*** AMD "Zen" thread (inc AM4/APU discussion) ***

[Stanners]

Stop with the childish bickering

 

[CAT-THE-FIFTH]

Cat i think the problem is your making mountains out of molehills and conjuring up issues that are not there...

The facts are as follows

1) Intel does not scale as well as AMD does with faster Ram
2) AMD Ryzen scales incredibly well with ram, to the point its a necessity to user the fastest ram possible
3) AMD are still working on providing the best ram support, in the meantime your mileage may vary in ram compatibility
4) Anyone can cherry pick reviews to show any point they are trying to make, we all know it, you know this to be true
5) Why cant we all just get along!

Its not that - look at the difference between the internal benchmark and the actual playthrough. Its very worrying since most sites use the canned benchmark.

I never realised it was such a huge difference - it makes me wonder whether half the AOTS CPU or GPU results are really that valid. Its like with Deus Ex:MD,ROTTR and Metro:LL where the internal benchmarks were pointless. I even tried out a GTX1080 and RX470 in the ROTTR benchmark and where the canned benchmark seem to show hardly an improvement under DX12,but in game it actually did seem to weirdly enough and more on AMD.

Is there any other site which uses their own playthrough,do they see the same issue or not??

I am not interested in the Core i7 7700k but only mentioned it due to RAM thingy,since £300+ is not worth it,in light of the R5 being under £250 and Intel really needs to lop off a decent amount of the price.

Edit!!

Thats another thing - how do we know if some of the poor results for AMD are down to the canned ones? It can go either way!

Most sites use the ROTTR internal benchmark to test CPU performance.

How do we know if that paints AMD in a worse light or not??

 

[DragonQ]

Look at the scaling from 2133MHZ to 3200MHZ. Remember,DF uses custom sequences.

That's an i5-6600, not an i7-7700K. I could be wrong but that could make a big difference. Also, I don't like the fact that these review videos never actually show you the hardware setup used, even in the description. I assume it's 1080p with a Titan X OC or something crazy? At least a 980 Ti is more realistic.

On the other hand, this video also shows no difference in most games (at least I think, all I can see is blurry screengrabs of the charts since my internet connection is awful right now):

 

[CAT-THE-FIFTH]

That's an i5-6600, not an i7-7700K. I could be wrong but that could make a big difference. Also, I don't like the fact that these review videos never actually show you the hardware setup used, even in the description. I assume it's 1080p with a Titan X OC or something crazy? At least a 980 Ti is more realistic.

Its not going much if any difference though - once I am at home I will try and see if I can find a Core i7 one. Some games it makes no difference,but other it does and FO4 is the worst so is an outlier. But here is their i7 7700K review:

http://www.eurogamer.net/articles/digitalfoundry-2017-intel-kaby-lake-core-i7-7700k-review

In certain games it can have huge improvements which high speed RAM and anyway most reviews even showing RAM scaling with Ryzen are doing something similar:

http://www.legitreviews.com/ddr4-me...tform-best-memory-kit-amd-ryzen-cpus_192259/4

But it does not answer why with the same RAM the internal and custom runs looks so different. I think maybe some more testing needs to done as it is only one data point so far.

 

[DragonQ]

But it does not answer why with the same RAM the internal and custom runs looks so different. I think maybe some more testing needs to done as it is only one data point so far.

I think that's fairly obvious - when doing internal testing of performance improvements it's likely that the testing team uses their own in-game benchmark, so in theory that will always be the best case scenario. It probably shows that the in-game benchmark is a poor representation of game performance more than anything else.

 

[SiDeards73]

That AOTS is garbage whichever way you look at it anyhow, its definitely not a great representation of DX12, personally for me, Division was a good DX12 implementation, Mantle under BF4 was also a good implementation of an API, Vulkan in Doom etc, pretty much any implementation where the results are good across the board (obviously Mantle being AMD only). But it is weird that the benchmark performance is not really indicative of the actual ingame performance, why would that be? only the devs can answer that honestly i guess.

My bro just ordered his 1700 rig yesterday, he went with 3600mhz ram, we discussed it last night, i said worse case you have to downclock the ram, and in the future you should be able to run it at its intended range. I dont think anyone, especially Ryzen owners will recommend 2400mhz ram for the CPU would they? given how well it scales with faster ram, you would have to be a bit daft to not try and get the most out of it, right now, yeah theres issues, in 6 months time im betting they wont be an issue then.

 

[Bobmunkhouse]

So will Zen+ next year support DDR5?

 

[DragonQ]

So will Zen+ next year support DDR5?

Very unlikely. I guess it's possible that the 3rd generation could have a dual IMC and a backwards-compatible AM4+ motherboard could be released that supports DDR5. All speculation though.

 

[CAT-THE-FIFTH]

I think that's fairly obvious - when doing internal testing of performance improvements it's likely that the testing team uses their own in-game benchmark, so in theory that will always be the best case scenario. It probably shows that the in-game benchmark is a poor representation of game performance more than anything else.

It seems a bit pointless then - not sure what they were thinking - having said that Deus ex and ROTTR internal benchmarks were pointless too.

That AOTS is garbage whichever way you look at it anyhow, its definitely not a great representation of DX12, personally for me, Division was a good DX12 implementation, Mantle under BF4 was also a good implementation of an API, Vulkan in Doom etc, pretty much any implementation where the results are good across the board (obviously Mantle being AMD only). But it is weird that the benchmark performance is not really indicative of the actual ingame performance, why would that be? only the devs can answer that honestly i guess.

My bro just ordered his 1700 rig yesterday, he went with 3600mhz ram, we discussed it last night, i said worse case you have to downclock the ram, and in the future you should be able to run it at its intended range. I dont think anyone, especially Ryzen owners will recommend 2400mhz ram for the CPU would they? given how well it scales with faster ram, you would have to be a bit daft to not try and get the most out of it, right now, yeah theres issues, in 6 months time im betting they wont be a

Actually do you have ROTTR? Don't use the rubbish internal benchmark under dx12. The village in the geothermal valley under dx12 showed some decent gains for me with an rx470 even with an ancient ib CPU.

 

[SiDeards73]

It seems a bit pointless then - not sure what they were thinking - having said that Deus ex and ROTTR internal benchmarks were pointless too.

Actually do you have ROTTR? Don't use the rubbish internal benchmark under dx12. The village in the geothermal valley under dx12 showed some decent gains for me with an rx470 even with an ancient ib CPU.

Naw, cant say ive ever been that interested in the Tomb Raider games lol, might get it to try though

 

[CAT-THE-FIFTH]

Naw, cant say ive ever been that interested in the Tomb Raider games lol, might get it to try though

They are an acquired taste - Lara does always seem starstruck all the time though. Shiny game tho. Interestingly enough the Tess on Polaris is actually not bad - the rx470 4gb even with hairworks upto the max did far better than I expected in W3.

Having said that the i7 7700k will need a hefty price cut soon - the R5 CPUs are going to be most of an R7 for gaming at £180 to £250.

The locked core i5 CPUs can't be overclocked and can only run RAM in spec.

Plus good riddens to the i3 7350k.

 

[humbug]

Nope it isn't since you have quoted their results before when they showed AMD in a positive light and all the people saying it is BS now are basically trying to bury stuff. You know very well that the other sites are using canned results from internal benchmarks which a number of you have yourselves have said in CPU and GPU threads are not a good indication of gameplay,and moreover you are yet another one who has not read it but are reacting. The level of flip-flopping is hilarious - so I expect next time when you and your mates try and use PCGH repeatedly to show AMD in a positive light I will call you out on it - it shows you had silly some of you are getting when you are basically trying to call out any website which does not show Ryzen 100% in a positive light which is basically what happens in the GPU section.

AMD themselves would not have sent out 100s of dev kits and admitted as much that gaming performance is not a good.

Moreover clinging to RAM speed is BS(using your words),because:
1.)BOTH test systems used the same RAM and they are using a less optimised older generation chipset for the Core i7
2.)Intel IMC overclocks more and can handle higher speed RAM better
3.)Running at 2400MHZ puts Intel in a worse light NOT AMD

Plus moreover all your deflecting does not change the fact,that for some reason even with 2400MHZ RAM,the internal benchmark seems FAR better than actual game play.

Moreover burying anything negative about Ryzen and going la!la!la!la! means diddly squat -
even the biggest AMD fan should want this highlighted since it means AMD can give feedback to its sponsored dev that they need to do a better job.

Trying to call out PCGH as a Intel shill is stupid when most of you were extensively using the FX8350 Crysis3 results on this forum to show how an FX8350 might do in future games. They are hardly an Intel shill site - using absurd logic are you telling me they were a AMD shill site too.

CAT you know as well as the rest of us using slow RAM on Ryzen has a significant negative impact on performance, why when before Ryzen testing CPU's was done with RAM at 3Ghz+ suddenly they are now testing with slow RAM, much slower than Ryzen is capable of, slower than even its base rating of 2667Mhz?

 

[ltron]

Pcgameshardware have now added the 6900k into the graph and it performs even worse than the 1800x in-game: http://www.pcgameshardware.de/Ryzen...als/AMD-AotS-Patch-Test-Benchmark-1224503/#a3

Something is clearly broken. Thankfully, AMD have not been caught cheating after all.

 

[SiDeards73]

Pcgameshardware have now added the 6900k into the graph and it performs even worse than the 1800x in-game: http://www.pcgameshardware.de/Ryzen...als/AMD-AotS-Patch-Test-Benchmark-1224503/#a3

Something is clearly broken. Thankfully, AMD have not been caught cheating after all.

What i find so disappointing in all of this is the rush to publication of these so called techsites to grab the headlines with world firsts etc, so they jumped all over the benchmark citing its not indicative of actual gameplay and at the same time smear AMD Ryzen all over the press, while not revealing the whole truth that also the 6900k is even worse in this scenario... its sensationalist reporting at its finest, pick up on the current crest of press around Ryzen and find a way to get some hits on your site with something controversial, sad thing is if they had presented the entire case correctly the first time they would have garnered more credit, instead they point the finger at AMD then quickly have to backtrack when its apparent it affects Intel chips equally or more so.

Another website to add to the "Dont bother reading" list, seriously all these techtubers and review sites that are springing up and more often than not totally useless.

I still maintain if i want an honest opinion of hardware, i will come to forums like this to get it from people who actually own and use it daily and find out their point of view.

 

[humbug]

There is something very odd going on there, the Minimum FPS here are whats CPU bound, look at them.

Are we supposed to think that the 1800X has a 29% higher IPC than the 6900K? and the 7700K a 180% higher IPC?

What is going on with this ATOS benchmark tool?

 

[Gzero]

I still maintain if i want an honest opinion of hardware, i will come to forums like this to get it from people who actually own and use it daily and find out their point of view.

The only problem with that is what I read recently on a certain subreddit of user reports:

My Fx-8350 was smooth. My Ryzen R7 is butterly smooth.

Okay then...

And now the latest is that input lag is lower! Because reasons: all in one chip bla bla bla.

 

[nkuk]

You have also obviously missed that the Intel setup also used the SAME 2400MHZ dual ranked RAM,too so is Intel also now gimped....

That does not change the fact that the article has false statements, contradictions, missing configuration information and missing testing information. The guy is making incorrect assumptions that are negatively affecting the results and making statements that might mislead the reader, and whether the findings are accurate or not, because the surrounding information is poorly presented it makes the results questionable.

 

[humbug]

The only problem with that is what I read recently on a certain subreddit of user reports:


Okay then...

And now the latest is that input lag is lower! Because reasons: all in one chip bla bla bla.

Sorry but he's right about the FX-8350, i'm not defending that chip, i had two (also an FX-9590) yes they are hot when overclocked, they use a lot of power and often in games an i5 is faster, i now have a 4690K... and actually yes, the FX-8 core chips in my experience are smoother. the Intel chips seem to have issues with the tiny amount of cache that they have compared with the old FX-8 chips. Ryzen also has vastly more cache than the 4 core i7's and its not just him and me saying Ryzen is smoother than the 7700K even where it is slower, half the reviewers are saying it.

This isn't the first time i have made a comparison like that vs the FX-8 core.

Intel's none enthusiast range seem to be built down to a price, just enough cache to get by and keep the die size small and cheap.

 

[silversurfer]

There is something very odd going on there, the Minimum FPS here are whats CPU bound, look at them.

Are we supposed to think that the 1800X has a 29% higher IPC than the 6900K?

Isnt ryzen recognised as helping min fps in some cases. It could also be about handling multiple threading not just IPC. Min FPS could be about wait states, latency not exactly IPC but say efficiency. In some cases ryzen is going to do things differently and possibly give some advantage in some scenarios ?

 

[harney]