Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[Riek]

What about those who don't overclock?

The 1400 will still be faster by a big margin.

HT alone will give a rather big uplift, but the cpu's you are comparing also have a big difference in frequency:

the i5 - 6400 has a boost frequency of 3.3GHz, the 1400 -> 3.45GHz. (probably gives similar ST performance)
The base frequency of the 6400 is 2.7GHz, the base frequency of 1400 is 3.2GHz.

So in any application (including games), the difference can go up to:
+ 4 additional threads
+ up to 20% frequency difference!!

tbh, the i5 6400 isn't competing at all;
They either need an i5 clocking well against the 4Ghz (to offset the lower amount of threads. note: the difference between 4-> 8 threads is higher than 8->16threads)
Or they need to enable HT on the i5 range to compete.

 

[mohit9206]

The 1400 will still be faster by a big margin.

HT alone will give a rather big uplift, but the cpu's you are comparing also have a big difference in frequency:

the i5 - 6400 has a boost frequency of 3.3GHz, the 1400 -> 3.45GHz. (probably gives similar ST performance)
The base frequency of the 6400 is 2.7GHz, the base frequency of 1400 is 3.2GHz.

So in any application (including games), the difference can go up to:
+ 4 additional threads
+ up to 20% frequency difference!!

tbh, the i5 6400 isn't competing at all;
They either need an i5 clocking well against the 4Ghz (to offset the lower amount of threads. note: the difference between 4-> 8 threads is higher than 8->16threads)
Or they need to enable HT on the i5 range to compete.

Yeah good point. Now only if amd had a G4560 competitor.

 

[KompuKare]

Graphs from the Youtube Video! 3.96ghz Ryzen 1700 wins 4 out of 6 tests versus 5.0ghz 7700k!! (Really it wins all 6 if you look at minimum frames.

Which assuming there isn't some crazy outline (some driver or other hugging everything for a few ms), min frames is the way most reviews should be ranked. Think GameGPU do it that way but far too many sites don't

 

[lolfail9001]

I don't think that operate a bus up to 3600MHz instead 1800MHz would be a problem apart from power drain...

Careful there, your expectations are a curse so far.

The 1400 will still be faster by a big margin.

HT alone will give a rather big uplift, but the cpu's you are comparing also have a big difference in frequency:

the i5 - 6400 has a boost frequency of 3.3GHz, the 1400 -> 3.45GHz. (probably gives similar ST performance)
The base frequency of the 6400 is 2.7GHz, the base frequency of 1400 is 3.2GHz.

So in any application (including games), the difference can go up to:
+ 4 additional threads
+ up to 20% frequency difference!!

tbh, the i5 6400 isn't competing at all;
They either need an i5 clocking well against the 4Ghz (to offset the lower amount of threads. note: the difference between 4-> 8 threads is higher than 8->16threads)
Or they need to enable HT on the i5 range to compete.

6400 has all core turbo of 3.1Ghz against 3.25 or something of 1400. So we shall see how they compete. Ah, who am i kidding, they (7400 and 1400) will even in gpu bound titles most likely.

 

[bjt2]

Careful there, your expectations are a curse so far.

The infinity fabric seems to be 32 byte wide. The bus between L1<->L2, L2<->L3 also is 32 byte wide and goes at core speed, so up to 4.1GHz. I don't see any problem also for infinity fabric bus.

 

[unseenmorbidity]

Its just reviewers starting to understand how to bm a game more than ryzen obviously have greater benefit of higher speed ram. Its old news. If we take 0.1 mins who wouldnt select a 1700 with 2667 ram any time of the weeks for these games anyway?

Its a matter of how you test in the games and what you are looking for. Computerbase frametime bm of 7700 vs 1800x shows same extreme uplift of ryzen. They just play different maps and select the hard situations. Had he done the same ryzen would also dominate here.

Its the old story. 4 extra cores obviously gives a huge benefit when the game can use it and in high load situations.

Besides:
Solving the ccx speed issue via faster ram obviously works everywhere and gives the benefit of faster ram as sideeffect but its also a brutal and expensive way of adressing it. New uefi and the ability to alter it via software is a bit more elegant but probably comes at some efficiency loss. Still for desktop that seems like a solution to me. Give the fabric a voltage hike alter the timings and let us oc it.

But at the end of the day writing software that fits the processor and not only Intel is the elegant and perf efficient solution. As explained in the technical thread there is several possibilities to adress and solve most of the ccx problem. At the same time adapting to the entire cache and core typology will give a nice boost imo.

I am still surprised how damn fast this processor is for desktop out the gate. Obviously its a balanced design but imo there must be some technical gems hidden in eg branch prediction that really is a good deal faster than what Intel does to outweight for this beeing a new platform.

Isn't that what they should be doing though? What's the point of loading up Watch Dogs 2, and walking back and forth in an empty alley?! The point is to see how the CPU performs in the actual game, and maybe make some predictions about which will be better in the near future.

I edited the pic of the benchmarks that was posted. I removed the irrelevant data, and the i7, since we are only talking about Ryzen. I also put in ratios from one tier to the next to show the gains.



EDIT: I included average gains for all the games tested, except Mass effect. It's a new game, and a pretty large outlier here.

Anyways, going from 3200 to 3600 MHz DDR4 netted about 3% gains in FPS. However, these are some of the most CPU intensive games out there. If he had tested a larger suite of games, then I would imagine the difference would be lower.

 

[krumme]

Isn't that what they should be doing though? What's the point of loading up Watch Dogs 2, and walking back and forth in an empty alley?! The point is to see how the CPU performs in the actual game, and maybe make some predictions about which will be better in the near future.

I edited the pic of the benchmarks that was posted. I removed the irrelevant data, and the i7, since we are only talking about Ryzen. I also put in ratios from one tier to the next to show the gains.



EDIT: I included average gains for all the games tested, except Mass effect. It's a new game, and a pretty large outlier here.

Anyways, going from 3200 to 3600 MHz DDR4 netted about 3% gains in FPS. However, these are some of the most CPU intensive games out there. If he had tested a larger suite of games, then I would imagine the difference would be lower.

Well i havnt looked at cpu gaming venchmarks for 10 years. Because the reviewers dont look for the hard places and situations. It makes the bm more or less useless imo.

Peopled complained tf2 ran slow in many matches back then and it was simply a single thread cpu bottleneck in big mp battles during some loads. Not remotely reflected in the typical sp bm that was used. Different worlds.

Today 10 years after people complain about their new i5 or i7 in bf1. They just got the tf2 memo so they understand as little as back then. Why does my rig run slow when all the bm shows its perfectly capable?

I have stopped saying its because its of their cpu. Its lost cause. Instead its something like "stop playing sniper and learn to play operations or go back to conquest" lol

 

[guachi]

Graphs from the Youtube Video! 3.96ghz Ryzen 1700 wins 4 out of 6 tests versus 5.0ghz 7700k!! (Really it wins all 6 if you look at minimum frames.

Look at those minimum 0.1% frame rates in GTA 5, Mass Effect, and Crysis 3...

I saw the video yesterday. If you set all the i7 results at a score of 100, the Ryzen 3200 results look like this:
.1% 114.0
1% 104.0
avg 98.8

The numbers for the .1% and 1% are really thrown off by the Crysis 3 results, but I'm not going to exclude data without good reason.

With the image quality settings and video card the reviewer is using, Ryzen is extremely competitive with the best Intel has to offer for gaming.

 

[HurleyBird]

Graphs from the Youtube Video! 3.96ghz Ryzen 1700 wins 4 out of 6 tests versus 5.0ghz 7700k!! (Really it wins all 6 if you look at minimum frames.

Look at those minimum 0.1% frame rates in GTA 5, Mass Effect, and Crysis 3...

Rise of The Tomb Raider is rather shamelessly absent from that picture:

 

[inf64]

@ HurleyBird

In ROFTM Ryzen @ 4Ghz with 3.2Ghz DDR4 had 4% lower 0.1%, 6% lower 1% and 8.5% lower average Vs 5Ghz 7700K with 3.2Ghz DDR4. I'd say that for the worst case scenario game for Ryzen, the performance difference is basically negligible at best at 1080p. Awesome scaling from Zen with high freq. memory.

 

[HurleyBird]

I'm not saying that all seven results taken together don't paint a good picture. They do, and even with RoTP factored in Ryzen looks superior (although these tests need to be verified from more sources. Do your job, journalists!), but that doesn't change the fact that leaving out the one bench where Ryzen doesn't look good is completely shameless.

 

[KompuKare]

I'd say that for the worst case scenario game for Ryzen, the performance difference is basically negligible at best at 1080p.

Well, worst case in this suite of games of course. The worst case possible gaming tests would probably be Fallout4 or Skyrim with lots of mods as they both scale very poorly with cores and it seems that scripts just run on one core (or that is a possible explanation for why Fallout4 mods with lots happening run so poorly). Been playing Bethesda's TES stuff since Daggerfall but they sure don't know how to program that well nor do much testing. But I love their open-world RPGs once fans have made mods.
As for MindBlank Tech's tests I didn't see how he tested BF1. If it was in single-player then where it really matters - intense multiplayer playing - Ryzen would most likely have won by quite a bit (in the BF1 test Intel was ahead on average by only 1.06%, ahead on min 1% by 1.84% and behind on average 0.10% by 3.99%). Krumme's point about multiplayer performance (or any gaming review which doesn't use the most stressful part of the game) is very important but in the case of the Battlefield series it has been almost totally ignored since Mantle whether due to lack of repeatability, laziness, incompetence or indifference from most reviewers.

Anyway a very good showing. If AMD can afford to quickly revise Ryzen and port it to higher clocking process they could even clean up in the PC Master Race gaming market. Guess Intel will finally have to become a bit more generous and offer 6C/12T for the same price or less than their current 4C/8T i7's. Would love to see these kind of benchmarks repeated with both running at 4.0GHz just to see get an idea what a Ryzen+ @ 5.0GHz would look like.

 

[CatMerc]

@ HurleyBird

In ROFTM Ryzen @ 4Ghz with 3.2Ghz DDR4 had 4% lower 0.1%, 6% lower 1% and 8.5% lower average Vs 5Ghz 7700K with 3.2Ghz DDR4. I'd say that for the worst case scenario game for Ryzen, the performance difference is basically negligible at best at 1080p. Awesome scaling from Zen with high freq. memory.

Except it's a game where a 6900K does far better than what Ryzen is putting out here. The extra cores aren't being utilized, despite them being utilized with the 6900K.

 

[sushukka]

Graphs from the Youtube Video! 3.96ghz Ryzen 1700 wins 4 out of 6 tests versus 5.0ghz 7700k!! (Really it wins all 6 if you look at minimum frames.

Look at those minimum 0.1% frame rates in GTA 5, Mass Effect, and Crysis 3...

The coolest thing still here is that AMD's 8core processor beats Intel's 4core processor which is their fastest processor targeting gaming sector...and at the same time this very same processor gives the Intel's flagship 8core HEDT 6950k hard time being very close beating it with about third of the price. It seems that Intel needs to merge at least two product lines in the near future for just not to look stupid. Impressive, most impressive AMD...

 

[unseenmorbidity]

For some reason, some people want to compare the Ryzen CPUs to both the 7700k, and the 6900k. Apparently Ryzen "sucks", if it cannot beat both simultaneously. As if that was ever going to be possible...

 

[MajinCry]

Well, worst case in this suite of games of course. The worst case possible gaming tests would probably be Fallout4 or Skyrim with lots of mods as they both scale very poorly with cores and it seems that scripts just run on one core (or that is a possible explanation for why Fallout4 mods with lots happening run so poorly). Been playing Bethesda's TES stuff since Daggerfall but they sure don't know how to program that well nor do much testing. But I love their open-world RPGs once fans have made mods.

Actually, Skyrim and Fallout 4 have multithreaded scripts. I don't know if each Papyrus script gets it's own thread, but States do. So if you have a script like the following:

auto State BeginState

   Event OnBeginState()
       GoToState("01Hellow")
       GoToState("02Hellow")
       GoToState("03Hellow")
   EndEvent
  
Endstate

State 01Hellow
   Event OnBeginState()
       Debug.Notification("Hello World! - 01")
   EndEvent
EndState

State 02Hellow
   Event OnBeginState()
       Debug.Notification("Hello World! - 02")
   EndEvent
EndState

State 03Hellow
   Event OnBeginState()
       Debug.Notification("Hello World! - 03")
   EndEvent
EndState

The three hello world states will each run on a separate thread. Not really a good idea to spawn threads for extremely simple workloads, as creating and destroying threads isn't "free". Shows the idea well enough.

There are other ways to thread scripts, but using states is the simplest one.

And yeah, Skyrim doesn't scale well with cores, but it does make use of them. I tested this by changing the affinity for TESV.exe in Task Manager. Changing it to use 3 cores of my Phenom II took a few fps off, setting it to use 2 did the same again, and setting it to 1 core put me in 20fps overlooking Solitude, whereas with four cores I was at around 35fps.

Fallout 4 likes more cores, especially since it has threadlocks up the yang. There's about 100k active at any given frame, so being able to spread those across 8 actual cores is much better than spreading them across 4 cores + 4 SMT threads.

 

[sushukka]

For some reason, some people want to compare the Ryzen CPUs to both the 7700k, and the 6900k. Apparently Ryzen "sucks", if it cannot beat both simultaneously. As if that was ever going to be possible...

True. However, 6900k and 7700k are from different product lines but only because Intel say so and have made us to believe that it has to be like that. The reason is of course that they can get premium from HEDT line what they wouldn't be able to do by introducing the same features to the consumer targeted i7. One more place where Intel plays with their current market position. One would say that it's business as usual, but as consumer I certainly don't like it. Now it's quite refreshing to see how AMD is breaking these artificial barriers forcing Intel to think their milking tactics little bit more.

 

[CatMerc]

For some reason, some people want to compare the Ryzen CPUs to both the 7700k, and the 6900k. Apparently Ryzen "sucks", if it cannot beat both simultaneously. As if that was ever going to be possible...

It should be close to 6900K at all times. Whether or not it beats a 7700K would be down to game multithreading.

 

[MajinCry]

It should be close to 6900K at all times. Whether or not it beats a 7700K would be down to game multithreading.

Ryzen's not all that close to Haswell. In single threaded workloads, sure. But in gaming, the cross-CCX draw call performance is abysmal; it's at Core 2 levels. On a single CCX, with 3200Mhz DDR4, Ryzen performs a tad better than Sandybridge at draw calls.

It would take even faster RAM for AMD to catch up to Haswell in intensive scenes, and if it uses more than four cores...Ew.

 

[thepaleobiker]

Ryzen's not all that close to Haswell. In single threaded workloads, sure. But in gaming, the cross-CCX draw call performance is abysmal; it's at Core 2 levels. On a single CCX, with 3200Mhz DDR4, Ryzen performs a tad better than Sandybridge at draw calls.

It would take even faster RAM for AMD to catch up to Haswell in intensive scenes, and if it uses more than four cores...Ew.

MajinCry - Just wanted to say I've spotted your profile on many comments in youtube vids I saw you linked to this thread a few times. Small (digital) world lol

Regards,
Vish

 

[unseenmorbidity]

Ryzen's not all that close to Haswell. In single threaded workloads, sure. But in gaming, the cross-CCX draw call performance is abysmal; it's at Core 2 levels. On a single CCX, with 3200Mhz DDR4, Ryzen performs a tad better than Sandybridge at draw calls.

It would take even faster RAM for AMD to catch up to Haswell in intensive scenes, and if it uses more than four cores...Ew.

What dark smelly orifice did you procure this information from?

 

[ultima_trev]

Any one use UserBenchMark?

I'm in the top 3% of Ryzen 1800X builds in UserBenchMark:

http://www.userbenchmark.com/UserRun/3219849

103% for CPU score. For what it's worth, the average of all i7 5960X baselines is 91% and 6900K 100% even. Now only if Ryzen could overclock as good as Haswell...

Edit: Top 1% in Passmark CPU Mark
http://imgur.com/a/yq2Eo

 

[Malogeek]

Now only if Ryzen could overclock as good as Haswell...

I wonder how it would rank in $/perf though

 

[MajinCry]

What dark smelly orifice did you procure this information from?

Parapa-pam! http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=threads/part-2-measuring-cpu-draw-call-performance.2499609/

MajinCry - Just wanted to say I've spotted your profile on many comments in youtube vids I saw you linked to this thread a few times. Small (digital) world lol

Regards,
Vish

Jyargurinsk greets you.

 

[unseenmorbidity]

Parapa-pam! http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=threads/part-2-measuring-cpu-draw-call-performance.2499609/

Jyargurinsk greets you.

Now show me the correlation with that and actual games run at realistic settings. You can't, because that information is about as useful as cinebench for making predictions.