Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

Page 60 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

inf64

Diamond Member
Mar 11, 2011
3,684
3,957
136
Yep take a look at score of their Ryzen chip in 3dmark physics: 12587.

Then take a look at the recent leak(originally via videocarz):
http://www.eteknix.com/amd-ryzen-3dmark-physics-score-leaked/

Ryzen 8C/16T ES @ 3.4/3.8Ghz scored 17800pts, Something wrong with one of those 2 results. Either the first leak was not true (videocardz source, unlikely) or the 2nd one is gimped(OCholic)

The score of 17K in Physics was recently confirmed by 2nd source(Xfastest) all with screenshots of the run.

Edit: Seems it was Firestrike Physics subtest. I stand corrected then.
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Chances are that in some of the tests, Ryzen needs faster memory to be competitive.
Reviews are gonna get messy as folks are not gonna give it the BW it needs.
 

USER8000

Golden Member
Jun 23, 2012
1,542
780
136
Yep take a look at score of their Ryzen chip in 3dmark physics: 12587.

Then take a look at the recent leak(originally via videocarz):
http://www.eteknix.com/amd-ryzen-3dmark-physics-score-leaked/

Ryzen 8C/16T ES @ 3.4/3.8Ghz scored 17800pts, Something wrong with one of those 2 results. Either the first leak was not true (videocardz source, unlikely) or the 2nd one is gimped(OCholic)

The score of 17K in Physics was recently confirmed by 2nd source(Xfastest) all with screenshots of the run.

Edit: Seems it was Firestrike Physics subtest. I stand corrected then.

Well the results are from January and also it is for 3DMark11. The chart looks very weird - the Core i7 6900K at 4.2GHZ is nearly 50% faster than a Core i7 6850K or a Core i7 6800K at 4.2GHZ with both running 3GHZ RAM??

I think it is that website just taking the mickey for some of its more alert readers.
 
  • Like
Reactions: psolord

Minkoff

Member
Nov 7, 2013
54
8
41
Chances are that in some of the tests, Ryzen needs faster memory to be competitive.
Reviews are gonna get messy as folks are not gonna give it the BW it needs.

At the end of the Ryzen announcement Lisa Su said: "...We Welcome you to review to your hearts content...", so I guess we'll see what "hearts content" of the reviewers will be, when alternative benchmarks, from users and enthusiasts, start to surface. :)
 

Shivansps

Diamond Member
Sep 11, 2013
3,835
1,514
136
https://www.computerbase.de/2017-02/cpu-skalierung-kerne-spiele-test/

Untitled.png




And they sent 1800X s to reviewers, which is compared to the 6900X, by them ;)

Keep in mind Intel HEDT chips have bigger cache and quad channel, that will yield better results, is just 15% for double number of cores, we know for a fact that Kaby Lake gives better gains than that by using faster ram.
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Yep take a look at score of their Ryzen chip in 3dmark physics: 12587.

Then take a look at the recent leak(originally via videocarz):
http://www.eteknix.com/amd-ryzen-3dmark-physics-score-leaked/

Ryzen 8C/16T ES @ 3.4/3.8Ghz scored 17800pts, Something wrong with one of those 2 results. Either the first leak was not true (videocardz source, unlikely) or the 2nd one is gimped(OCholic)

The score of 17K in Physics was recently confirmed by 2nd source(Xfastest) all with screenshots of the run.

Edit: Seems it was Firestrike Physics subtest. I stand corrected then.

3DMark 11 Physix scales very well with memory BW.
They got the 6900k at 4.2GHz doing 17497 and then same clocks but the DRAM at 3000 gets 19906 so Ryzen will always be behind here with dual chan.
In some other benches slow memory could be a bottleneck that from a certain point is not there anymore but not in this one.
 

french toast

Senior member
Feb 22, 2017
988
825
136
Keep in mind Intel HEDT chips have bigger cache and quad channel, that will yield better results, is just 15% for double number of cores, we know for a fact that Kaby Lake gives better gains than that by using faster ram.
Memory bandwidth scaling in games drops off after 3600mhz DDR4 does it not? In any case faster Memory than that is ridiculously expensive to be pairing with a 350$ cpu.

Food for thought, R7 1700 has 20mb of cache, also its IMC is more efficient at extracting maximum bandwidth out of its memory than intel, so ddr4 3200mhz on Ryzen might be equivalent to intel at 3400 or even 3600mhz.
Very interesting if you look at those benchmarks and try and fit a 1700 into it. :)
 

Shivansps

Diamond Member
Sep 11, 2013
3,835
1,514
136
Memory bandwidth scaling in games drops off after 3600mhz DDR4 does it not? In any case faster Memory than that is ridiculously expensive to be pairing with a 350$ cpu.

Food for thought, R7 1700 has 20mb of cache, also its IMC is more efficient at extracting maximum bandwidth out of its memory than intel, so ddr4 3200mhz on Ryzen might be equivalent to intel at 3400 or even 3600mhz.
Very interesting if you look at those benchmarks and try and fit a 1700 into it. :)

You cant compare AMD and Intel cache size, size is not everything, is about the algorithm used to access it.

About the memory i guess we will soon know it, too many dark things about the IMC lately. not sure how you could be sure its more efficient than anything whiout seeing actual reviews and benchmarks.
 

Agent-47

Senior member
Jan 17, 2017
290
249
76
any latency teething issuesare likely to make NO difference in real world applications due in part to ryzens excellent cache system.

.

The slower IMC could explain certain discrepancies: Why pure integer/FP/SSE performance is better than Intel but overall scores like PN and Physics is lower in Passmark (whose memory utilization is bigger than the cache size).

Lower latencies is pop up every time there is a cache miss.

But this does not cripple the performance which is the bottom line. And maybe its by design in order to leave a hefty room for future improvement in case Intel tries to counter attack (all they have to do is release a CPU with better IMC to turn the table)
 

Shivansps

Diamond Member
Sep 11, 2013
3,835
1,514
136
But this does not cripple the performance which is the bottom line. And maybe its by design in order to leave a hefty room for future improvement in case Intel tries to counter attack (all they have to do is release a CPU with better IMC to turn the table)

Those two statements contradict each other.
 

GroundZero7

Member
Feb 23, 2012
55
29
91
The slower IMC could explain certain discrepancies: Why pure integer/FP/SSE performance is better than Intel but overall scores like PN and Physics is lower in Passmark (whose memory utilization is bigger than the cache size).

Lower latencies is pop up every time there is a cache miss.

But this does not cripple the performance which is the bottom line. And maybe its by design in order to leave a hefty room for future improvement in case Intel tries to counter attack (all they have to do is release a CPU with better IMC to turn the table)

It isn't quite that simple. Ryzen has stupid good multithreading, and PE. You can't just make a better IMC and make a new cache system. Intel needs a new architecture to compete with Ryzen's MT.

AMD has been working on the building blocks of Ryzen, and Vega for at least 7 years. Intel can't just snap it's fingers and come up with something to compete with them.
 

french toast

Senior member
Feb 22, 2017
988
825
136
You cant compare AMD and Intel cache size, size is not everything, is about the algorithm used to access it.

About the memory i guess we will soon know it, too many dark things about the IMC lately. not sure how you could be sure its more efficient than anything whiout seeing actual reviews and benchmarks.
But this is somewhat answered with leaks is it not? Ryzen beats broadwell (slightly ) clock for clock in most games tested, if you factor that into that german benchmark database as well as the increased memory efficiency of ryzen you get an interesting conclusion :)

As 6900k is likely to be running at 3.7ghz all core in those tests and maybe 4ghz on one we can conclude that 6900k is going to be quite abit faster than a 65w cpu that probably runs at 3.2 all core turbo only, but still it would look very good in that chart.
I think over 15 games 7700k will win in 10, but when both overclocked to 4ghz and 5ghz respectively i think that will be reversed.
 

Agent-47

Senior member
Jan 17, 2017
290
249
76
Those two statements contradict each other.

I should have been more clear.

At the moment it does seem AMD is capable of keeping up with Intel's latest. If Intel chooses to release new CPU with better IPC and MT (when ever that maybe), aMD can release an improved IMC to counter it along with better IPC and improved process.
 

Shivansps

Diamond Member
Sep 11, 2013
3,835
1,514
136
What games leaks? the only one i see was the chinese 1700X vs 6800K and it was very bad result one.

Also, if a remember well, 6800 and 6900K ACT turbo is 3.5Ghz.
 
Last edited:
  • Like
Reactions: frozentundra123456

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136

Oh, no you've posted some 720p benchmarks and defeated me.

Why didn't you also post the results for 1080p, which some people actually game at?

We both already know the answer, and it's because the higher-core counts only have 5-7% instead of 40%. Care to speculate or guess what the results would look like if they posted benchmarks for 1440 or 4k? [H] had an article a while ago that similarly looked at scaling, more limited in that it only tested with a single game, but also included 1440 results.

As long as the 4-core chips enjoy a comfortable bit of OC headroom over HEDT parts, they'll have their niche in the market and Intel won't have to budge price all that much, maybe drop around $50 to ~$300 or so. It really depends on how good the R3 chips are since there'll be able to OC better given all of the dark silicon to soak up and spread the heat around and what AMD wants to sell them for.

One possibility that hasn't been discussed would be if using Ryzen Master to manually shut off 4-cores that you could get an 8-core Zen to have similar clocks in which case you can configure the processor to give you the best of both worlds and tailor it to best fit the task, but we don't know how easy it will be to do that or if it allows hitting similar OC levels.

However, the fact is that at higher resolutions, additional cores amount to barely any performance gains and that because Intel's x700K can hit higher clock speeds it will overcome that relatively small scaling advantage and give overall better results.
 

french toast

Senior member
Feb 22, 2017
988
825
136
What games leaks? the only one i see was the chinese 1700X vs 6800K and it was very bad result one.
The 1800x beats 6900k on at least 2 games shown, BF1 and sniper elite 4 (SE looked to be over 10%) the Chinese link that saw both 6800k and 1700x locked at 3.4 saw ryzen matching or leading broadwell even in ST sensitive games.
 

malitze

Junior Member
Feb 15, 2017
24
49
51
Oh, no you've posted some 720p benchmarks and defeated me.

Why didn't you also post the results for 1080p, which some people actually game at?

We both already know the answer, and it's because the higher-core counts only have 5-7% instead of 40%. Care to speculate or guess what the results would look like if they posted benchmarks for 1440 or 4k? [H] had an article a while ago that similarly looked at scaling, more limited in that it only tested with a single game, but also included 1440 results.

As long as the 4-core chips enjoy a comfortable bit of OC headroom over HEDT parts, they'll have their niche in the market and Intel won't have to budge price all that much, maybe drop around $50 to ~$300 or so. It really depends on how good the R3 chips are since there'll be able to OC better given all of the dark silicon to soak up and spread the heat around and what AMD wants to sell them for.

One possibility that hasn't been discussed would be if using Ryzen Master to manually shut off 4-cores that you could get an 8-core Zen to have similar clocks in which case you can configure the processor to give you the best of both worlds and tailor it to best fit the task, but we don't know how easy it will be to do that or if it allows hitting similar OC levels.

However, the fact is that at higher resolutions, additional cores amount to barely any performance gains and that because Intel's x700K can hit higher clock speeds it will overcome that relatively small scaling advantage and give overall better results.

Because if you want to focus on the CPU you try to take the GPU out of the equation as good as possible.