Ryzen 7 1700X vs Kaby Lake 7700k

[tribeca600]

Just got my Ryzen 7 1700X system up and running, specs in my signature. I eventually chose the Ryzen 7 8-core over 7700k due to some valid arguments - that multi-core is the future and from what I read software is constantly being optimized for the new platform. As for gaming, it seems more about optimization for the new architecture rather than the processor itself. I am sure once software catches up this chip is going to shine even more than it does now.

 

[krumme]

As I am sure you know but have conveniently forgotten [H] looked into this: https://www.hardocp.com/article/2017/05/26/definitive_amd_ryzen_7_realworld_gaming_guide/

If you want to game the 7700K is just faster. Doesn't matter the resolution, Xfire, etc.

Except it isnt faster in a new game like bf1 for the 0.1% mins that matters. H omits graphs and min numbers ( frametimes preferably) excactly in THE game where there is a sizable difference for subjective gameplay. Its a mess to meassure but thats how the gaming is.
For the rest of the time they talk 4% differences up like 67 vs 70 fps on ultra makes a subjective difference. Quite funny when for the gpu test they would consistently say it would be the excact same gaming experience.

Where a 7700k makes sense vs 1700 is for the high framerate 144 monitors and for games that cant use the more cores. Here ryzen is still limited in some of the half old games.
Your avg framerate is not cpu limited on 60fps monitors but gpu limited.

Eg. On bf1 MP, for the 99.9% percentile the frametime for 7700k is 18.2 ms while its aprox 10 ms for the 6900 and 11 ms for the 6850 and 1800x. Cores matters where it counts.

https://www.computerbase.de/2017-03...d-1-dx11-multiplayer-frametimes-in-percentile

And in some dips the 7700k goes above 24ms, while the higher core machinces dont have those dips:
https://www.computerbase.de/2017-03...-frametimes-ryzen-7-1800x-gegen-core-i7-7700k

Granted there is games where the ryzen is still stuck on the mins, but hopefully the updates that we have seen will come to the last handfull that needs it.

 

[tribeca600]

Except it isnt faster in a new game like bf1

The thing is that you are looking at current, the future is absolutely multi-core given limitations of clock speed increasing. As I said, when software is optimized for multi-core the gains will be phenomenal, much faster than what you are experiencing now. Intel may respond with cutting prices on their 8-core you never know but as of now that is not the case.

 

[CHADBOGA]

The thing is that you are looking at current, the future is absolutely multi-core given limitations of clock speed increasing. As I said, when software is optimized for multi-core the gains will be phenomenal, much faster than what you are experiencing now.

You do realise it is not always easier to get software that is currently not optimised for multi-core, to be speeded up tremendously by trying to use moar cores.

Many tasks are just serial in nature.

 

[krumme]

You do realise it is not always easier to get software that is currently not optimised for multi-core, to be speeded up tremendously by trying to use moar cores.

Many tasks are just serial in nature.

As Dan Baker explains we need new game models where there is no main thread. He is bringing a new engine to market that does use new principles.

It will hopefully be a major change but its also evident that a rewriting of engines to such new structures is a monumental task and will take years.

I remember his ideas was heard first time at a q&a for mantle back at bf4 launch and we are yet to see the first iteration of the engine not to say games. Its a giganic uptake but its also the only way forward. But there is a solution comming even for gaming.

 

[Eug]

If you're mostly gaming, get the 7700K. If you're mostly doing productivity tasks, get a Ryzen CPU.

Well, it would depend upon which productivity tasks of course. I've been reading some of the Puget Systems reviews which cater to photo and video editors, and they provide some useful insight IMO.

For Photoshop work, the 7700K won most of the benchmarks over Intel's own 6-core or even 10-core models. No, it's not an i7 vs Ryzen benchmark, but nonetheless I think it does say something.

OTOH, for video work, the 6-core 6850K usually beat the 4-core 7700K, especially for 4K footage. For 1080p footage the performance was a lot closer, however.

 

[Eug]

^^^ Actually, it turns out some of that Ryzen testing has already been done. Below are some Lightroom results.

https://www.pugetsystems.com/labs/a...1700X-1800X-Performance-910/#BenchmarkResults

For image exports, Intel chips with more than 4 cores dominated. They were ahead of Ryzen 1700X by a whopping 41-57%. 7700K was behind Ryzen, but by only about 10-11%. For almost every other test though, 7700K won, beating out both Ryzen and Intel's own 6 to 10 core chips.

Yeah, a lot of this is likely due to to software design and optimization, but it's not as if this will magically be fixed any time soon.

 

[Topweasel]

Well, it would depend upon which productivity tasks of course. I've been reading some of the Puget Systems reviews which cater to photo and video editors, and they provide some useful insight IMO.

For Photoshop work, the 7700K won most of the benchmarks over Intel's own 6-core or even 10-core models. No, it's not an i7 vs Ryzen benchmark, but nonetheless I think it does say something.



OTOH, for video work, the 6-core 6850K usually beat the 4-core 7700K, especially for 4K footage. For 1080p footage the performance was a lot closer, however.

Photoshop recently got an update to expand core usage. Do you know when those benchmarks came out?

 

[richierich1212]

Photoshop recently got an update to expand core usage. Do you know when those benchmarks came out?

Benchmarks were from March 3rd, around Ryzen release.

 

[frozentundra123456]

Except it isnt faster in a new game like bf1 for the 0.1% mins that matters. H omits graphs and min numbers ( frametimes preferably) excactly in THE game where there is a sizable difference for subjective gameplay. Its a mess to meassure but thats how the gaming is.
For the rest of the time they talk 4% differences up like 67 vs 70 fps on ultra makes a subjective difference. Quite funny when for the gpu test they would consistently say it would be the excact same gaming experience.

Where a 7700k makes sense vs 1700 is for the high framerate 144 monitors and for games that cant use the more cores. Here ryzen is still limited in some of the half old games.
Your avg framerate is not cpu limited on 60fps monitors but gpu limited.

Eg. On bf1 MP, for the 99.9% percentile the frametime for 7700k is 18.2 ms while its aprox 10 ms for the 6900 and 11 ms for the 6850 and 1800x. Cores matters where it counts.

https://www.computerbase.de/2017-03...d-1-dx11-multiplayer-frametimes-in-percentile

And in some dips the 7700k goes above 24ms, while the higher core machinces dont have those dips:
https://www.computerbase.de/2017-03...-frametimes-ryzen-7-1800x-gegen-core-i7-7700k

Granted there is games where the ryzen is still stuck on the mins, but hopefully the updates that we have seen will come to the last handfull that needs it.

Cherrypick all you want, overall, 7700k *IS* faster link [/URL}. Even in BF1 according to this source. And amazingly, even with only half the cores, it is only 25% slower in applications.

 

[krumme]

Cherrypick all you want, overall, 7700k *IS* faster link [/URL}. Even in BF1 according to this source. And amazingly, even with only half the cores, it is only 25% slower in applications.

Computerbase test real world 64 mp gameplay amiens. And then they show the 0.1% mins.
Its 100% consistent with what you will experience in the game. Its real world testing.
4c is just a huge no go in this game. The dips is to the 24ms or below.
I just watched on of the kids playing it on something i5 hsw and its just barely working. He avoids the 64mp battles and stay on smaller teams. And no an i7 doesnt solve the problems as evident from eg cb test.
Its visible to everyone that owns and play the game.
Bf1 actually issues 10 threads fui.
Just get the new skl 7800x if the r7 1700x doesnt cut it for other applications.

 

[jpiniero]

Computerbase test real world 64 mp gameplay amiens. And then they show the 0.1% mins.

Oddly DX12 is much faster on the 7700K. I can't tell what GPU they are using though.

 

[krumme]

Oddly DX12 is much faster on the 7700K. I can't tell what GPU they are using though.

Dx12 is just worse crap on ryzen than core. It works ok single player and for weak amd bd cores but it gives occasional stuttering in multiplayer and is therefore useless. Its imo a huge dissapointment but thats how it is. But it also shows avg framerates is 100% useless.

I have played bf1 ryzen and core. And core with both 970 maxwell and 470 polaris. Dx12 just doesnt work.

And yes its odd ryzen is worse but its buggy anyway.

 

[krumme]

Aa a follow up to this api issue. I remember in bf4 at some 64 maps there was cpu limitations. It didnt strike as hard as bf1. But when mantel was functional on 7970 and 290 for about the half year it was working and not buggy the speedup and experience was stellar. It took some months to get there and support stopped but it showed the potential.
The physics in bf series phenomenal so its hard to complain but it would be nice if modern i5 could run this flawlessly even in the destruction and man heavy situations.

 

[Topweasel]

Oddly DX12 is much faster on the 7700K. I can't tell what GPU they are using though.

Chances are a Nvidia based card. AFIK Nvidia still hasn't fixed the issue in DX12 where it it is optimized specifically for a 4c8t i7 so that we dont see the performance penalty you would normally get when switching to DX12 on Nvidia video cards. It caps thread usage on that configuration even for Intel's X lineup. Games that are well threaded to begin with (in DX11) like BF1 MP you can see where core scaling is fine till you switch to DX12.

 

[Shivansps]

You can talk about this all you want, but until the main thread in games gets replaced for something else, and we get a new high end api with full MT support and OPTIONAL low level optimizations that does not create a shitload of work on devs, games are going to be ST limited in some way or another.

And we are talking about well intro 5 to 10 years in the future, current hardware does not matter.

Also consoles need to have more power than a low end cellphone CPU. So maybe after the next gen.

 

[itsmydamnation]

You can talk about this all you want, but until the main thread in games gets replaced for something else, and we get a new high end api with full MT support and OPTIONAL low level optimizations that does not create a shitload of work on devs, games are going to be ST limited in some way or another.

And we are talking about well intro 5 to 10 years in the future, current hardware does not matter.

Also consoles need to have more power than a low end cellphone CPU. So maybe after the next gen.

Removal of main thread already started happen LAST GENERATION on consoles, you are fighting a loosing war. Think about this. PS5 in late 2019, 7nm 8-12 core (1 or 2 ccx) TDP 25 watts taking around 50mm sq, that will have significantly more throughput then a 7700K.

Not 5 years not 10, 2 years, its 1915 and your a train salesmen.......

 

[tribeca600]

watch this video:
https://www.youtube.com/watch?v=aNK7EtXrtjs

 

[w3rd]

Cherrypick all you want, overall, 7700k *IS* faster link [/URL}. Even in BF1 according to this source. And amazingly, even with only half the cores, it is only 25% slower in applications.

Not true..!

I have a 4790k @ 4.2Ghz and it chokes on BF1 64 Player. That is why I am upgrading from a 4core i7, to at least an 8core at minimum, for gaming. Just like when I went from dual core to quad.. I am not going to upgrade to another 4core i7, that would be stupid for anyone. Intel even knows this, ergo x299.

If you are building a rig to handle today's games, then you are doing it wrong. And will always be behind the curve.



8 Cores is the new mainstream gaming Platform.
And 6 Core tucks right in under that, as the perfect min/max system, for a base gaming rig. 4 cores are dead. Doesn't matter how fast they go, I don't play old games.

 

[krumme]

Bf1 MP 64 after Computerbase testing for frameverlaub/framerun for the slowest 5 sec part:
Minimum: 1600x
Base: 1800x/6850/7800
Top: i asume its going to be 7820x

I might upgrade to 7820x this august when m atx boards arive and run it stock to keep power under control and more comparable to the 3.8 ryzen i run now. I ryn the computer in a closet so 100c and 350w from the wall is a 100% no go.

Heck people upgrade for 20% perf on gfx too. We might as well get used to it on cpu too.

 

[jpiniero]

Chances are a Nvidia based card. AFIK Nvidia still hasn't fixed the issue in DX12 where it it is optimized specifically for a 4c8t i7 so that we dont see the performance penalty you would normally get when switching to DX12 on Nvidia video cards. It caps thread usage on that configuration even for Intel's X lineup. Games that are well threaded to begin with (in DX11) like BF1 MP you can see where core scaling is fine till you switch to DX12.

Yeah but the issue more is that the 7700K gains like 10% in DX12 and has more consistent frametimes than DX11. I imagine it's hard to properly test MP.

There are other factors to consider of course such as the L3 size and speed, memory bandwidth, etc.

 

[Topweasel]

Yeah but the issue more is that the 7700K gains like 10% in DX12 and has more consistent frametimes than DX11. I imagine it's hard to properly test MP.

There are other factors to consider of course such as the L3 size and speed, memory bandwidth, etc.

That's because of the Nvidia GPU and not that the 7700 is better. In DX11 the game is threaded well enough that when computer base did their Ryzen CPU review, the 6900 won in BF1. The 6950 came in second and Ryzen came in third.

In DX12 all three of them lost performance when using a API that is supposed to be more threaded. The 7700 was on of two CPU's to gain performance. Ended up leap frogging all three of them to take the too spot.

It's actually pretty visible in all games with DX11 and DX12. CPU's scaled fine in DX11 and stopped scaling in DX12 and miraculously on i7 4c8t setups it's the only configuration with Nvidia graphics cards that don't take a hut when switching to DX11 (the one game they don't see a significant bump, they lose like 1%).

Whereas we know that because of the Nvidia arch that DX12 is should always see a dip in performance due to the software scheduler.

 

[krumme]

Yeah but the issue more is that the 7700K gains like 10% in DX12 and has more consistent frametimes than DX11. I imagine it's hard to properly test MP.

There are other factors to consider of course such as the L3 size and speed, memory bandwidth, etc.

Yes. But try to look at cb "Die langsamsten 5 Prozent Frametimes im Vergleich"
For 1800x vs 1700 in dx12 and you will see the 7700k have dips frequently above 16ms and also a few above 24ms.
Surely 1800x is a trainwreck here but the 7700k will get bads dips too.
No matter what you select on 7700k you get the dips. On the 1800x you can select dx11 and get rid of it.

Surely it will be interesting how skl x performs in bf1. So we get some insight into how the new cache impacts perf.

 

[moinmoin]

That's because of the Nvidia GPU and not that the 7700 is better.

More specifically it's Nvidia's GPU driver which appears to be better threaded for DX11 than DX12. The guesses as to why that I've seen so far were that Nvidia essentially realized their own version of Mantel/DX12/Vulkan style threading in its DX11 driver while the approach in DX12 can't be supported directly in their current hardware, adding an additional overhead in their driver layer.

 

[Shivansps]

Removal of main thread already started happen LAST GENERATION on consoles, you are fighting a loosing war. Think about this. PS5 in late 2019, 7nm 8-12 core (1 or 2 ccx) TDP 25 watts taking around 50mm sq, that will have significantly more throughput then a 7700K.

Not 5 years not 10, 2 years, its 1915 and your a train salesmen.......

2 years? it looks like mayor game engines companys did not get the memo them, for it to be in 2 years time, it needed to be on stable branch 3 years ago, minimum.
The new game making concept needs to be fully documented on engines, from there is a 4 to 5 years waiting until the first games come out.

And on PC we need a new API, DX11 is not cut out for the job and DX12 is a hassle to implement, so you need to consider this as well. No such API exist, nor is planned either.

Zen consoles is fine, at least they whould not be the crap they are right now, yet until main thread concept is fully replaced (thats has not happened yet and it does not seems to be close either) we still going to be tied to ST in some way.