Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[SPBHM]

A bit tongue in cheek video, still very amusing

I know this is just a joke but, I would be curious to see the result running at a higher resolution, if I'm not mistaken a Core 2 Duo gets less than 60fps running Q2 in software mode 1280x960.

 

[unseenmorbidity]

I know this is just a joke but, I would be curious to see the result running at a higher resolution, if I'm not mistaken a Core 2 Duo gets less than 60fps running Q2 in software mode 1280x960.

I believe it is capped at 1000 fps.

 

[SPBHM]

I believe it is capped at 1000 fps.

and it hits that cap at 320x240, but at 1280x960 or whatever is the maximum resolution for software rendering I don't think it's going to hit the cap for most of the run.

 

[looncraz]

Well, that's maddening... some idiot forgot to rename the FRAPS benchmark files while doing bench marking runs for frame times in Windows 10 and now has no idea what ~5 hours of gaming results go for which configuration...

I guess I will have to play games all day tomorrow again, after I finish Windows 7 testing... and remember to rename the benchmark files!!!!!!!!! Thankfully, I had the wherewithal to backup the Windows 10 installation before moving to Windows 7.

 

[cytg111]

Seriously? And w

interesting times ahead.

Do you know who's been silent in all these though? Intel. Where are the price cuts already? Hehe. I wouldn't want to play a poker game against Intel. I mean, no reaction at all? Damn!

Its the smart move, no need to lower prices wihout dents in the earnings.

The people that bought core 2 quad vs core dual core had nearly 2 years extra gaming on their cpu.
2500 vs 2600 show similar pattern though the difference is less because 8t is far less uplift vs real cores.
1600x obviously will look a good deal better than 1700 especially stock at gaming. Those graphs will be all over.
But changing cpu is imo not the best upgrade path. Easy yes.
The price difference for 6c vs 8c is so small people should do some real thinking before forking out for a 6c when a game like bf1 already issues 10 threads. If an 8c can keep you going just a year extra its peanuts vs total cost when you build a new pc.

An 1700 system is 80 usd cheaper than a 7700. Imo its a cheap way to play for perhaps 8 years and upgrade gpu 3 times.

+1. And do the total cost of a ownership of a system, what is r5 -> r7 really? another 2%? 5%? Thats why i never skip on the CPU when building new.

 

[cytg111]

and it hits that cap at 320x240, but at 1280x960 or whatever is the maximum resolution for software rendering I don't think it's going to hit the cap for most of the run.

So while Zen should easily be able to hit 1000, why doesnt it? The horsepower is surely there.. what is holding back 5%?

 

[PotatoWithEarsOnSide]

Well, that's maddening... some idiot forgot to rename the FRAPS benchmark files while doing bench marking runs for frame times in Windows 10 and now has no idea what ~5 hours of gaming results go for which configuration...

I guess I will have to play games all day tomorrow again, after I finish Windows 7 testing... and remember to rename the benchmark files!!!!!!!!! Thankfully, I had the wherewithal to backup the Windows 10 installation before moving to Windows 7.

Can't you just check the times that they were created?

 

[looncraz]

Can't you just check the times that they were created?

Sure, if I knew which configuration I tested at those times I've tested so many configurations and settings... and I only have a handful of screenshots that I took with the settings.

 

[lolfail9001]

Sure, if I knew which configuration I tested at those times

Well, self-deprecating is appropriate indeed.

 

[lolfail9001]

Canned Linux gaming benches in 4k (and some in 1080p): http://www.phoronix.com/scan.php?page=article&item=nvidia-1080ti-ryzen&num=1

Spoiler: they contain 1 very nasty surprise. Or 2 nasty surprises, if you think that Win10 scheduler is to blame.

EDIT: I must note that i have no clue if Win10 RTC issue applies to Linux, and i am actually out of clues to explain the very nasty outlier.

 

[leoneazzurro]

Canned Linux gaming benches in 4k (and some in 1080p): http://www.phoronix.com/scan.php?page=article&item=nvidia-1080ti-ryzen&num=1

Spoiler: they contain 1 very nasty surprise. Or 2 nasty surprises, if you think that Win10 scheduler is to blame.

EDIT: I must note that i have no clue if Win10 RTC issue applies to Linux, and i am actually out of clues to explain the very nasty outlier.

In the review they say that on Linux practically no game scales over 4 cores. And I think the state of optimization is something to debate, because we have also an outlier with Ryzen having double performance than 7700K

 

[dfk7677]

Canned Linux gaming benches in 4k (and some in 1080p): http://www.phoronix.com/scan.php?page=article&item=nvidia-1080ti-ryzen&num=1

Spoiler: they contain 1 very nasty surprise. Or 2 nasty surprises, if you think that Win10 scheduler is to blame.

EDIT: I must note that i have no clue if Win10 RTC issue applies to Linux, and i am actually out of clues to explain the very nasty outlier.

I really don't think these can really be used. Look at the disparity of the Tomb Raider bench.

 

[lolfail9001]

In the review they say that on Linux practically no ame scales over 4 cores.

That's a fact, since absolute majority of games are either too old to be threaded well or are ports of games from Windows using a DX wrapper, that works as a very heavy bottleneck on it's own. The best threaded games is like Dota 2 and Metro LL (both scaling to 5 cores relatively well), that's it.

And I think the state of optimization is something to debate, because we have also an outlier with Ryzen having double performance than 7700K

State of optimization is not up to debate, it is downright terrible. As for the nasty outlier, i think we have either stumbled upon bug in software Michael uses to automate testing, or a bug in Linux clock measurements as well. Sadly, he is not the sort of guy to make an investigation when something goes wrong.

I really don't think these can really be used. Look at the disparity of the Tomb Raider bench.

Tomb Raider is the only outlier here, though.

 

[piesquared]

What does Mafia III do differently that other games can't?

http://www.techspot.com/review/1348-amd-ryzen-gaming-performance/page2.html


woahh a little big

 

[leoneazzurro]

When you have four samples and 3 outliers, I think the sample is far from being representative.

 

[looncraz]

Canned Linux gaming benches in 4k (and some in 1080p): http://www.phoronix.com/scan.php?page=article&item=nvidia-1080ti-ryzen&num=1

Spoiler: they contain 1 very nasty surprise. Or 2 nasty surprises, if you think that Win10 scheduler is to blame.

EDIT: I must note that i have no clue if Win10 RTC issue applies to Linux, and i am actually out of clues to explain the very nasty outlier.

Linux testing needs to be done with the latest kernel - even 4.9 is too old.

I plan on compiling Linux from the ground up with Ryzen optimizations maxed out sometime this next week. I was hoping to have my board before then... no way I want to be doing that while sitting in my wife's chair... not with my back.

 

[PotatoWithEarsOnSide]

That's the article that had the 1800x as within 2.5% of the 6900k if you took the best result regardless of SMT on/off, though the article itself claims 12%...on dodgy math that no-one has been able to explain since I first brought it up.

 

[piesquared]

That's the article that had the 1800x as within 2.5% of the 6900k if you took the best result regardless of SMT on/off, though the article itself claims 12%...on dodgy math that no-one has been able to explain since I first brought it up.

Other sites are also showing similar results for Mafia III. Whatever this game is doing, other games should be able to do the same. Is there some magic voodoo going on? Has this game already been optimized for Ryzen architecture?

 

[lolfail9001]

Linux testing needs to be done with the latest kernel - even 4.9 is too old.

Yeah, the llc patch landed few weeks after 4.9 merge window closed. I guess that explains the visible inconsistency of some Ryzen results there.

 

[DisEnchantment]

Linux testing needs to be done with the latest kernel - even 4.9 is too old.

I plan on compiling Linux from the ground up with Ryzen optimizations maxed out sometime this next week. I was hoping to have my board before then... no way I want to be doing that while sitting in my wife's chair... not with my back.

Please if possible could you check if DCA is impacted on Ryzen Platform.

 

[looncraz]

Please if possible could you check if DCA is impacted on Ryzen Platform.

My Linux testing will be next week at the earliest. Just doing Windows testing is a pain because I have to test everything in so many configurations. I've had to limit how many because of all the inconsistencies... I even wrote an app to test the hardware clock accuracy (in Linux) and call cost to try and see why some of these variations are happening (good news: they're basically perfect - 100% match to Intel specifications).

Linux Mint 18.1, interestingly, sees Ryzen as having only 4MB of cache with one CCX enabled. I don't think DCA (direct cache access, correct?) is enabled on the live boot image. We'll see what happens when I build fresh from Gentoo with every Ryzen optimization in place (including march=zenver1).

 

[imported_jjj]

1700 review at stock and 4.05GHz with a Radeon RX 480 8GB https://www.vortez.net/articles_pages/amd_ryzen_1700_review,1.html

 

[DisEnchantment]

Linux Mint 18.1, interestingly, sees Ryzen as having only 4MB of cache with one CCX enabled. I don't think DCA (direct cache access, correct?) is enabled on the live boot image. We'll see what happens when I build fresh from Gentoo with every Ryzen optimization in place (including march=zenver1).

Yes, Direct Cache Access. I have R7 running Linux at home, but it is in a VM so not ideal to test. But at work its an Intel Machine..

 

[PotatoWithEarsOnSide]

Canned Linux gaming benches in 4k (and some in 1080p): http://www.phoronix.com/scan.php?page=article&item=nvidia-1080ti-ryzen&num=1

Spoiler: they contain 1 very nasty surprise. Or 2 nasty surprises, if you think that Win10 scheduler is to blame.

EDIT: I must note that i have no clue if Win10 RTC issue applies to Linux, and i am actually out of clues to explain the very nasty outlier.

Sorry to come back to this, but why did he test with kernel 4.9 when he'd already tested 4.10 and 4.11 a week or so ago? Yes, new GPU, but still doesn't shed any light as to why he'd gimp Ryzen for the 1080ti.

 

[beginner99]

1700 review at stock and 4.05GHz with a Radeon RX 480 8GB https://www.vortez.net/articles_pages/amd_ryzen_1700_review,1.html

These results are all over the place. In other reviews Ryzen sucked in exactly these 2 games while here it shines compared to say a 7700k. However these other reviews all seem to be done with a GTX1080. I think what would be great bench to compare CPU + GPU combo. Maybe the NV driver is causing an issue with Ryzen...