we're going to have to wait long.

smacl

Samuel T. Cogley wrote: »

Very well put. I don't agree with all of it (mainly his time scales).

Great video and thanks for sharing. I'd tend to agree on time scales and add a year, which is typical slippage on new development. I'd also tend to think he's a bit off on it being anything to do with AMD to optimise old games, the ball's clearly in the developers court as to whether they consider there's worthwhile return on investment in reworking and re-releasing old games. Seems to have been the case for Sony with PS3 games going to PS4, but that was really to bolster their new platform.

Slutmonkey57b

The argument really boils down to "how lazy are game developers?". AMD is giving them 8 cores on both next gen consoles and pc's on effectively the same architecture. MS is giving them as much of the same development platform between xbox and pc as possible. There's really no excuse for not having good multi-threaded code at this point, as another poster pointed out you are effectively throwing away most of the machine's available power if you don't do it.

The answer, unfortunately, is probably "very lazy".

smacl

Slutmonkey57b wrote: »

The answer, unfortunately, is probably "very lazy".

I'd say its a bit more complex than this, in that multi-threaded is more time consuming to develop, much more time consuming to debug and and a pretty reliable source of serious bugs in late beta releases. e.g. I started investigating issues surrounding multi-threaded debugging about six years ago, as described here and it took another two years of quite intensive work to get my head around it. Most game developers don't have that luxury, and the companies they're working for are more concerned about getting the current release out the door in saleable* condition with the minimum time and money spent. That said, all the better engines now support multi-threading, e.g. Unreal so I would expect most new games based on the larger engines to utilise multi-threading reasonably effectively. The use of post release bug fixing / patching also means that the release date for a game isn't the line in the sand it used to be, such that taking development risks isn't as expensive as it once was.

What I think (hope!) we're seeing is a return to a time where we'll see very large year on year performance increases for PCs as the major bottleneck of single CPU thread is removed.


(as opposed to bug free )

Venom

Slutmonkey57b wrote: »

The argument really boils down to "how lazy are game developers?". AMD is giving them 8 cores on both next gen consoles and pc's on effectively the same architecture. MS is giving them as much of the same development platform between xbox and pc as possible. There's really no excuse for not having good multi-threaded code at this point, as another poster pointed out you are effectively throwing away most of the machine's available power if you don't do it.

The answer, unfortunately, is probably "very lazy".

I think game devs being able to get away with their games just having to run at 720-1080p at 30fps max has a good deal to do with it as well seeing most games are developed for console peasants.

If the Scorpio is as powerful as rumors claim we might see a change in this mentality and benefit as PC gamers.

BloodBath

Apparently we're going to be getting updated Ryzens early next year with another update the year after that all on AM4. Probably more refined with higher clocks. Then in 3-4 years we'll be getting 7nm cpu's. Probably 16/32's in the desktop space. Another whopping 100% die shrink and the last great silicon shrink. They are going to struggle to go much lower than 7. Total diminishing returns past that point.



http://wccftech.com/amd-pinnacle-ridge-cpu-zen-2-core/