AMD RyZen CPU Reviews

Gubbi said:
It's probably something less nefarious, like using non-temporal stores to a higher degree, which Ryzen seems to take issue with (or something similar which turns out to be a pathological case on Ryzen)
Click to expand...
That's likely. I am wondering whether Ryzen has more limited write combining buffer storage than Intel, or whether there's some problems with the coherence algorithms (AMD is different than Intel). This is a plausible reason, since GPU drivers use a lot of write combining when writing data to GPU buffer (write combining is similar to temporal stores). I have seen huge perf increases (10x+) by slightly modifying our write combining code on consoles.

Nvidia's DX11 drivers are also internally multi-threaded. The driver could exchange across cores, causing problems when data moves from CCX to another.
 
Silent_Buddha said:
Which still doesn't explain why NVidia cards do just fine on Intel CPUs but performance drops massively on Ryzen. And certainly doesn't explain performance failing in Dx9 on Ryzen when CPU limited, either. Dx11 games appear to be the only ones unaffected by this effect as I haven't seen a report of a Dx11 game having abnormally low performance using an NVidia card on Ryzen. It's always possible though.

It's just bad drivers at the moment. Hopefully, it gets fixed soon.

Right now I'm just very very glad I haven't pulled the trigger on Ryzen yet as I play a lot of games that don't get benchmarked and I'd hate to have NVidia's drivers ruin my gaming experience until they get fixed. At this point Vega can't come out soon enough. And hopefully it's at least competitive at the high end, it'd be nice to have some options again. I usually try to keep my video cards around for at least 2 years, but OMG do I want to get rid of this 1070 ASAP. Of course, this is assuming that AMD doesn't bork the drivers for Vega somehow. /sigh.

Regards,
SB
Click to expand...
You miss the aspects that raises pointers beyond Ryzen just like my comment about Civ 6.
You do not see anything wrong with a 480 beating the GTX1080 on Intel 7600k?
By that reasoning we might as well say AMD has DX11 driver problems because their performance tanks with Fallout 4, when in reality it comes back to the game-rendering engine design not ideal for said hardware.

But the drops are in those games I mentioned and not others (unless one deliberately sets the resolution to 720p), all the games have a similar trend in terms of Civ 6/Deus Ex Mankind/Hitman/Total War Warhammer development.
This is happening in those games for both DX11 and DX12 in relative terms on both machines (albeit further exacerbated on Ryzen for Nvidia).
Yet others scale very well with Nvidia on Ryzen such as AoTS (probably one of the better developed games around DX12), Sniper Elite 4, Gears of War 4 (apart from the 720p behaviour), and quite a lot of DX11 and a few of the more recent DX12 games work well with Nvidia on Ryzen.
Like I said those games with weird performance for Nvidia on DX11/DX12 also tend to be the ones optimised for AMD or from the period of devs making an engine a fudge of DX11 and DX12, and this is also on Intel machines just like I mentioned with Civ 6.

And we could say the same about AMD as their DX11 performance is trending lower than expected on Ryzen as well in certain games, but again certain games so should we say then they also have a driver problem.

Now I agree part of it could be driver problem involved in all of this (or may help anyway) but the unusual behaviour is finding its way back to certain games while others perform extremely well for Nvidia on both platforms.
This seems a more complex problem involving multiple factors to the specific games that seem around DX11-DX12 fudging period (games mentioned so far) to possibly something the Nvidia hardware/software is not happy with in terms of the Ryzen platform exacerbating a specific situation.

But Civ6 performance is way beyond driver issues for Nvidia and comes back to possibly optimisation only for AMD at an engine/game level, a game lead by graphics engineers who were also involved in implementing Mantle with AMD's support into Civ engine back in 2014.
 
Last edited:
CSI PC said:
You miss the aspects that raises pointers beyond Ryzen just like my comment about Civ 6.
You do not see anything wrong with a 480 beating the GTX1080 on Intel 7600k?
By that reasoning we might as well say AMD has DX11 driver problems because their performance tanks with Fallout 4, when in reality it comes back to the game-rendering engine design not ideal for said hardware.

But the drops are in those games I mentioned and not others (unless one deliberately sets the resolution to 720p), all the games have a similar trend in terms of Civ 6/Deus Ex Mankind/Hitman/Total War Warhammer development.
This is happening in those games for both DX11 and DX12 in relative terms on both machines (albeit further exacerbated on Ryzen for Nvidia).
Yet others scale very well with Nvidia on Ryzen such as AoTS (probably one of the better developed games around DX12), Sniper Elite 4, Gears of War 4 (apart from the 720p behaviour), and quite a lot of DX11 and a few of the more recent DX12 games work well with Nvidia on Ryzen.
Like I said those games with weird performance for Nvidia on DX11/DX12 also tend to be the ones optimised for AMD or from the period of devs making an engine a fudge of DX11 and DX12, and this is also on Intel machines just like I mentioned with Civ 6.

And we could say the same about AMD as their DX11 performance is trending lower than expected on Ryzen as well in certain games, but again certain games so should we say then they also have a driver problem.

Now I agree part of it could be driver problem involved in all of this (or may help anyway) but the unusual behaviour is finding its way back to certain games while others perform extremely well for Nvidia on both platforms.
This seems a more complex problem involving multiple factors to the specific games that seem around DX11-DX12 fudging period (games mentioned so far) to possibly something the Nvidia hardware/software is not happy with in terms of the Ryzen platform exacerbating a specific situation.

But Civ6 performance is way beyond driver issues for Nvidia and comes back to possibly optimisation only for AMD at an engine/game level, a game lead by graphics engineers who were also involved in implementing Mantle with AMD's support into Civ engine back in 2014.
Click to expand...

SB isn't going to be a happy camper until he has an AMD CPU + GPU setup again. Maybe you've missed some other threads in the past few months...
 
I don't play like any of the games these reviews test. :) But it looks like there's not much reason to not buy Ryzen. I mean I'm happy here with 2500K and 3570K yet and it beats the snot out of those most of the time. Granted most people can't help being all hip and only judge it against the latest Intel stuff.

I feel like one really needs some weird corner case to justify an upgrade from even a Sandy Bridge system. Something like an absolute need for 100 fps in the latest games.

What I wonder about is when the Ryzen APUs will come. AMD isn't going to be getting much OEM volume until they have that CPU + GPU integration. Better power characteristics for mobile, cheaper than a discete GPU, etc.
 
Last edited:
swaaye said:
SB isn't going to be a happy camper until he has an AMD CPU + GPU setup again. Maybe you've missed some other threads in the past few months...
Click to expand...

Other than browser issues (which are a major pain) I've had less to complain about recently. The issues in Warframe and GW2 appear to have been mostly addressed in the latest drivers for the most part.

That said, the browser issues (can't go more than a few days before NVidia's driver causes Edge to just close or Chrome to complain about lack of memory, doesn't happen on the 290) and overall sluggishness on the 1070 compared to the 290 when memory starts to get full (multiple apps, games, browser windows, explorer windows, etc. open) are infuriating at times.

Game performance however has been stellar, until the CPU hits 100%, at which point the 290 doesn't get nearly as sluggish as the 1070 does in Windows. So on the 290, I can keep my idle games running while I open up something like GW2, Warframe, Xcom 2, etc. But on the 1070, I have to close some of those idle games (losing progression bonuses) in order to not have performance tank. Annoying, but not deal breaking.

I'm perfectly happy with an Intel CPU, however. It's just that Intel's 8 core CPU happens to be quite significantly more expensive than Ryzen. I've been wanting to move up to an 8 core CPU for a while as I tend to have a lot of applications open both for work and leisure. This would also likely alleviate some of the issues I'm having the with 1070. I was actually looking forward to pairing Ryzen 7 with the 1070 to see how things went. I'm now waiting for NVidia to fix their drivers before upgrading. Either that or waiting for Vega to launch before upgrading. Whichever comes first (I'm putting my money on Nvidia fixing their drivers first).

You see me as wanting an AMD GPU and CPU. I see it as wanting to have some choices. I'd like to be able to choose between AMD and Intel CPUs, but haven't been able to do due to AMDs CPUs sucking big time for quite a while. I'd like to be able to choose between AMD and NVidia GPUs, but haven't really been able to as AMD just haven't been competitive this generation.

Regards,
SB
 
Last edited:
CSI PC said:
You miss the aspects that raises pointers beyond Ryzen just like my comment about Civ 6.
You do not see anything wrong with a 480 beating the GTX1080 on Intel 7600k?
By that reasoning we might as well say AMD has DX11 driver problems because their performance tanks with Fallout 4, when in reality it comes back to the game-rendering engine design not ideal for said hardware.

But the drops are in those games I mentioned and not others (unless one deliberately sets the resolution to 720p), all the games have a similar trend in terms of Civ 6/Deus Ex Mankind/Hitman/Total War Warhammer development.
This is happening in those games for both DX11 and DX12 in relative terms on both machines (albeit further exacerbated on Ryzen for Nvidia).
Yet others scale very well with Nvidia on Ryzen such as AoTS (probably one of the better developed games around DX12), Sniper Elite 4, Gears of War 4 (apart from the 720p behaviour), and quite a lot of DX11 and a few of the more recent DX12 games work well with Nvidia on Ryzen.
Like I said those games with weird performance for Nvidia on DX11/DX12 also tend to be the ones optimised for AMD or from the period of devs making an engine a fudge of DX11 and DX12, and this is also on Intel machines just like I mentioned with Civ 6.

And we could say the same about AMD as their DX11 performance is trending lower than expected on Ryzen as well in certain games, but again certain games so should we say then they also have a driver problem.

Now I agree part of it could be driver problem involved in all of this (or may help anyway) but the unusual behaviour is finding its way back to certain games while others perform extremely well for Nvidia on both platforms.
This seems a more complex problem involving multiple factors to the specific games that seem around DX11-DX12 fudging period (games mentioned so far) to possibly something the Nvidia hardware/software is not happy with in terms of the Ryzen platform exacerbating a specific situation.

But Civ6 performance is way beyond driver issues for Nvidia and comes back to possibly optimisation only for AMD at an engine/game level, a game lead by graphics engineers who were also involved in implementing Mantle with AMD's support into Civ engine back in 2014.
Click to expand...

And yet the issues ONLY come up on Ryzen. With Intel CPUs performance is as expected. However, with Ryzen on NVidia drivers there are absolutely massive performance drop-offs compared to Intel systems in some games.

AOTS was a good example where the engine wasn't optimal for Ryzen as it affected performance on both NVidia and AMD GPUs. They patched it and now both AMD and NVidia GPUs operate as they should.

On Rocket League (dx9), Civ6 (dx12), and ROTR (dx12) the AMD GPUs perform as expected relative to their performance on Intel CPUs. For NVidia, performance drops massively on Ryzen when the game is CPU limited but performs as expected when the game isn't CPU limited. That's completely not like the AOTS situation where both AMD and NVidia GPUs were affected.

Regards,
SB
 
Silent_Buddha said:
You see me as wanting an AMD GPU and CPU. I see it as wanting to have some choices. I'd like to be able to choose between AMD and Intel CPUs, but haven't been able to do due to AMDs CPUs sucking big time for quite a while. I'd like to be able to choose between AMD and NVidia GPUs, but haven't really been able to as AMD just haven't been competitive this generation.
Click to expand...
If Vega is impressive, I might buy one just to have an AMD GPU around again. The newest AMD hardware I have is a long-abandoned-by-AMD unlocked 6950. I had a 290X for a bit years ago but it was just too hot and noisy for my liking. It uses something like 100W more power than a 6970.

I see the appeal in a 6-8 core Ryzen setup, though I'm not sure I really see the need for myself. I just don't feel any performance problems with any new games I've played since I got a 1070.
 
Silent_Buddha said:
And yet the issues ONLY come up on Ryzen. With Intel CPUs performance is as expected. However, with Ryzen on NVidia drivers there are absolutely massive performance drop-offs compared to Intel systems in some games.

AOTS was a good example where the engine wasn't optimal for Ryzen as it affected performance on both NVidia and AMD GPUs. They patched it and now both AMD and NVidia GPUs operate as they should.

On Rocket League (dx9), Civ6 (dx12), and ROTR (dx12) the AMD GPUs perform as expected relative to their performance on Intel CPUs. For NVidia, performance drops massively on Ryzen when the game is CPU limited but performs as expected when the game isn't CPU limited. That's completely not like the AOTS situation where both AMD and NVidia GPUs were affected.

Regards,
SB
Click to expand...
It is a mix, Ryzen just shows it more clearly so further exacerbates it.
Please answer this.
If it was just Ryzen, why does Civ 6 also have performance issues on Intel for Nvidia?
A custom MSI GTX1080 Gaming X vs Sapphire 480 Nitro, both on the 7600K at 1080p resolution.
GTX1080: 48.12 fps
480: 50.01 fps

That is a good custom GTX1080 (an enthusiast tier performance card) against an ok custom 480 (a mainstream tier performance card).
So yeah it is more game related in that instance and that it stands out more on Ryzen with these very specific games (Hitman/Deus Ex: Mankind/Warhammer), the situation is just more exacerbated with Ryzen but there is an underlying issue on both platforms for the Nvidia hardware and specifically these games comes back to what I mentioned earlier.

TBH Anandtech really needs to revisit the games they intend to use in the new layout going forward (Rocket League is not a great choice just like some of the MOBA that are optimised towards Nvidia), remember people were unhappy when Fallout 4 being used by publications as is it weighted heavily for Nvidia due to engine design while Quantum Break DX12 was a similar situation reversed, while RoTR is a maniac and can show various results depending upon map-scene tested.
But there are decent games out there that are more neutral and should be the more recent games that are known to be well developed DX11 or DX12, such games should make up at least half the list if also using weighted titles.

I know some other publications are re-evaluating the games to use when considering Intel and Ryzen with GPUs these days, especially with how DX12 games are starting to improve as well.
Cheers
 
Last edited:
CSI PC said:
It is a mix, Ryzen just shows it more clearly so further exacerbates it.
Click to expand...
I think the secret lies in the CPU limitation, whenever that kicks in, Ryzen performance stumbles on NV compared to Intel. Using Anand's numbers these points are noticeable:

-Civilization 6 @1080p has the 1060 and 1080 scoring the same fps, the same for the 480 and Fury. This is a text book CPU limitation situation. Here the AMD cards perform a little better on Ryzen than NV cards.

-Tomb Raider @1080p, has the same thing basically. On Ryzen, the Fury is barely faster than 480, the 1080 is barely faster than 1060 on Ryzen. Another obvious situation of CPU bottleneck. Switch to Intel and the 1080 is now 54% faster.

Another interesting observation here is the i5 7400, a 3.0GHz 4 core CPU, the weakest of the bunch and the most CPU limited, however the RX 480 scores the best on that processor by 12%! Could be an anomaly or a mistake. But if not, it fits well into the narrative of CPU limitation exposing the problem.

86412.png


-Rocket League exposes this further @1080p DX9: The 1080 is barely any faster than Fury and 480. The Fury is actually slightly slower than 480. Another CPU limitation story. The 1080 pulls far ahead only @4K and Intel. On Ryzen even @4K, it is still slower than Fury and barely faster than 480.
 
I am not sure if this is the right thread but this is related to Ryzen.
Sorry for posting a video but this one is really worth looking at (disregard the clickbait title).
I have to say i agree with Adore's conclusions. This seems like a big blunder on Tom's part. And a prime case of 1 outlier skewing results heavily.
 
For me the important thing about this is that in reddit when they accuse them to be Intel biased they wrote a PR BS basically saying that their sells team do not talk with their reviews team...when they only needed to use arguments and evidence to back up their advises.

And yes I don't see a single reason to buy a 7500 at the current price...but anyways lets assume tom's just crew up and not anything else that could be actually really serious and concerning.
 
DeF said:
I am not sure if this is the right thread but this is related to Ryzen.
Sorry for posting a video but this one is really worth looking at (disregard the clickbait title).
I have to say i agree with Adore's conclusions. This seems like a big blunder on Tom's part. And a prime case of 1 outlier skewing results heavily.
Click to expand...

I love the part with the statistics. I think reviewers should (if available) provide actual probability density functions/graphs for overclocking. And if not available, put a disclaimer like "Top Secret" , maybe "Glassball", diagonally over the whole review. Just half kidding. ;)
 
DeF said:
I am not sure if this is the right thread but this is related to Ryzen.
Sorry for posting a video but this one is really worth looking at (disregard the clickbait title).
I have to say i agree with Adore's conclusions. This seems like a big blunder on Tom's part. And a prime case of 1 outlier skewing results heavily.
Click to expand...
Too add in top of that:


Funny how they pointed out how tom's contradict itself about best performance and then about best value...
 
You must log in or register to reply here.

Similar threads

A
AMD: Zen 3 Announced [2020-10-08]
10 11 12
Replies
220
Views
91K
davis.anthony
D
D
OG XBox was planned to launch with an AMD CPU until last minute.
4 5 6
Replies
112
Views
12K
Nesh
N
D
Intel Launches Comet Lake (Core 10th Series)
2
Replies
25
Views
28K
Deleted member 7537
D
G
AMD 4000 series APU for mobile
Replies
4
Views
6K
mczak
M
F
AMD RyZen Threadripper 2 CPU Reviews [2018]
Replies
4
Views
5K
Alexko
A
Back
Top