AMD Ryzen 7 1800X (14nm Zen)

[CAT-THE-FIFTH]

Ryzen 2 early next year it seems:

https://translate.google.com/transla...-text=&act=url

I think that is where it will be at TBH.

[CAT-THE-FIFTH]

Another day another potential issue which might have hampered performance at launch:

http://hwbot.org/newsflash/4335_ryze...ect_benchmarks

Edit!!

Just ban Gigabyte boards for using old AGESA and make mainboard tab mandatory for AM4 submissions. Not sure about MSI or ASRock but it's not possible to do this on ASUS.

I seriously think with any AMD launch you need +3 months added to the launch to actually get an accurate picture!!

[watercooled]

OFC you are going to get reviewers of varying competency and bias,but seriously I really think this social media spread "conspiracy" that reviewers are against AMD is just nuts - its a way to quietly attack the media for reviews which are less than 100% positive when AMD has said they need to improve performance,and we remember what happened in the past. Seriously they are different designs,and just like an FX8350 had different strengths and weaknesses,or a Phenom II.

I'm not meaning to imply some sort of conspiracy or journalism integrity issue, it's more of a rant about how results are interpreted, and in some cases presented. Some reviews do have very suggestive wording, e.g. AMD did not 'tell reviews not to use 1080p benchmarks' as a couple of sites are implying and ranting about it, as I understand it they suggested at least running some higher resolution benchmarks for comparison. I still think it was a faulty move on AMD's part as, of course, this sort of attention was almost guaranteed regardless of their intention. And at the end of the day, it's a guide - it's not binding, and for anyone who doesn't know, review guides are the norm for products like CPUs, GPUs, etc.

The strengths and weaknesses thing is true of any CPU, so I think it's a little unfair to compare it to an 8350 (assuming you meant it that way). It's not nearly the same sort of compromised release as Bulldozer. Like I say there is no single 'best CPU' - you have to base judgement on the priorities for a given system. Ryzen can stand alone as a solid processor regardless of price comparisons.

It's like some people are angry and disappointed that Ryzen isn't the fastest CPU across the board, and all for $200. It looks like Zen is a very balanced design and a solid first step on a new series of cores, besting many optimistic predictions and even AMD's own public goals.

I agree with you about Zen2 - Zen had to be finalised at some point to meet deadlines and AMD have IIRC openly stated that some planned features didn't make it into gen1, likely due to time and transistor budget constraints. Having said that, I think they prioritised well with the features they concentrated on. I expect a couple of things may be wider AVX and maybe the clock speed of the on-die CCX (it seems to have the same width as Intel's ring bus but runs at DDR clocks rather than core clocks like Intel's implementation). TBH I can't think of any other obvious areas.

[CAT-THE-FIFTH]

I'm not meaning to imply some sort of conspiracy or journalism integrity issue, it's more of a rant about how results are interpreted, and in some cases presented. Some reviews do have very suggestive wording, e.g. AMD did not 'tell reviews not to use 1080p benchmarks' as a couple of sites are implying and ranting about it, as I understand it they suggested at least running some higher resolution benchmarks for comparison. I still think it was a faulty move on AMD's part as, of course, this sort of attention was almost guaranteed regardless of their intention. And at the end of the day, it's a guide - it's not binding, and for anyone who doesn't know, review guides are the norm for products like CPUs, GPUs, etc.

The strengths and weaknesses thing is true of any CPU, so I think it's a little unfair to compare it to an 8350 (assuming you meant it that way). It's not nearly the same sort of compromised release as Bulldozer. Like I say there is no single 'best CPU' - you have to base judgement on the priorities for a given system. Ryzen can stand alone as a solid processor regardless of price comparisons.

It's like some people are angry and disappointed that Ryzen isn't the fastest CPU across the board, and all for $200. It looks like Zen is a very balanced design and a solid first step on a new series of cores, besting many optimistic predictions and even AMD's own public goals.

I agree with you about Zen2 - Zen had to be finalised at some point to meet deadlines and AMD have IIRC openly stated that some planned features didn't make it into gen1, likely due to time and transistor budget constraints. Having said that, I think they prioritised well with the features they concentrated on. I expect a couple of things may be wider AVX and maybe the clock speed of the on-die CCX (it seems to have the same width as Intel's ring bus but runs at DDR clocks rather than core clocks like Intel's implementation). TBH I can't think of any other obvious areas.

Yes,but even the FX8350 and Phenom II CPUs had strengths and weaknesses - even AMD got to the same level of IPC as Skylake,there will be situations where AMD would trash Intel and other areas where it would lose. If you listen to what David Kanter had to say about the design in this talk it was quite interesting:
https://www.pcper.com/reviews/Proces...w-David-Kanter

But AMD has done themselves zero favours with this launch - even on OcUK forums there are reports of some motherboard models made by Asus bricking themselves and if you look at my previous post on this page,another set of issues have cropped up.

Now I have noticed people trying to blame the motherboard companies exclusively,even there is noise they only had a few weeks to get final BIOSes out.

This is another part of the problem - if reviewers ended up getting the worst of the boards AMD sent out,they will most likely have encountered problems,as opposed to those who had better boards.

You see people framing the more negative reviews which have worse results as them being against AMD - but the issue again is the launch was sloppy.

I have said this for years - AMD KEEPS doing launches like this. Can you remember the last time a brand new AMD CPU or GPU launch for desktop didn't have an issue,that people like us to explain to somebody that it was fixable?? R9 290X anybody?? Now see how Hawaii is probably one of the most solid GPUs designed in the last few years.

The problem is it really undersells their products big time and why even when AMD has graphics cards which are the equal or better than Nvidia they still sell less cards. They are seen as "problematic" on average than competitors and it means the average Joe or Jane goes for Intel or Nvidia. Intel and Nvidia are also far better at hiding problems too.

Zen isn't another Athlon 64 - its more like a Phenom II of sorts(but with better power consumption though) so maybe more like a Phenom II plus launch. I also do keep hearing some of the stuff on other forums about the death of Intel and I think it is premature. Its why AMD getting Ryzen 2 out in a year is good news. In the past they had tended to drag out the time between releases,which could make them fall back a bit and it means Coffee Lake will be most likely fighting Zen 2 not Zen.

[watercooled]

Yes,but even the FX8350 and Phenom II CPUs had strengths and weaknesses - even AMD got to the same level of IPC as Skylake,there will be situations where AMD would trash Intel and other areas where it would lose.

Oh I see what you mean - I suspected I might have been misunderstanding that bit.

Now I have noticed people trying to blame the motherboard companies exclusively,even there is noise they only had a few weeks to get final BIOSes out.

That's probably just fanboys being fanboys. The way I see it, there are a number of problems that could and should have been ironed out by AMD prior to release.

Zen isn't another Athlon 64 - its more like a Phenom II of sorts(but with better power consumption though) so maybe more like a Phenom II plus launch.

Yeah that's one way of looking at it. Zen seems more competitive than Phenom II was at launch though.

[CAT-THE-FIFTH]

Yeah that's one way of looking at it. Zen seems more competitive than Phenom II was at launch though.

Well you can't say its an Athlon 64 though - the Athlon 64 beat Intel in almost any workload including gaming with all its teething problems and Intel had nothing competitive outside an emergency edition Pentium 4,and even then the Athlon X2 came along and dished out more punishment. Its not the original Athlon,since again,it dominated almost every Intel CPU at the time,although Intel did have the P3,but they had more issues getting the clockspeed up:

http://www.tomshardware.co.uk/intel-...eview-230.html

The Athlon and Athlon 64 had chipset issues and that was partially down to Intel apparently putting pressure on companies to favour them for motherboards,so when you take that into consideration its even more amazing what they did. Then the Athlon XP also could compete on all levels and could beat Intel too even in gaming and productivity but things were a bit closer at the time,as the P4 could get close towards the end,especially when AMD couldn't push the clockspeeds high enough.

OTH,the XP-M was one heck of a chip if you could get one.

Ok,maybe not a Phenom II Plus,but its probably more between a Phenom II and a Athlon XP.

[mapesdhs]

I have a Xeon E3 1230 V2/Core i7 3770 which is locked and stays between 3.5GHZ to 3.7GHZ,so if you had a 5GHZ Core i7 2700K you would be doing far better. ...

Ah I see! Yes indeed, an oc'd 2700K is quite a step up, despite IB's IPC advantage.

Note that I do specifically focus on the 2700K. The 2600K is also a good CPU of course for oc'ing, but its potential does vary, hard to be sure what any particular unit can handle. By contrast, every 2700K I've obtained (seven so far) can run at 5GHz with ease, and at a sensible voltage. I'm sure one could push them further (5.2, maybe 5.3?), but I've never tried, 5 is plenty and sits well with the eyeballs on a CPU-Z. Here's my system (been using it for about 5 years now):

http://www.overclock.net/lists/display/view/id/2415471

I plan on upgrading to 4K soon though, for which I'll switch it around with my other system, fit a 1080 Ti and likely fit an SM961 to hold game data (then the 2700K will drive the HDTV, for which a 980 is plenty):

http://www.overclock.net/lists/display/view/id/6211115

... One thing SKL/KL do better is the memory controller - going from 2133MHZ to 3000MHZ RAM can yield upto 20% improvements in performance. ...

Very much depends on the task, but yes some definitely benefit, especially in the prosumer space. After Effects benefits from faster RAM (I observed a 10% slowdown in RT3D renders when changing RAM from 2133 to 1866), and of course lots of it. Friend of mine is working on a project atm which easily gobbles 40GB, hence why I reckon he'd be better off waiting for Naples (the 3930K system I built for him already has 64GB).

A lot of games though, based on site reviews, are less affected by RAM speed, oftem even less so by latency. Amuses me when a RAM kit review commentary says one brand "beats" another for a particular game when the difference isn't even a fraction of the margin of error.

In fact there are one or two hints Ryzen might be that way too - its why I hope in a month or two BIOSes might improve - if that could yield a similar improvement in older games,it might change my perception of things.

Must admit, reading all the forum comments, etc., I'm thinking of leaving my 1700X build for just a little while, perhaps until the end of March, let the BIOS updates roll through somewhat. I'll have enough to do benching a 1080 Ti anyway.

Its these blasted titles running on older engines which are where I see the CPU being pushed.

That was something which always annoyed me about the way Crysis was lauded so much as a benchmark: nobody bothered to ask whether the way it hammered GPUs was perhaps because the engine just wasn't that well written, no matter how good the visuals looked (and I do love the visuals). The speedups obtained with Crysis Warhead suggested this was indeed the case to some extent. Eventually though, my wanting to play at higher custom detail/res pushed the load onto the GPU, so for me CPU power was less relevant. I was pushing the VRAM limit of two 580 3GB SLI due to the custom mod settings I use, high res shadows, very long draw distances, etc. A single 980 worked better, but still only manages 45fps (reeeally high detail! I hate popup in games).

Ian.

[scaryjim]

... The way I see it, there are a number of problems that could and should have been ironed out by AMD prior to release. ....

Whilst there were almost certainly issues that AMD knew about before launch, I think it's worth remembering that some of these issues are things that AMD themselves couldn't necessarily fix.

For instance, they can't rewrite the windows scheduler themselves. Nor can they write fixes to BIOS/UEFI bugs or inconcsistencies. After all, mobo manufacturers have had AM4 specifications for at least 6 months - we've seen OEM AM4 systems at retail last year, and Ryzen ES have been doing the rounds for months. You have to question why those board manufacturers needed to release multiple BIOS updates in the week prior to release...

AMD don't have a huge amount of market share or financial muscle to throw around, sadly. They're a relatively small fish in a relatively big pond, so they're more likely to be dictated to by other companies than vice versa. I'm sure they'll have had hard deadlines for the launch, set for business reasons that we're not privy to (investors are a tricky breed to keep happy ). So it's quite possible that they've been working on these issues for months but have been sandbagged by other companies not giving them high priority.

It could also be the main reason they're holding the Ryzen 5 & 3 launches back - if they knew the Windows 10 scheduler makes Zen's single-threaded performance look bad it would make sense to launch the chips that are meant to be used for monstering multi-threaded tasks first and hold back the lower core-counts for once the scheduler gets fixed...

[mapesdhs]

Well you can't say its an Athlon 64 though - the Athlon 64 beat Intel in almost any workload including gaming with all its teething problems and Intel had nothing competitive outside an emergency edition Pentium 4,and even then the Athlon X2 came along and dished out more punishment. ...

I found RAM speed was a terrible bottleneck for P4s. I had a Dell 650 with two P4 XEONs, but its DDR266 RAM really held it back, it was stomped by an Athlon64 using DDR400. For 3D tasks, as the res/detail went up, the AMD machine could be as much as six times faster:

http://www.sgidepot.co.uk/misc/mysystemsummary.txt

Back then, the only thing I missed about moving off the Dell 650 was losing the hw U320 SCSI RAID for the C-drive, but of course later SSDs came along to save the day.

... if they knew the Windows 10 scheduler makes Zen's single-threaded performance look bad it would make sense to launch the chips that are meant to be used for monstering multi-threaded tasks first and hold back the lower core-counts for once the scheduler gets fixed...

That's a very good point, sounds distinctly plausible. Surprising that tech site commentaries don't mention this.

Ian.

[scaryjim]

... That's a very good point, sounds distinctly plausible. Surprising that tech site commentaries don't mention this. ...

Well, it's wild speculation on my part, so it'd be a brave journalist who made it a selling point of an article

OTOH if you look at AMD's launch videos, they talk a lot about prosumer and creative workloads, then say "Oh, but we know you're interested in gaming too". The focus was definitely on those heavily threaded workloads where Ryzen 7 shines. I think we'll learn more from the timing of the other releases, and in particular the quad core parts that will make up the lower Ryzen 5 and Ryzen 3 CPUs - they're not going to monster multi-threaded tasks anywhere near as well, so AMD really needs to be getting the most out of their single threaded performance. I suspect that means we'll be waiting a while longer for those parts, while scheduler fixed and drivers get released and tested properly...

[kalniel]

I think it's more that the professional market is worth a lot more in terms of revenue - they know they don't have long to steal market share from Intel so hit with the best return as soon as possible.

[CAT-THE-FIFTH]

I found RAM speed was a terrible bottleneck for P4s. I had a Dell 650 with two P4 XEONs, but its DDR266 RAM really held it back, it was stomped by an Athlon64 using DDR400. For 3D tasks, as the res/detail went up, the AMD machine could be as much as six times faster:

http://www.sgidepot.co.uk/misc/mysystemsummary.txt

Back then, the only thing I missed about moving off the Dell 650 was losing the hw U320 SCSI RAID for the C-drive, but of course later SSDs came along to save the day.

Well I had an XP 2800(Barton core) and a P4 2.53GHZ(Northwood B) at the time,and I think the Northwood B was about the closest Intel got to actually trying to reach parity with the Athlon line....until the Athlon 64 came and blew the back doors of the P4 off!!

The XP 2800 was more a commiseration purchase though - I was after an XP-M and only OcUK sold them at the time,and stock was like gold dust.

Ah I see! Yes indeed, an oc'd 2700K is quite a step up, despite IB's IPC advantage.

Note that I do specifically focus on the 2700K. The 2600K is also a good CPU of course for oc'ing, but its potential does vary, hard to be sure what any particular unit can handle. By contrast, every 2700K I've obtained (seven so far) can run at 5GHz with ease, and at a sensible voltage. I'm sure one could push them further (5.2, maybe 5.3?), but I've never tried, 5 is plenty and sits well with the eyeballs on a CPU-Z. Here's my system (been using it for about 5 years now):

http://www.overclock.net/lists/display/view/id/2415471

I plan on upgrading to 4K soon though, for which I'll switch it around with my other system, fit a 1080 Ti and likely fit an SM961 to hold game data (then the 2700K will drive the HDTV, for which a 980 is plenty):

http://www.overclock.net/lists/display/view/id/6211115

Very much depends on the task, but yes some definitely benefit, especially in the prosumer space. After Effects benefits from faster RAM (I observed a 10% slowdown in RT3D renders when changing RAM from 2133 to 1866), and of course lots of it. Friend of mine is working on a project atm which easily gobbles 40GB, hence why I reckon he'd be better off waiting for Naples (the 3930K system I built for him already has 64GB).

A lot of games though, based on site reviews, are less affected by RAM speed, oftem even less so by latency. Amuses me when a RAM kit review commentary says one brand "beats" another for a particular game when the difference isn't even a fraction of the margin of error.





Must admit, reading all the forum comments, etc., I'm thinking of leaving my 1700X build for just a little while, perhaps until the end of March, let the BIOS updates roll through somewhat. I'll have enough to do benching a 1080 Ti anyway.





That was something which always annoyed me about the way Crysis was lauded so much as a benchmark: nobody bothered to ask whether the way it hammered GPUs was perhaps because the engine just wasn't that well written, no matter how good the visuals looked (and I do love the visuals). The speedups obtained with Crysis Warhead suggested this was indeed the case to some extent. Eventually though, my wanting to play at higher custom detail/res pushed the load onto the GPU, so for me CPU power was less relevant. I was pushing the VRAM limit of two 580 3GB SLI due to the custom mod settings I use, high res shadows, very long draw distances, etc. A single 980 worked better, but still only manages 45fps (reeeally high detail! I hate popup in games).

Ian.

I also use a SFF rig which is partly why I didn't go the K series route and the Xeon E3 series were well under £200 which was a decent saving. Regarding the RAM,certain titles,could actually see a decent uplift from RAM being run at 3000MHZ instead of 2133MHZ of upto 20% with SKL/KL which was surprising,and indicates Intel is probably starting to hit memory bandwidth limitations in some ways too.

Also,Crysis was awesome - even without mods,and with mods looks awesome even now. It was the last time I actually got excited about a game to justify building a new rig for(also it was for work related stuff too).

[watercooled]

Whilst there were almost certainly issues that AMD knew about before launch, I think it's worth remembering that some of these issues are things that AMD themselves couldn't necessarily fix.

That is true. To an extent you'd hope AMD would have been pushing the likes of MS and their board partners to get this sort of issue resolved before release, but a combination of timing and the points you mention are plausible reasons things haven't quite happened that way.

It could also be the main reason they're holding the Ryzen 5 & 3 launches back - if they knew the Windows 10 scheduler makes Zen's single-threaded performance look bad it would make sense to launch the chips that are meant to be used for monstering multi-threaded tasks first and hold back the lower core-counts for once the scheduler gets fixed...

Yeah that's exactly what I've been thinking - get *something* out the door and put up with some launch-day negativity in order to get this sort of issue ironed out, both for the other CPUs and what are probably more high-volume (on the consumer side) APUs. TBH it's a potentially smart way of forcing re-testing of Ryzen with newer software etc.

[DanceswithUnix]

IMO a very level-headed look at Ryzen, particularly gaming performance, and highlights some of the faulty reasoning for using very low resolutions as 'forward looking' benchmarks. TBH it's something I've sort-of noticed in recent benchmarks but not really had it register - Piledriver has begun closing the gap or outperforming the 2500k it was pitted against in gaming benchmarks on its release, and again some places were playing down 'real-world' benchmarks (i.e. benchmarking games at settings you'll actually use), claiming that the performance difference would become worse as games became 'more demanding' in the future, or as more powerful GPUs became available. That simply hasn't happened. It seems it's not quite that easy to extrapolate future gaming performance.

That was an interesting video, it does amuse me when I see the 8350 doing well in modern benchmarks (I bought mine for simulating payment systems, which I no longer do but it hasn't let me down in my meagre gaming needs enough for replacement so far).

In the past I have benchmarked games at different resolutions and found some strange anomalies in the results. It seems to me that games and drivers are tuned for popular resolutions, which makes a lot of sense. Game writers will target 1080p as monitors are cheap and plentiful, and people at that resolution probably can't afford the latest graphics card and could do with all they help they can get. Popularity in benchmarks then makes the graphics card driver writers target the specific cases as a humanly undetectable 5% frame rate difference can be the difference between winning and losing against competing cards which translates to sales.

Driver writers optimising for popular resolutions helps players at those resolutions, I don't want that effort going into the likes of 640x480, I think that is what annoys me most when seeing low res benchmarks.

[ik9000]

Driver writers optimising for popular resolutions helps players at those resolutions, I don't want that effort going into the likes of 640x480, I think that is what annoys me most when seeing low res benchmarks.

but how else will they prepare to release it on mobile in 5 years time?

[CAT-THE-FIFTH]

but how else will they prepare to release it on mobile in 5 years time?

I was looking at some of the old reviews of the Athlon,Athlon XP,Athlon 64 and Athlon X2 and there were plenty of silly low resolution and normal resolution tests - it didn't matter what resolution 9/10 they won and it was fine back then and nobody was moaning!

We need to be careful getting too much into reading social media,everybody - soon 1080P won't enough as we are now hearing rumblings that is too low even though that is the most common resolution. At this rate everybody will be testing at 5K by next week!