[Various] AMD Ryzen Reviews (Official)

[StyM]

XS (chew)
Ryzen - Return of the Jedi

Youtube
Ryzen Tech Talk
The AMD Ryzen™ Processor Ecosystem
Overclocking the AMD Ryzen™ Processor
AMD SenseMI Technology – Pure Power
AMD SenseMI Technology – Smart Prefetch

AMD RYZEN 7 REVIEW... WE DROP IT
Helium - AMD Ryzen Gaming PC Build (2017)
AMD Ryzen R7 1800X Review: An i5 in Gaming, i7 in Production
#0156 - Ryzen 1800x vs 5960x vs 7700k in the ultimate showdown
AMD Ryzen 7 1800X & 1700X Review: Live Up to The Hype?
The AMD Ryzen 7 1800X Review: Now and Zen
ZEN BENCHMARKS! Ryzen 7 1800X Review vs 6850K, 7700K & FX-8350
AMD Ryzen 7 1800X CPU Review
FIRST OFFICIAL Ryzen 7 1800X Benchmarks! Is AMD BACK?
AMD Ryzen 7 1800X Review - Finally, Competition!
Ryzen 7 1700X: The new sweet spot CPU?
AMD Ryzen 7 - Launch Review
RYZEN 1800X vs INTEL 6900K (1700X vs 6800K)

AnandTech
The AMD Zen and Ryzen 7 Review: A Deep Dive on 1800X, 1700X and 1700

HardwareCanucks
The AMD Ryzen 7 1800X Performance Review

Guru3D
A New Era Has Ryzen
AMD Ryzen 7 1700X Review

KitGuru
AMD Ryzen 7 1800X CPU Review

Hexus
Review: AMD Ryzen 7 1800X (14nm Zen)
Review: AMD Ryzen 7 1700X (14nm Zen)

PcPer
The AMD Ryzen 7 1800X Review: Now and Zen

ArsTechnica
AMD Ryzen 7 1800X review: Good, but not for gamers

Bit-tech
AMD Ryzen 7 1800X and AM4 Platform Review

TweakTown
AMD Ryzen 7 1800X CPU Review - Intel Battle Ready?

HardwareZone
AMD Ryzen 7 1800X vs. Intel Core i7-7700K: Next-gen flagship CPU matchup!

OC3D
AMD Ryzen 7 1800X CPU Review

TomsHardware
AMD Ryzen 7 1800X CPU Review

GamersNexus
AMD Ryzen R7 1800X Review: An i5 in Gaming, i7 in Production

Legit Reviews
AMD Ryzen 7 1800X, 1700X and 1700 Processor Review

TechSpot
AMD Ryzen Review: Ryzen 7 1800X & 1700X Put to the Test

Phoronix
AMD Ryzen 7 1800X Linux Benchmarks

eTeknix
AMD Ryzen 7 1800X AM4 8-Core Processor Review

ServeTheHome
AMD RYZEN 7 1700X LINUX BENCHMARKS

HardOCP
AMD Ryzen 1700X CPU Review

HotHardware
AMD Ryzen 7 1800X, 1700X, And 1700 Review And Benchmarks: Zen Brings The Fight Back To Intel

[FlanK3r]

thanks for updates!

[Ket]

I was right on the money opting for a 1700 to review, I knew nobody would be that interested in it just because it didn't have a "X" on the end of it Gigabyte Aorus Gaming 5 is still delayed because of Gigabyte falling behind (thanks for that Giga ) but maybe thats a blessing in disguise, gives them a bit more time to churn out a extra UEFI or two prior to me starting to review it.

Now I can say something though, (just seen its 5pm here ) I will say at lower resolutions for gaming Ryzen isn't so hot, but I'm confident thats because of memory latency which should be addressed as UEFIs mature.

[AliG]

Seems about right. I found it suspicious that AMD only released benchmark data for a small set of tests, and now these confirm Ryzen has tons of raw compute - but the instruction logic needs to be refined for edge cases.

Curious to see how many people still keep their pre-orders after seeing this (I know of a few who already cancelled due to the game performance).

[Dimitriman]

I find the lower gaming performance weird compared to the rest of the tests. But in essence I see two problems: bad overclocking and too low clocks to compete in gaming. Both of these can be ironed out with new revisions and better process tweaking. This is a brand new arch and I remember even the old Athlons were not stellar at launch. This needs some time and AMD needs to get their quad/hex cores well beyond 4.0 Ghz to compete will with 7700k.

[Sparky]

Maybe it will help me feel better about not being able to upgrade right away anyway lol

[AliG]

I find the lower gaming performance weird compared to the rest of the tests. But in essence I see two problems: bad overclocking and too low clocks to compete in gaming. Both of these can be ironed out with new revisions and better process tweaking. This is a brand new arch and I remember even the old Athlons were not stellar at launch. This needs some time and AMD needs to get their quad/hex cores well beyond 4.0 Ghz to compete will with 7700k.

I think the gaming performance has a lot more to do with the memory latency and lack of bandwidth. Probably doesn't help too much that L1 cache isn't stellar either (though their L2 is rather impressive).

[Ket]

Seems about right. I found it suspicious that AMD only released benchmark data for a small set of tests, and now these confirm Ryzen has tons of raw compute - but the instruction logic needs to be refined for edge cases.

Curious to see how many people still keep their pre-orders after seeing this (I know of a few who already cancelled due to the game performance).

I'll be keeping my Zen system, I'm confident with UEFI maturity (lets be honest current UEFIs are still in beta stages) and some old fashioned thermal improvements to key areas on mainboards that 1700s and the 1800X will see around 4.3GHz which is very impressive for a Octacore. I'd still even hedge my bets on slightly higher clocks than that albeit with some actual tweaking on the manufacturing side of Zen. I've already got some specialist 10W/m-K thermal pads on their way so if overheating board components are a problem (they will be, stock pads are always crap) there won't be with the pads I'm going to use. I might even go insanity overkill and get some 1750W/m-K pads, those things are so good you basically don't need heat pipes or fans, assuming the heatsink design is at least competent and good quality.

[Metroid]

I was right on the money opting for a 1700 to review, I knew nobody would be that interested in it just because it didn't have a "X" on the end of it

I was also in the 1700 groupie since the very beginning hehe

I find the lower gaming performance weird compared to the rest of the tests. But in essence I see two problems: bad overclocking and too low clocks to compete in gaming. Both of these can be ironed out with new revisions and better process tweaking. This is a brand new arch and I remember even the old Athlons were not stellar at launch. This needs some time and AMD needs to get their quad/hex cores well beyond 4.0 Ghz to compete will with 7700k.

I'm very certain that there are many things to improve, Ryzen 7 was kind rushed.

[chew*]

For the 6 core crowd the cheaper 65w is likely a good candidate as well.


Some of the lower game performance can be attributed possibly to what i said. better performance fixed vs dyanamic speeds. But i dont think that will help all games. Some like raw speed.

[Metroid]

https://www.pcper.com/reviews/Proces...ll-E-Kaby-Lake

a good clock for / per clock comparison, 1800x x 6900k both at 3.5ghz.

Well for gaming, Ryzen 1800x is not as good as 6900k but for almost all the others apps, Ryzen is better at fixed 3.5ghz, given the price, its a no brainer, so Ryzen 7 is the clear winner, now if Intel equalize 6900k price to x1800 then Intel wins the battle, so far Ryzen is much better overall. So my point is, for gaming buy a i7 7700, for everything else buy a Ryzen 7 1700.

[chew*]

Id like to see game tests done with core only on ryzen.

Smt off. If you game 16t is pointless. Thats exactly how i plan to use it.

I have no clue if it makes a difference but the chip 4c8t is not as fast as it is in 8c mode in MT.

I have no clue if it will gain any clock headroom either like that. Ran out of time.

There are many variables and ways to use it however.

[NEOAethyr]

On intel most of the time smt/ht is just fine, in some instances in linux it can hold you back a little bit, it's generally just scheduling issues.
I would use it with 16threads dude...
The more the better...
Gaming wise, 8t's are fine but if you wanna dl or something while gaming, even communicating in multiplayer, it could help a little.
I don't think it's generally gonna cut your perf in half.
If you wanna check that I would recommend the old call of duty's, they are all single threaded games.
Found that out big time when messing with 3d vision.
If you got say mw2 or even mw3, check with afterburner the cpu stuff and fps with smt on/off.
I'm not sure what other games I would recommend for single thread testing, mafia2 would be one but I don't know if it uses more then one core or not.
Farcry obviously not one to check with, but why not I guess.
Just check with whatever games you have and report back on what you eventually do .

[Metroid]

Smt off. If you game 16t is pointless.

Yeah smt is pointless, so to rephrase, dont buy a 7700, buy a 7600k for gaming.

I'm trying to find a review with smt off, still no luck.

[xlink]

http://hexus.net/tech/reviews/cpu/10...nm-zen/?page=9

However, and somewhat interesting to note, switching off the chip's SMT capability increased the average frame rate from 79fps to 85.8fps, suggesting that code is not running efficiently when there's SMT involved. Hopefully this problem will be fixed by a game-patch update.

Going back to SMT, switching it off also increases the Hitman score, from 91.4fps to 95.6fps, suggesting, once again, that having it active is definitely hindering performance. In fact, running Ryzen in non-SMT mode offers more performance in every scenario, and this is something that AMD needs to be concerned about.

There are some bugs still.

Regardless, this is a movement in the RIGHT direction. AMD is no longer "embarrassing" from a gaming perspective and is VERY competitive in other domains.
As threading gets better, and we're starting to get to that point, it'll do better and better in newer games relative to the 7700ks of the world. At the same time, older games run well enough and even if there isn't heavy threading it's "good enough" such that it isn't outclassed by a dual core i3.

Also all the AMD fanboys can now safely say bulldozer sucks.

I'm excited. I might wait a little bit before pulling the trigger though.

[AliG]

Also all the AMD fanboys can now safely say bulldozer sucks.

lol that made my day

[Metroid]

http://hexus.net/tech/reviews/cpu/10...nm-zen/?page=9



There are some bugs still.

Regardless, this is a movement in the RIGHT direction. AMD is no longer "embarrassing" from a gaming perspective and is VERY competitive in other domains.
As threading gets better, and we're starting to get to that point, it'll do better and better in newer games relative to the 7700ks of the world. At the same time, older games run well enough and even if there isn't heavy threading it's "good enough" such that it isn't outclassed by a dual core i3.

Also all the AMD fanboys can now safely say bulldozer sucks.

I'm excited. I might wait a little bit before pulling the trigger though.

Thanks, in the Intel side it was the same on my old i7 920, smt off gave a performance boost in games and single threads apps. Not sure if that issue still on intel cpus after i7 920 as I did not test. I guess I will have to conduct few tests with the Ryzen 7 1700 concerning that. I mean to have smt turned off --> 8 cores as in 8 cores is amazing if given performance improves at least 5%.

[NEOAethyr]

@Metroid
Yeah it has a small effect on newer intel's, I believe it to be mostly a os scheduling issue.
Take windows for ex., there's times were it will lock on a single task, and make you wait on window handles and other interrupt types.
Been that way since we had dual cores, if you guys want better smt perf, it's not necessarily the games fault, but most likely the os.

It's something I've been watching for the last year on linux, it's still a mess really.
Windows doesn't generally bounce the threads around as much (losing context and cache by doing so).
But there's definitely times where it screws up because it's not a true multi tasking os.

5% is normal in my opinion, because that's probably about around what I've seen on my ivy-e.
What's more worth it?, %5 or 45%... (45% going for stuff like cinebench with smt on)

One issue on my intel though is if you hammer the crap out of the cpu, all threads, it'll stutter everything.
That was in windows though, haven't messed with that in linux, I generally try to stay around 80% max or so.
I noticed the other day my cpu usage wasn't that high when gaming so it's all good (I need a new vga card though so who knows).

[chew*]

Nope shutting smt off lowered tdp. They were using a stock setting right? Well lower tdp the chip will turbo up just fine all cores.

This is why i ignored all that turbo xfr dyanamic tdp and ran at a fixxed speed for testing.

But yes ipc in mt tasks when i compared cinebench 8c beat 4c/8t. I expected that.

No clue if carries over to games@ fixxed speeds but one can hope.

[Metroid]

@Metroid
Yeah it has a small effect on newer intel's, I believe it to be mostly a os scheduling issue.
Take windows for ex., there's times were it will lock on a single task, and make you wait on window handles and other interrupt types.
Been that way since we had dual cores, if you guys want better smt perf, it's not necessarily the games fault, but most likely the os.

It's something I've been watching for the last year on linux, it's still a mess really.
Windows doesn't generally bounce the threads around as much (losing context and cache by doing so).
But there's definitely times where it screws up because it's not a true multi tasking os.

5% is normal in my opinion, because that's probably about around what I've seen on my ivy-e.
What's more worth it?, %5 or 45%... (45% going for stuff like cinebench with smt on)

One issue on my intel though is if you hammer the crap out of the cpu, all threads, it'll stutter everything.
That was in windows though, haven't messed with that in linux, I generally try to stay around 80% max or so.
I noticed the other day my cpu usage wasn't that high when gaming so it's all good (I need a new vga card though so who knows).

Thanks for the info, when I conducted the tests, performance while smt was off was as much as 35% on Windows XP and as much as 8% on Windows 7 in 2009, Windows 10 was not around so probably the performance constraints might be less on Windows 10. Performance constraints was huge in the Windows XP as Windows XP was never really built for multi-threading. It supported and that was that.

This is why i ignored all that turbo xfr dyanamic tdp and ran at a fixxed speed for testing.

But yes ipc in mt tasks when i compared cinebench 8c beat 4c/8t. I expected that.

Exatcly, somehow is efficient, smt x the cpu raw cores is inefficient, however it gives perf boost with a lot more heating, for me the extra heating never justified the perf, Ryzen 7 1700 might overclock better with smt off. I wonder if somebody tested the overclock with smt off, in theory, less heat, more headroom for overclocking. Well that was what AMD stated about Ryzen. Ryzen 1700 with 8 cores, TDP at 65 watts with smt on is really incredible.

[KiSUAN]

http://www.hardware.fr/articles/956-...ct-smt-ht.html

[StAndrew]

Seems about right. I found it suspicious that AMD only released benchmark data for a small set of tests, and now these confirm Ryzen has tons of raw compute - but the instruction logic needs to be refined for edge cases.

Curious to see how many people still keep their pre-orders after seeing this (I know of a few who already cancelled due to the game performance).

I wouldn't cancel my processor pre-order but I would cancel my moboard. I'm sure there will be much better offerings in the not too distant future. However, on the Anandtech review (IIRC) AMD already mentioned they are working on some short fall in the design and are coming out with a Re-Ryzen CPU... Yeah, I'll wait

[xlink]

I wouldn't cancel my processor pre-order but I would cancel my moboard. I'm sure there will be much better offerings in the not too distant future. However, on the Anandtech review (IIRC) AMD already mentioned they are working on some short fall in the design and are coming out with a Re-Ryzen CPU... Yeah, I'll wait

Saying that they're working on the next iteration of Ryzen isn't exactly surprising... that IS their main line of business.

[Metroid]

http://www.hardware.fr/articles/956-...ct-smt-ht.html

Thanks, now all fits well. Ryzen team has something to fix. Intel 6900k arch is a lot more efficient with smt on.

I wouldn't cancel my processor pre-order but I would cancel my moboard. I'm sure there will be much better offerings in the not too distant future. However, on the Anandtech review (IIRC) AMD already mentioned they are working on some short fall in the design and are coming out with a Re-Ryzen CPU... Yeah, I'll wait

I don't mind waiting as well, this performance degradation while smt is active/on should be top priority to be fixed on next Ryzen CPU's.

[StAndrew]

Saying that they're working on the next iteration of Ryzen isn't exactly surprising... that IS their main line of business.

But to admit there are some pretty critical improvements already identified and being worked on is pretty significant considering they mentioned this before NDA was even lifted...

Yes its too be expected given this is a very new architecture.