AMD Zen Discussion Thread

Lukker-

I wasn't expecting a 4.3 all core boost, that is pretty good tbh. I wonder will 4.5 Ghz one core turbo be achievable with XFR2?

Cuddlesworth

And all the reviews are out.

TDLR,

Productivity beasts.

They are faster then the older generation but on pure single threaded loads(like games) the 8700k is still faster when both overclocked. Gap is closed though.

If you were streaming and gaming on the same machine, much better then the 8700K.

4.3 would be the max, 4.1-4.2 would be the majority for OC(possibly, sample size is too small).

You need a good cooler, they can pull serious power when OC'd(XFR2 included).

god's toy

Inviere wrote: »

Indeed, that chip stood out to me as serious value.

Said I was going to skip the 2nd one but I don't know, it does look good.

Cuddlesworth

There seems to be some controversy.

A small selection of reviewers have cpu's that beat down the 8700k pretty badly in games, like anandtech. There are a lot of factors involved but its going to be interesting to see where those results came from. In anandtech's review, its not even that the 8700k did worse, its that their Ryzen procs did significantly better then other people. Far better then can be explained by memory/XFR or other obvious factors.

wotzgoingon

Cuddlesworth wrote: »

There seems to be some controversy.

A small selection of reviewers have cpu's that beat down the 8700k pretty badly in games, like anandtech. There are a lot of factors involved but its going to be interesting to see where those results came from. In anandtech's review, its not even that the 8700k did worse, its that their Ryzen procs did significantly better then other people. Far better then can be explained by memory/XFR or other obvious factors.

I just checked that out now and from what I see they put stock 8700k against stock 2700X's. From what I gather a stock 2700X will perform near enough to the levels you can overclock it to thanks to precision boost 2 and XFR 2. And most other reviewers overclocked the 8700k to 5GHz and compared which will always favour the 8700k in games that rely more on single threaded performace. It's nice to see Ryzen reach 180 Cinebench R15 single threaded score which is the same as a non overclocked 6700k. As I always say I am all for AMD and will buy one of their processors eventually still debating on this gen or next. The cost of decent RAM is putting me off the move.

There was loads of click bait thumbnails on YouTube yesterday saying it OC's to 5.something ghz. I thought fanboyism only existed in the console realm.

Lukker-

Deleted User wrote: »

There was loads of click bait thumbnails on YouTube yesterday saying it OC's to 5.something ghz. I thought fanboyism only existed in the console realm.

If you follow the review circuit you will see they very much exist for all sides, AMD, Intel, Nvidia. Saying anything against any company will get you called a shill by the other side etc.

Venom

Deleted User wrote: »

There was loads of click bait thumbnails on YouTube yesterday saying it OC's to 5.something ghz. I thought fanboyism only existed in the console realm.

I saw some articles pop up on my phones news feed about the 5 GHz overclocks and benchmarks but it was down to LN2 cooling so not something realistic in the real world for everyday use.

K.O.Kiki

GamersNexus review is live!

(other reviews via Reddit)

Looks like AMD managed to lower the voltages & heat a lot compared to Zen 1000-series, but OC max is around 4.2GHz.

Lukker-

I'm really curious to see what Intel have up their sleeve for below 14nm

I reckon they are struggling to get high clocks and that's why they've stayed with the process for so long. Probably why they re pushing multithreading now because they are gonna have to switch from being IPC focused to multicore

AMD's 12nm Ryzen hasn't got the best clock speed, but the IPC is up there, the power draw of the new Ryzens is crazy, at stock the R7 2700X pulls about the same as a 4 core 4 thread i3 8350k. Infinity fabric has come on a good bit as well as latency so that certainly helped close the gap on Intel.

RN, I would still advise Intel for gaming exclusively, but any sort of mixed work load, potentially wanting to stream etc or even just wanting to future proof for multithreaded games and app it would have to be a Ryzen.

TerrorFirmer

That Anandtech review is really dodgy. In their benchmarks a Ryzen 2600 is crushing the 8700K in games? I don't think so. No other review has those findings or even close.

wotzgoingon

TerrorFirmer wrote: »

That Anandtech review is really dodgy. In their benchmarks a Ryzen 2600 is crushing the 8700K in games? I don't think so. No other review has those findings or even close.

I was reading up on that and they had the latest patches on all CPU's and motherboards used for both Intel and AMD so for Intel they had the latest patches for Meltdown and Spetre and most other reviewers didn't. So for that reason their results are lower supposedly.

TerrorFirmer

I really don't think the patches would affect that much though. I'm not saying there's no impact, but the charts on Anandtech seem really weird to me. They have Ryzen 1600 coming very close to 7700K for example in gaming benchmarks!

Even taking minor performance hits into account, I can't believe they would actually cripple Intel CPU's to that level.

Venom

TerrorFirmer wrote: »

I really don't think the patches would affect that much though. I'm not saying there's no impact, but the charts on Anandtech seem really weird to me. They have Ryzen 1600 coming very close to 7700K for example in gaming benchmarks!

Even taking minor performance hits into account, I can't believe they would actually cripple Intel CPU's to that level.

Yeah if the patches to fix those security issues hit the CPU's performance that much, it would have been all over the various tech sites and social media channels long before now.

K.O.Kiki

BitWit tested a 2700X on a low-mid B350 board:

Serephucus

No graphs. TL;DW?

Cuddlesworth

TerrorFirmer wrote: »

That Anandtech review is really dodgy. In their benchmarks a Ryzen 2600 is crushing the 8700K in games? I don't think so. No other review has those findings or even close.

Its weird but some other smaller reviewers saw similar numbers(but its a very small minority), they are retesting at the moment. Whats even weirder is that its not that the Intel proc's did badly because their results were similar to the 8700k's release benchmarks. Its that their Ryzen proc's did significantly better. No real explanation though, computerbase did cover most of the bases and while the 2700x was within 5% of the 8700k with memory and OC its just odd that anandtech results show that.

Slutmonkey57b

I suppose one of the interesting questions is what is the fundamental difference between ryzen chips.

The original 1700 performed identically to the 1800x in quite a few scenarios. The 2000 series is just following the same pattern. It's unlikely games are going to use more than 4 threads, so the extra volts don't help. Ryzen is heat and voltage limited, so binning and cooling can have a significant impact. AMD aren't enforcing hard cutoffs in the design to prevent processors overlapping like Intel did until kaby lake. Intel's various sleights of hand tuned down the noise about spectre but in the areas of the tech press that go beyond canned benchmarks and gaming, some really bad findings came out for Intel.

I don't presume anandtech's numbers are 100% in this case, but I would trust them to dig into understanding the numbers a lot more than most, so it'll be interesting to see what they uncover.

Lukker-

Slutmonkey57b wrote: »

I suppose one of the interesting questions is what is the fundamental difference between ryzen chips.

The original 1700 performed identically to the 1800x in quite a few scenarios. The 2000 series is just following the same pattern. It's unlikely games are going to use more than 4 threads, so the extra volts don't help. Ryzen is heat and voltage limited, so binning and cooling can have a significant impact. AMD aren't enforcing hard cutoffs in the design to prevent processors overlapping like Intel did until kaby lake. Intel's various sleights of hand tuned down the noise about spectre but in the areas of the tech press that go beyond canned benchmarks and gaming, some really bad findings came out for Intel.

I don't presume anandtech's numbers are 100% in this case, but I would trust them to dig into understanding the numbers a lot more than most, so it'll be interesting to see what they uncover.

Plenty of games are pretty well multithreaded. But it's a fair point. Not much difference between a 2600 and a 2600x expect you get a better cooler and perhaps .1 Ghz more all core overclock.

K.O.Kiki

Serephucus wrote: »

No graphs. TL;DW?

B350: 4.1GHz
X470: 4.2GHz

Slutmonkey57b wrote: »

I suppose one of the interesting questions is what is the fundamental difference between ryzen chips.

The original 1700 performed identically to the 1800x in quite a few scenarios. The 2000 series is just following the same pattern. It's unlikely games are going to use more than 4 threads, so the extra volts don't help. Ryzen is heat and voltage limited, so binning and cooling can have a significant impact. AMD aren't enforcing hard cutoffs in the design to prevent processors overlapping like Intel did until kaby lake. Intel's various sleights of hand tuned down the noise about spectre but in the areas of the tech press that go beyond canned benchmarks and gaming, some really bad findings came out for Intel.

I don't presume anandtech's numbers are 100% in this case, but I would trust them to dig into understanding the numbers a lot more than most, so it'll be interesting to see what they uncover.

The differences between 2600/2600X & 2700/2700X this generation are very simple

X version gets a better stock cooler & XFR (auto-overclock).

wotzgoingon

Last gen non X chips were recommended as you could overclock them to X's performance but this gen due to the small price difference between X and non X it is recommended to go for the X versions due to small bit better binning and even if you run them stock they perform better than overclocked last gen.

Venom

K.O.Kiki wrote: »

B350: 4.1GHz
X470: 4.2GHz


The differences between 2600/2600X & 2700/2700X this generation are very simple

X version gets a better stock cooler & XFR (auto-overclock).

The stock coolers that come with the 2700X and 2600X are pretty meh from what I've seen online. TTL from OC3D strongly recommends getting even the most basic after market cooler to replace them.

Jayz2cents has done a vid on the benefits of overclocking the CPU vs memory vs both.

wotzgoingon

Venom wrote: »

The stock coolers that come with the 2700X and 2600X are pretty meh from what I've seen online. TTL from OC3D strongly recommends getting even the most basic after market cooler to replace them.

But he was stress testing if you just use the PC for gaming then the stock 2700X CPU cooler is fine and suppose to be pretty quiet as well.

Venom

wotzgoingon wrote: »

But he was stress testing if you just use the PC for gaming then the stock 2700X CPU cooler is fine and suppose to be pretty quiet as well.

No, right out of the box there were issues with the cooler from the 2700X. The base plate is not smooth like every other CPU cooler on the market and instead has weird ridges running across the base. The fan is also crazy loud according to TTL.

Is it possible to get the 2700X cheaper without the AMD fan? I'd much rather have what every savings there would be going towards something like a Noctua D15s.

wotzgoingon

Venom wrote: »

No, right out of the box there were issues with the cooler from the 2700X. The base plate is not smooth like every other CPU cooler on the market and instead has weird ridges running across the base. The fan is also crazy loud according to TTL.

Is it possible to get the 2700X cheaper without the AMD fan? I'd much rather have what every savings there would be going towards something like a Noctua D15s.

I seen his video alright but others were saying it was decent.

They really should sell some without coolers alright especially the 2700X as that cooler would add a bit to the cost and pointless really when most will buy a aftermarket cooler. But you could sell it to someone with a 2600/X.

K.O.Kiki

Only €30 difference between the non-X / X Ryzen 2000 series.

Wraith Prism cooler: from Kitguru:



It's made by CoolerMaster.

K.O.Kiki

Joker compares Wraith Prism to a BeQuiet Dark Rock 4 ($75)

6C difference in temps.
4-5fps difference in FPS/1% lows.
noise test @ 6:52

z0oT

Here's a review where there's little difference between the 8700k & 2700X in games at 1080p. The difference here from other sites is that the testing was done with a GTX 1070 as opposed to a 1080 Ti.
https://pcgamesn.com/amd-ryzen-7-2700x-review-benchmarks

To me this is a more relevent result, since there are more people out there with something like a 1070 class graphics card than a 1080 Ti. (I've a 1070 myself)

Cuddlesworth

That's more of a argument as to how and why reviewers test gaming performance. A crusade that Jim from Adored has been going on about for a while.

The gist is that reviewers use the most powerful card at a "relativity" low resolution and graphics setting under the assumption that it reduces the graphics bottleneck. But this puts a huge emphasis on single threaded performance through high frame rates because the main scheduling thread has to do everything in a very short space of time. Which suits whatever proc can do the best IPC at the highest frequency. And Nvidia's drivers put a huge emphasis on single threaded performance too, as they offload graphics scheduling to the CPU vs AMD's hardware scheduling design.

Meanwhile, the number of people running a 1080p monitor with a 5.2ghz 8700k and 1080Ti needing 200+fps is minimal.

Once you drop into decent graphics settings at normal resolutions today the differences basically disappear.

z0oT

Cuddlesworth wrote: »

That's more of a argument as to how and why reviewers test gaming performance. A crusade that Jim from Adored has been going on about for a while.

The gist is that reviewers use the most powerful card at a "relativity" low resolution and graphics setting under the assumption that it reduces the graphics bottleneck. But this puts a huge emphasis on single threaded performance through high frame rates because the main scheduling thread has to do everything in a very short space of time. Which suits whatever proc can do the best IPC at the highest frequency. And Nvidia's drivers put a huge emphasis on single threaded performance too, as they offload graphics scheduling to the CPU vs AMD's hardware scheduling design.

Meanwhile, the number of people running a 1080p monitor with a 5.2ghz 8700k and 1080Ti needing 200+fps is minimal.

Once you drop into decent graphics settings at normal resolutions today the differences basically disappear.

I'm well aware of why they do it like that and what the implications are.

Your comment about the 5.2GHz 8700k & 200FPS is pretty much the point I was making.