Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[ManyThreads]

We are assuming that your assumption on the GPU's role is wrong. I'm 99.9% sure that it is about the difference in GPUs.
The create results are scaling in a very different way between the 2 systems and that is odd. The diff in perf is very odd too as it is too large to be IPC related
Testing on CPU at least takes the GPU out of it, if you can't do better than that and use the same GPU in both systems.
BTW do share results if you test on CPU.

quickly butchered your graph in IrfanView. This suggests that the 2x950 are the limiting factor.

Thanks - I will try and roll the same test with GPU acceleration unchecked. I am at least curious. I won't be able to do it for a little while on the 1800X machine, but I can test my own setup with the 3770K this weekend.

What could the explanation be then for the fact that moving from a 560 Ti (1GB) to a Titan X had no noticeable effect on speed on the exact same system running the exact same processes? That is what keeps bothering me about the assumption that the GPU is in fact making a difference, but I am open to anything at this point.

Point is that when core amount increases you'll lose clock speed, cannot have both. This applies to all processors on the market from mainframe to x86. I'll doubt Intel's new 6 or 8 core 7700K will reach even near the clock speeds than the current 4core one. So basically you'll have to wait for lower nm processes (around 2years) or just put the money where you get the most of it. Currently it seems that Ryzen is having the best bang for the buck especially when keeping the future needs in mind.

I realize I can't have both, but you can currently OC a 6-core i7 6850K easily to 4.4 Ghz, the exact same as my 3770K. My thinking is if Skylake X or Coffee Lake 6 core's allow 4.5-4.8 GHz or so with an OC, I can have my cake and eat it too with both a core count increase and significant clock speed increase. I am not too concerned about price unless it's outrageous. Or if future Ryzen allows for a significantly bigger OC, that could work too. My problem is that for my side job (real estate photography) I equally rely on both heavily multi-threaded tasks *and* heavily single threaded tasks. Trying to find the best crossover point of cores and Ghz for a CPU is proving difficult.

 

[lolfail9001]

I've posted this before, but I have own both a Ryzen 1700 and an i5 6600k. 6600k is overclocked to 4.5ghz, the 1700 is overclocked to 3.9ghz.

The Ryzen is wayy better in games. . When reviewers do their reviews, they have ONLY the game running and compare FPS. That is not realworld. When I use my PC, I have youtube, podcast, music, browsers, vent, skype, etc open. This is where Ryzen pulls ahead. The ability to use multiple programs and not have a reduced FPS while gaming.

I use my computer for Streaming, and here it is a no contest win for the Ryzen cpu, but unless you are only playing the game and thats it, its almost always a better experience with the 8 cores and 16 threads.

I have posted this before too, but may i see some numbers? I mean, seriously, it is a shame that reviewers almost never do that sort of testing.

P. S. Youtube/music/podcast/browsers have little to no performance impact in general, just saying.

 

[imported_jjj]

I've seen a lot of different benchmarks, and there are some where Ryzen has a better minimum FPS than the i7 7700K, but I've seen more games where the i7 7700K had higher minimums. Clearly you showed an example of a game where Ryzen excels, probably due to Crysis scaling well with more cores. Hopefully we'll see more games that scale well across more cores.

Min FPS are usually 1 frame spikes and i don't know if that's of much relevance.
Avg is a better metric or maybe mid 90% would be even better.
If you have a certain FPS target, that kind of graph is nice as it kinda shows you consistent highs and lows.
The testing at high FPS is not very useful but reviewers are fixated on that so we got little good data. Even in Crysis if you game at 40-60 AVG FPS there might be no difference but we don't quite know that.
Plenty of games do scale even today and that was very visible with Intel's HEDT platform already.

Ryzen does underperform in some games right now- with a lot of speculation on reasons.
I think the games that have a perf penalty when SMT is enabled are the ones where we can see a fix soon. SMT shouldn't be a penalty and a minor game update should turn SMT into a gain.
Future ST vs MT is easy. ST doesn't scale while MT is underexploited and has room to scale.

 

[sushukka]

Simulating AMD Ryzen 5 1600X, 1500X Gaming Performance

http://www.techspot.com/review/1360-amd-ryzen-5-1600x-1500x-gaming/

Graz you screenshotted the worst part of Ryzen's results and posted it here. It was also the first test in the review, which brings the thought you didn't even read the rest of it. In the conclusion there were also following quotes:
With the exception of Far Cry Primal, it looks like the quad-core Ryzen CPUs will destroy the higher-end dual-core Kaby Lake processors such as the 7350K. Even at 4.8GHz, the 7350K was no match in the more CPU-intensive titles.
It was interesting to find that when paired with the GTX 1070, the Ryzen CPUs were actually able to pull ahead in games such as Mafia III, while folks equipped with a sub-$300 current-gen GPU won't see any difference between the quad-core Ryzen and Kaby Lake CPUs.
As things stand, we're still waiting for more games that better utilize Ryzen cores. It looks like plenty of them are on the horizon, but we currently have games that make Ryzen look average and not as good as it actually is.
What we can tell you today is that the 1500X is going to be a remarkable value at $190, a price that also includes a 'Wraith Spire' cooler -- quite the package.

I'm just wondering what have Intel done to have such a fine legion of fanboys. From business perspective that company has abused the market in so many ways that they nearly reach Microsoft's level. Everybody knows that, wants some change and still here we are: best overall performing processor, from the welcomed underdog player, lower TDP, best price-performance level, future safe upgradeability and...and...still people are whining of some minor nuisance in max FPS in some specific game or when their memory speeds reach only 3000Mhz instead of 3200Mhz. I mean wtf, get the head out of the box and look at the bigger picture. This time you really don't need to even act like a martyr but can invest your money to get the best cost-perfomance product and at the same time help the market find the right balance against Intel's domination.

 

[richierich1212]

I realize I can't have both, but you can currently OC a 6-core i7 6850K easily to 4.4 Ghz, the exact same as my 3770K. My thinking is if Skylake X or Coffee Lake 6 core's allow 4.5-4.8 GHz or so with an OC, I can have my cake and eat it too with both a core count increase and significant clock speed increase. I am not too concerned about price unless it's outrageous. Or if future Ryzen allows for a significantly bigger OC, that could work too. My problem is that for my side job (real estate photography) I equally rely on both heavily multi-threaded tasks *and* heavily single threaded tasks. Trying to find the best crossover point of cores and Ghz for a CPU is proving difficult.

It's probably going to be tough to get the Skylake-X processor to clock as high as you're expecting. I'm going to guess the best ones will hit 4.4-4.5GHz max. That's still very, very fast.

Thanks for sharing your testing though. It's good to get real world results from folks here on the forum.

 

[IEC]

Cherry picking is a timeless tech forum tradition.

95% of the time Ryzen is wonderful. In the 5% of cases where it isn't (yet), you get the same un-optimized results posted here day after day crowing about Ryzen "failing" in whatever metric. Sometimes missing that there is already an update which fixes performance (Dota2 +25% being the most recent example).

I'd wager the vast majority of these corner cases will get fixed in months if not weeks, considering the sheer number of well-threaded use cases demonstrating that Ryzen has plenty of horsepower under the hood. As well as seeing that the gap is already being closed in some applications and games with patches.

 

[imported_jjj]

What could the explanation be then for the fact that moving from a 560 Ti (1GB) to a Titan X had no noticeable effect on speed on the exact same system running the exact same processes? That is what keeps bothering me about the assumption that the GPU is in fact making a difference, but I am open to anything at this point.

Maybe you are recalling it wrong, maybe you measured total time only,maybe the software changed since.

Ofc ideally you would want to test with same GPU in both systems as CPU only will favor more cores for the task that should be run on GPU.
At the very least you'll see how much of a difference the GPU makes.

 

[kush120]

I have posted this before too, but may i see some numbers? I mean, seriously, it is a shame that reviewers almost never do that sort of testing.

P. S. Youtube/music/podcast/browsers have little to no performance impact in general, just saying.

They certainly do have an impact. Watching HD content on one monitor while playing games on another is alone enough to reduce FPS quite substantially.

Edit: Watch this video at 8k and tell me you would have no performance impact during gaming.
www.youtube.com/watch?v=f0NdOE5GTgo

 

[unseenmorbidity]

Who watches 8k videos?

 

[Rifter]

Who watches 8k videos?

Not me, but i certainly would if i had a 8k monitor!

 

[unseenmorbidity]

Sometimes I watch videos in 144p, just because of my data cap. I have to have one of the worst internet plans out there.

 

[lolfail9001]

They certainly do have an impact. Watching HD content on one monitor while playing games on another is alone enough to reduce FPS quite substantially.

I have literally ran a well threaded benchmark while watching some 720p vid (my internet cannot handle 1080p half of the day, also it is literally HD content) on my lowly i5. Impact was 2% across the board. I afford a thought that it would be larger at 1080p and beyond, though.

Edit: Watch this video at 8k and tell me you would have no performance impact during gaming.

So, HD content or 4k and beyond? I mean, sure, if your niche involves working with 4k+ video while gaming, then yes, Ryzen stomps i5 every day of the week and smokes afterwards.

 

[happy medium]

Did you really just take the worst results of that review?

is was a pain in the A to link, I took the first one and gave up.

 

[ultima_trev]

My opinion so far on Ryzen 1800X versus i5 6600K... I came from an i5 6600K on a GA Z170 HD3 board which I ran 24/7 at 4.2 GHz with RAM at 2666/timings 15 15 15 30. Suicide runs in 3DMark were done at 4.5 GHz and RAM at 2800 (timings identical). RAM at 3000 proved unstable.

I run my 1800X on a GA AB350 Gaming at 3.92 GHz for 24/7 usage and 4.05 for 3DMark suicide runs. RAM (4x8G DIMMs) is too unstable at 2666 for me at any timings so for now I'm sticking to 2400/timings 14 14 14 14 30 (default for my kit is 17 17 17 17 39 BTW).

My only complaint is the variance of Combined Score results in 3DMark Firestrike. I've seen it as low as 5300 and as high as 6605, even if identical settings were used. My i5 on the other hand consistently scored around 9200.

Other than that, in games (I mostly play Witcher 3, Watch Dogs 2, Crysis 3, The Old Republic, WoW and StarCraft II) the experience is identical between them... Except in Witcher 3 and Watch Dogs 2 which seem notably smoother on the Ryzen CPU.

Standard office/desktop apps feel the same between the two.

In conclusion, I thought for sure the i5 6600K OC I had would prove better in normal desktop usage but so far it's pretty much a wash. Considering where AMD was with Bulldozer, Ryzen is all sorts of a winner.

 

[ozzy702]

It's probably going to be tough to get the Skylake-X processor to clock as high as you're expecting. I'm going to guess the best ones will hit 4.4-4.5GHz max. That's still very, very fast.

Thanks for sharing your testing though. It's good to get real world results from folks here on the forum.

I'd be really surprised if even poor Skylake-X six cores don't hit at least 4.4ghz with golden chips hitting 4.7ghz. Only time will tell though.

 

[tential]

I have posted this before too, but may i see some numbers? I mean, seriously, it is a shame that reviewers almost never do that sort of testing.

P. S. Youtube/music/podcast/browsers have little to no performance impact in general, just saying.

This is why I never recommended the i5. i7 is bare minimum. Now for me, it's HEDT platform from Intel, or Ryzen 7. I'm sure some people can enjoy Ryzen 5. Ryzen 5 is ideal for gamers IMO as it gives flexibility and is cheap.
Otherwise, just get the HEDT platform. I wouldn't get any more i7s that are 4 core 8 thread. You can keep a processor for awhile, just buy a good one and have it last a long time.

 

[imported_jjj]

TrustedReviews on cons for the 1700 -"High power draw when overclocked"

 

[Ajay]

TrustedReviews on cons for the 1700 -"High power draw when overclocked"

Mooks- compared to what, a 4.5 GHz 7600k??
Because, uhm 4 cores == 8c/16t.

...typo

 

[bystander36]

Min FPS are usually 1 frame spikes and i don't know if that's of much relevance.
Avg is a better metric or maybe mid 90% would be even better.
If you have a certain FPS target, that kind of graph is nice as it kinda shows you consistent highs and lows.
The testing at high FPS is not very useful but reviewers are fixated on that so we got little good data. Even in Crysis if you game at 40-60 AVG FPS there might be no difference but we don't quite know that.
Plenty of games do scale even today and that was very visible with Intel's HEDT platform already.

Ryzen does underperform in some games right now- with a lot of speculation on reasons.
I think the games that have a perf penalty when SMT is enabled are the ones where we can see a fix soon. SMT shouldn't be a penalty and a minor game update should turn SMT into a gain.
Future ST vs MT is easy. ST doesn't scale while MT is underexploited and has room to scale.

Most reviews use 300 ms to 1 second time frames to determine minimums. Not a single frame.

 

[imported_jjj]

Most reviews use 300 ms to 1 second time frames to determine minimums. Not a single frame.

Then it's not min FPS.
Anyway can you provide some examples as the vast majority don't even look at min FPS and i can't find any that test the way you mention.

 

[bystander36]

Then it's not min FPS.
Anyway can you provide some examples as the vast majority don't even look at min FPS and i can't find any that test the way you mention.

Your previous post said it is not good when you thought they took a single frame, and you are saying that if they take the minimum FPS based on 300 ms to 1000ms, it's wrong. If both ways are wrong, what is the right way to do it?

And minimum FPS doesn't have to be a single frame. It's what ever frame time they chose, but it should be relatively small.

 

[bystander36]

Anyway can you provide some examples as the vast majority don't even look at min FPS and i can't find any that test the way you mention.

This is a little tricky, as most sites won't specify that minimum time frame. That said, they usually mention what software they use. If they use FRAPS, it uses 1 second intervals for their minimums (you can see this with their output files). If they use MSI Afterburner, it uses 300ms for their intervals (though I'm not sure who uses MSI Afterburner. FRAPS was always used, PresentMon is usually used for DX12).

I'm not sure what PresentMon uses as it's minimum frame time, but you can tell by the results it cannot be a single frame, as FCAT will show their longest frames as much lower than minimum FPS shown with PresentMon.

 

[sushukka]

I'd be really surprised if even poor Skylake-X six cores don't hit at least 4.4ghz with golden chips hitting 4.7ghz. Only time will tell though.

Time will tell yes, but I'm little bit sceptic when looking the past so far. Broadwell-E is the newest generation where Intel have 6-10cores available. Fastest 6950K have 3,6GHz, Turbo boost 3,8Ghz and max single core 4GHz. On server side you only have four core Skylakes available where fastest is 3,7GHz with 4GHz turbo. Newest and best Kaby Lake's (7700K, 7600K) 4,2GHz/3,8GHz normal speed, turbo clock drops about 100MHz when single, dual or four cores are in use. As the same behaviour happens through Intel's product line, it seems that the multicore clock penalty will hit Intel likewise. When having 6 cores or 8 cores, the difference with AMD seems to diminish/vanish. Fastest 8 core Xeon has base 3,6GHz and turbo 4Ghz. When adding cores and taking the biggest 22 core Intel the base speed is only 2,2GHz.

In Intel's roadmap the first desktop Octacore (Coffee-Lake) processors are going to be seen not until 2018. It seems that the only possibility to have higher frequencies in these chips is lower the nm process where Intel has been struggling for some time (10nm was first to be announced 2015). Coffee-Lake will still be 14nm. Current status is that maybe at the end of this year Intel can ship the first 10nm processors (Cannonlake). However, currently Global Foundries is finalizing its 7nm process and first mass production of 7nm chips will be available probably on the second half of the next year or early 2019. In other words, things are getting exciting.

 

[Ajay]

Skylake-X is manufactured on the same high frequency 14nm + process as Kabylake. I'd be surprised if top bins didn't hit 4.6 GHz fairly easily.
Coffeelake will have a six core max (not 8).

 

[Zstream]

Has there been any proof of this besides the BF1 case a certain users post all over the place?

It's like HiFi sector. Every measurement tells you there is no difference between the $10 and $1000 cable yet the sound has more vibrance and clarity. I mean you have to justify the purchase somehow.

Why could Ryzen be more fluid? Because there is a CPU bottleneck and if you constantly run at say 60 fps is feels smoother than say a 7700k that runs at 120 fps, then gets to a GPU heavy area and drops to 80 fps and then back to 120 fps. How could you fix this? Get a FreeSync or Gsync monitor. The issue with stutter and smoothness is that above a certain threshold what we perceive as unsmooth are changes in framerate and not a generally low framerate.

I've been playing with my new box and it's clearly more fluid, especially when I'm doing multiple things. However that could be just he addition if more cores, but it's still the product and AMD has it.

I'm coming from a Haswell 1231.