Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[Mopetar]

Yep, our #1 aim is to rigorously indoctrinate and brainwashed others until they are purify of all dissenting notions that Intel chips are superior. Welcome to our bizarre cult where conformity is absolute and discord is punishable by sticking the latest amd CPU up the offender's rear end.

Chill dude, we aren't what you think we are...

I never said you were and if you read what I wrote, you'd see it's quite the opposite. You don't get that many people around here who are hardcore mindless fans of anything because the forums are mostly general (there aren't separate Intel/AMD forums) and moderated. Maybe you occasionally see a few posters who just outright refuse to buy Intel (or brand X) for whatever reason, but they're pretty few and far between and aren't constantly pushing that opinion either.

The kinds of people that are just rabidly pro-something (or occasionally less in favor of something, but against something else) don't hang around here because most people won't constantly validate their beliefs or they have to deal with those beliefs being challenged.

 

[agouraki]

What does For Honor use so much CPU processing power for anyway? It's just a simple fighting game with pretty graphics slapped on.

the thing is,i really want to see benchmark on for honor with 720p this will prove that its a game/windows optimization thing.

 

[Mockingbird]

I have a feeling that there's a flaw in Zen microarchitecture and that's why it performs so poorly in games.

It's not a fatal one unlike the one in Bulldozer.

AMD knows what it is, but it won't be corrected until Zen+, or possibly as early as Raven Ridge.

In fact, the reason that quad-core Ryzen won't be release until H2 2017 is probably because it will be based on Raven Ridge.

Until then, AMD will make BS excuses such as: games haven't been optimized for Ryzen yet.

 

[piesquared]

I have a feeling that there's a flaw in Zen microarchitecture and that's why it performs so poorly in games.

It's not a fatal one unlike the one in Bulldozer.

AMD knows what it is, but it won't be corrected until Zen+, or possibly as early as Raven Ridge.

In fact, the reason that quad-core Ryzen won't be release until H2 2017 is probably because it will be based on Raven Ridge.

Until then, AMD will make BS excuses such as: games haven't been optimized for Ryzen yet.

Except some have received micro code updates that have boosted performance significantly. So i'm more inclined to believe that variances are due to teething issues.

 

[inf64]

I have a feeling that there's a flaw in Zen microarchitecture and that's why it performs so poorly in games.

Please stop posting nonsense. Zen is not performing "poorly" in games. It is on average 10-15% slower than 7700K, stock vs stock. You people make it sound like Zen is unable to run games at 30fps, it is incredible.

 

[Atari2600]

I have a feeling that there's a flaw in Zen microarchitecture and that's why it performs so poorly in games.

1. It doesn't run poorly. Only folks with a limited ability to interpret data think that.

2. A microarchitecture flaw that is specific to games? Do you think intel only packed in a 32bit GAME instruction per die as opposed to Intel's 4x 512b GAME instructions per core?

 

[moonbogg]

14nm Finfet LP is a low power transistor, the normal operative range is 2-3Ghz, beyond that and power and of course heat starts to spiral. One issue in Zen with gaming is that it says 4Ghz but it isn't really there, the power needed cannot be sustained and it has a throttle of sorts. The CPU is never functioning at 100% capacity.

I have reason to suspect that Zen + will not use LP transistors. It is easy to change, Zambezi went from SOI to Planar in Vishera so transistors are not locked to uARCH, AMD can change to 14nm FINFET High Power.

This is terrifying. Are you serious?

 

[R0H1T]

This is terrifying. Are you serious?

No, looks like something's lost in translation.

 

[french toast]

1. It doesn't run poorly. Only folks with a limited ability to interpret data think that.

2. A microarchitecture flaw that is specific to games? Do you think intel only packed in a 32bit GAME instruction per die as opposed to Intel's 4x 512b GAME instructions per core?

LOL.
Zen has a couple of severe bottlenecks i can see, it can only dispatch 2 on screen characters per core, but SMT enables "double pumping" of on screen charaters, but they must be half the size.
Intels superior uarch has 2x more throughput of on screen charaters, not 8 half size charaters as such, but a whopping 4 full size charaters per core.
Ryzens neural net predictor hower enables harder game modes, it knows exactly what the player will do next, using xfr to turbo the enemies reaction times.

Its tough pick as both GAME ipc are not comparable really. I think the higher throughput of intel is better for large multiplayer maps that i play.

 

[Mechanical Man]

I have a feeling that there's a flaw in Zen microarchitecture and that's why it performs so poorly in games.

It's not a fatal one unlike the one in Bulldozer.

AMD knows what it is, but it won't be corrected until Zen+, or possibly as early as Raven Ridge.

In fact, the reason that quad-core Ryzen won't be release until H2 2017 is probably because it will be based on Raven Ridge.

Until then, AMD will make BS excuses such as: games haven't been optimized for Ryzen yet.

I hope RR is done in TSMC 16FF+. That would remove the clock wall of Ryzens 4GHz most likely. Then, RR could actually be massively interesting offering.

 

[french toast]

I hope RR is done in TSMC 16FF+. That would remove the clock wall of Ryzens 4GHz most likely. Then, RR could actually be massively interesting offering.

Samsung 14nm LPU is what ive been speculating for months.

 

[unseenmorbidity]

I don't see them giving up the low power consumption in favor of higher clocks. The real money is in the server market, which favors that aspect. I foresee we will continue to get trimmed down server cpus.

 

[french toast]

I don't see them giving up the low power consumption in favor of higher clocks. The real money is in the server market, which favors that aspect. I foresee we will continue to get trimmed down server cpus.

Stilt hinted something, he might be messing around though.

 

[imported_jjj]

You guys need to dial down expectations for Raven Ridge in desktop.
It's a notebook first platform so it's not gonna be designed for 95W and high clocks.
And the socketed desktop SKUs could be late as it makes sense to focus on the core market first.- this is speculation so don't take it as a fact.

 

[Glo.]

Samsung 14nm LPU is what ive been speculating for months.

Is it any better than what we have seen to this day?

 

[Glo.]

Stilt hinted something, he might be messing around though.

Can you give any specific link? I do not follow everything what happens on this forum.

 

[moonbogg]

Except some have received micro code updates that have boosted performance significantly. So i'm more inclined to believe that variances are due to teething issues.

Who and where? I want to see it for real.

 

[french toast]

Can you give any specific link? I do not follow everything what happens on this forum.

Cant post links sorry, look in his technical ryzen thread last page, cant remember who suggested LPU, but heard about it last year and speculated zen+ might use it.
Samsung announced 14nm LPU as their 4th generation finfet process Q3 last year, details are light but it is thought it offers 10% higher performance, maybe at the expense of leakage, dont quote me on that.
Its a higher performance HP version i know that.

 

[formulav8]

In fact I used to be a fan of AMD, but the hype, whining and complaining, grade school level smack talk, and general obnoxiousness of the AMD contingent on the internet, instead of making me want to buy an AMD product, as is obviously their objective, has quite the opposite effect, and makes me want to never buy any AMD product, no matter how good it is.

You must Really hate NVidia then

 

[Puffnstuff]

It is on average 10-15% slower than 7700K, stock vs stock.

New architecture vs old should have the new showing the greatest performance in every way. Granted Ryzen is far better than its predecessors but it still lags Intel in most cases. When 4 Ryzen cores are pitted against an equal amount of Haswell or Skylake cores at the same clocks it loses which should not be the case for a superior performing chip. While I am happy that AMD has made a tremendous improvement to their CPU its still not the jump I was led to believe it was based upon their hype. I'm hoping to see additional improvements to it as time goes on that will increase L1 cache performance, raise throughput and IPC to match or beat Intel.

 

[looncraz]

Please stop posting nonsense. Zen is not performing "poorly" in games. It is on average 10-15% slower than 7700K, stock vs stock. You people make it sound like Zen is unable to run games at 30fps, it is incredible.

THANK YOU!

Ryzen 1700 performs like a 2600k @ 4.5Ghz in games... DOOM AND GLOOM!

Oh, wait, that 1700 is only running 3.0~3.7Ghz?!?

Holy sh...

Wait - there are patches coming to make it even better?

...

On a side note: I'm buying a Fury to test with Ryzen - I'm more interested in the driver behavior compared to nVidia driver behavior.. my RX 480 will be another data point, but I need to be able to eclipse its performance for testing - which the Fury can handily do (while using much more power, naturally).

Then I'll have to figure out what to do with one of those cards

 

[Mockingbird]

1. It doesn't run poorly. Only folks with a limited ability to interpret data think that.

2. A microarchitecture flaw that is specific to games? Do you think intel only packed in a 32bit GAME instruction per die as opposed to Intel's 4x 512b GAME instructions per core?

What I want is something I can believe in.

Show me a few games where Ryzen 7 1800X was trailing the Core i7-6900K, but after some updates/optimizations run as well on Ryzen 7 1800X as on the Core i7-6900K.

Obviously, I am talking about cases where the games are not bottlenecked by GPU, etc.

 

[piesquared]

Who and where? I want to see it for real.

It was posted a few pages back.

https://www.reddit.com/r/Amd/comments/5xdxvy/the_overclocker_magazine_that_moment_when_your/

Bear in mind there is no confirmation on this.

 

[looncraz]

This is terrifying. Are you serious?

In a way, it's true. AMD has done some 'clock stretching' - up to about 7% - for the last couple of generations to save power. This only happens in response to hitting a TDP limit and it happens so fast that you can't really see it with software monitoring. It will look like the base clock is jumping around when this happens, but usually only within 1~2Mhz. But when you have a multiplier of 40, that's a pretty big impact.

This will happen less with better cooling, better VRMs, or just more efficient chips. I witnessed this behavior on the x4 845. The performance, however, is superior to normal turbo modes - and XFR generally means you will be at a multiplier of 40.25~41, so you will actually be at or above 4Ghz if XFR is running...

This all goes away in OC mode, however. But that has its own costs - OC mode, frankly, sucks. My 24/7 overclock will be a refclock/bclk overclock unless I find using Windows to manage the power states doesn't have much of a draw back, then I'll set it to ~4Ghz and let the P-states be manually handled by the software.

Ryzen will take a lot of testing to find out what is best.

 

[desura]

If I were building a PC today good chance I would get this chip. It is within Intel by a negligible unnoticeable amount but is significantly cheaper and it also has better multitasking performance such as gaming while streaming.

But my ivy bridge is more than fast enough for my uses.