Ryzen-A Fail for Gamers?

[richaron]

So, I was right. You moved the goal post, and the crux of your argument is "Santa's shop is invisible!".

Yes, I am aware of that phrase, as I hear it often from religious philosophers. Which is why I find it ofd that the next line you use includes "scientific mind".

A person with a scientific mind wouldn't make arguments for the existence of something with absolutely no evidence. I have repeatedly asked for evidence, and your rebuttal is "have faith". Very scientific.

From the very start my only goal has been to correct your claim that low resolution benchmarks are useless.

Anyone who has followed our conversations can easily see your strawman arguments, red herrings, and most recently what they call an "argument from ignorance" trying to shift the burden of proof.

It's stupid to pretend that low resolution benchmarks has no value, how much "real world" value they may have is analogous to synthetic benchmarks and rightly debatable. But to pretend there is zero value is stupid.

 

[krumme]

An example the illustrate the problem is bm in bf1.
You can get a skl i7 to go under 60 fps in certain situations and modes. Its absolute cornercase but a zen or 6900 never gets those dips.
But when bm in normal mp 64 situations at say 1080 the 4c high freq i7 is faster avg and for frametimes.
How do we adress such issue?
Clearly lowering resolution only goes to a certain degree to spell out cpu limitations. The problem beeing that cpu limitation is far from beeing a precise term. If we omit the basic problem of testing the H way beeing expensive and hardly reproducible results it seems we need to identify where the cpu is bottleneck.

Perhaps some strict synthetic gaming test stressing all parts and showing the info is the better way to go. Its more a detailed profile and not a meaningless sum of all subtest.

Then comparing demands from engines and future engines to that bm profile will give better insight to the game you play.

Then lets imagine you set up your personal gameprofile. That profiles changes over time. Learn from you. And so does the gameengine profiler and the cpu bm profiler based on continuous new data.

In an more open source environment with transparency that could be a platform for new tech and cpu development.

 

[unseenmorbidity]

I didn't notice this being posted yet. Quite informative on the subject:

Very informative, ty.

 

[beginner99]

After this, who ever says Ryzen Gaming performance is a fail is Trolling.

R7 1800X 4C 8T @ 4GHz vs Core i7 7700K 4C 8T @ 4GHz

Again? Didn't this already get cleared up in another Ryzen thread. It's ok to prefer one or the other brands but in the end, you can't be taken seriously if you post this over and over again.

I call comparing an OCed vs. a downclocked CPU trolling. Actually the charts only proof the 7700k is in these benches far superior as it wins most of them with 1 hand tied behind it's back. Since the 1800x is OCed the 7700k should be running at least at 4.8 Ghz if not higher meaning 20% faster...

I still haven't decided if I buy 7700k, Ryzen or wait for Skylake-X. But posts like this are for sure pushing me towards Intel. So if you actually are an AMD marketing guy (as I think you are), what you are doing is counter-productive. I'm telling you this because I want AMD to succeed but I', not buying their products out of pity. They need to deliver.

 

[richaron]

I didn't notice this being posted yet. Quite informative on the subject:

I actually like this video for a number of reasons. And I have been a proponent of "moar cores" on this site for many years, the rewards of which we are only just seeing (re video).

But again, I see this dude as setting up a strawman argument.

What I mean is that there are 2 distinct arguments here:
1) whether low resolution benchmarks have value.
2) whether low resolution benchmarks can be "extrapolated" into future performance.

The maker of the video invented point #2 whether deliberately or not. They then spent the rest of the video arguing against point #2.

But there is no logical connection to point #1.

 

[tamz_msc]

I actually like this video for a number of reasons. And I have been a proponent of "moar cores" on this site for many years, the rewards of which we are only just seeing (re video).

But again, I see this dude as setting up a strawman argument.

What I mean is that there are 2 distinct arguments here:
1) whether low resolution benchmarks have value.
2) whether low resolution benchmarks can be "extrapolated" into future performance.

The maker of the video invented point #2 whether deliberately or not. They then spent the rest of the video arguing against point #2.

But there is no logical connection to point #1.

If it is not to gauge point #2, then why would you do low-resolution tests in the first place? What, according to you, is the point of low-resolution tests?

 

[Head1985]

Good test.

Ryzen is clearly faster. When I say to my friends that Cry Engine 3 is the best in terms of managing cpu resources, my friends think I am crazy!

Faster?Look at start-ryzen 80-83fps, 6700k ht on 90-93fps and 6700k ht off 70fps.When running 6700k with HT on can reach more than 100fps with avg around 80-90fps.
Faster than 6700k HT off.Slightly slower than 6700k HT ON.

Btw can you make video with 2500k?

 

[richaron]

Did they assume that though? This sounds like a strawman argument to me. But if some testers did make this claim they're jumping to conclusions.

As you should well know; low resolution benchmarks are an attempt to remove the GPU from the equation as much as possible, thus making it ~more~ a CPU vs CPU test. I never said it was particularly relevant. Just like synthetic memory bandwidth tests may not be particularly relevant to real world applications.

All I said was that to completely discount these tests because you, or someone else, doesn't understand the scope of the information provided is stupid.

In response to a claim that they run low resolution tests to "extrapolate" future performance.

Please don't make me go searching for such easy to find posts. The urge to edit & fix is painful.

Edit: I meant edit & fix spelling, grammar, & sentence structure. Not content. Don't judge me.

 

[scannall]

The thread has obviously devolved into a my (insert brand name) is better than your (insert brand name) bash fest filled with confirmation bias. Until testing is objective the subjective skew will continue to permeate the data rendering it useless. Looks like the scientific double blind testing method might be required here to remove any biases. I just want to see pure unbiased results that just present the truth rather than a pile of excuses one way or the other for the results.

It's more that some people are pushing their own use case as being all important. And that everyone should follow their assessment. The benchmarks that actually matter to *me* are the productivity ones for example. The 1080p ones don't hold any value for my own use case. But it doesn't bother me in the least if someone else's needs are different, and require them to look at different benchmarks being more important to them. The low rez twitch games probably don't care in the slightest about production or content creation benchmarks either.

Now, if people would just use the benchmarks that actually matter to them, and leave people alone who need different things this thread could move right along with some real information and benefit.

 

[richaron]

It's more that some people are pushing their own use case as being all important. And that everyone should follow their assessment. The benchmarks that actually matter to *me* are the productivity ones for example. The 1080p ones don't hold any value for my own use case. But it doesn't bother me in the least if someone else's needs are different, and require them to look at different benchmarks being more important to them. The low rez twitch games probably don't care in the slightest about production or content creation benchmarks either.

Now, if people would just use the benchmarks that actually matter to them, and leave people alone who need different things this thread could move right along with some real information and benefit.

^ 100% agree.

Take what you need/want from all the information provided.

But don't yell "Fake!" at something you don't like. Or someone will most likely correct you.

 

[french toast]

Err... I'm as pro AMD as anyone but you proved your point by underclocking the 7700k and maxing out the OC on the 1800x? What does that prove again?

It dispels the fud that some repeatedly spew on this forum that ryzen has sub par "gaming ipc".
Clock for clock and core for core ryzen is within 5% of skylake, ryzens problems with inter ccx communication when not necessary, ie windows and games being numa aware.
Im quite sure atenra didnt mean r7 1800x is better at current gaming whilst comparing a downclocked 7700k, that would be stupid and i feel hes smarter than that.

 

[french toast]

Nope I said it was stupid to completely discount low resolution benchmarks. You've obviously confused these results in the past so I decided to correct you.

I didn't state any case where these benchmarks were relevant or irrelevant; again I said it was stupid to discount these values as meaningless. Just because you, or other people, don't understand the scope of this information doesn't mean it's useless.

We have been over this a thousand times already, it all depends on what you take from certain benchmarks and what conclusion you are trying to draw.
720p gaming benchmarks have a place as they stress the cpu, but these benchmarks should be treated as synthetic like cinebench and not used to draw conclusions future performance, for that purpose they are completely useless, threads are the best indicator of future performance not 720p gaming.

Low res benchmarks do have a purpose as a synthetic test for geeks and corner cases like ultra fps competitive gaming, perhaps for niche scenarios where a consumer want to play the same game for 4 years whilst planning a gpu buy.
Reviewers have been using such low res benchmarking to draw the wrong conclusions about future performance, unless testing equally threaded cpus its basically spreading mis information.

 

[unseenmorbidity]

You know why low res tests are interesting?
Because benchmarkers don't bother to search for the spots that have the highest cpu demands, seeing the difference in low res tests shows you which cpu has less chances of dropping to low fps when hitting such a spot even if you are playing at higher resolutions.

So, you want a bad benchmark done, so you can try to extrapolate whether there will be potential performance drops? Why not just look for people doing proper benchmarks of the game?

This all depends on the person doing the benchmark, and whether there is a built in benchmark.

Gamer's Nexus uses the notoriously awful gta benchmark. He also supposedly test watchdogs2 by walking up and down an empty alley, and looking at the sky.

Jayztwocents runs the first mission on gta. Computerbase.de plays bf1 in multiplayer. Joker test watchdog by playing the game.

 

[PotatoWithEarsOnSide]

Faster?Look at start-ryzen 80-83fps, 6700k ht on 90-93fps and 6700k ht off 70fps.When running 6700k with HT on can reach more than 100fps with avg around 80-90fps.
Faster than 6700k HT off.Slightly slower than 6700k HT ON.

Btw can you make video with 2500k?

What did you have motion blur set as in each test?
I ask because there looked like a fair amount of motion blur on the 6700k footage, but less on the R7.

edit: watched it back, and both seemed to have the same level of motion blur.

 

[krumme]

If it is not to gauge point #2, then why would you do low-resolution tests in the first place? What, according to you, is the point of low-resolution tests?

Smack down. Haha. Incredible this logical way to adress the basic of this problem havnt been forwarded so short to the point before. But it shows how used we are to that way of benchmarking.

 

[Head1985]

What did you have motion blur set as in each test?
I ask because there looked like a fair amount of motion blur on the 6700k footage, but less on the R7.

edit: watched it back, and both seemed to have the same level of motion blur.

settings are in both videos at beginning.Its same setting he used with ryzen.

 

[AtenRa]

Again? Didn't this already get cleared up in another Ryzen thread. It's ok to prefer one or the other brands but in the end, you can't be taken seriously if you post this over and over again.

I call comparing an OCed vs. a downclocked CPU trolling. Actually the charts only proof the 7700k is in these benches far superior as it wins most of them with 1 hand tied behind it's back. Since the 1800x is OCed the 7700k should be running at least at 4.8 Ghz if not higher meaning 20% faster...

I still haven't decided if I buy 7700k, Ryzen or wait for Skylake-X. But posts like this are for sure pushing me towards Intel. So if you actually are an AMD marketing guy (as I think you are), what you are doing is counter-productive. I'm telling you this because I want AMD to succeed but I', not buying their products out of pity. They need to deliver.

For the second time, this is apples to apples, clock to clock. I havent said this is Default vs Default and its very clear in the title. You are free to choose what ever you want but when clock to clock and number of Cores/Threads to number of Cores/Threads Ryzen is so close to SkyLake then calling Ryzen Gaming performance a Fail is trolling.

As for the AMD Marketing BS keep it to yourself, i havent said that 4GHz 4C/4T Ryzen is equal to Core i7 7700K but it sure is extremely close clock to clock and Cores/Threads to Cores/Threads against the Skylake.
And between you and me, your reaction to my post would have made me believe you are the marketing guy that was irritated by the fact that a 4GHz 4C/8T Ryzen CPU at half the price being so close to a Intel 4GHz 4C/8T Skylake.

 

[AtenRa]

It dispels the fud that some repeatedly spew on this forum that ryzen has sub par "gaming ipc".
Clock for clock and core for core ryzen is within 5% of skylake, ryzens problems with inter ccx communication when not necessary, ie windows and games being numa aware.
Im quite sure atenra didnt mean r7 1800x is better at current gaming whilst comparing a downclocked 7700k, that would be stupid and i feel hes smarter than that.

That is exactly what i was trying to illustrate, i thought it was obvious by the title but it seems that some people dont really like to see clock to clock and core/threads to core/threads comparisons any more.

 

[Puffnstuff]

The one certain thing with Ryzen is that it is a fail with AMD investors. This is exactly what a company on the rebound doesn't need after a new product release.

Trolling and off-topic are not allowed
Markfw
Anandtech Moderator

 

[Head1985]

For the second time, this is apples to apples, clock to clock. I havent said this is Default vs Default and its very clear in the title. You are free to choose what ever you want but when clock to clock and number of Cores/Threads to number of Cores/Threads Ryzen is so close to SkyLake then calling Ryzen Gaming performance a Fail is trolling.

As for the AMD Marketing BS keep it to yourself, i havent said that 4GHz 4C/4T Ryzen is equal to Core i7 7700K but it sure is extremely close clock to clock and Cores/Threads to Cores/Threads against the Skylake.
And between you and me, your reaction to my post would have made me believe you are the marketing guy that was irritated by the fact that a 4GHz 4C/8T Ryzen CPU at half the price being so close to a Intel 4GHz 4C/8T Skylake.

Nice comparison, but they should use faster memory.2400mhz is just huge bottleneck for kabylake.Also testing tomb raider with benchmarks is not best thing.Only geothermal valley is heavy on cpu(and there kaby is 19% faster)

Btw its not bad at all.Its around haswell IPC.

 

[krumme]

The one certain thing with Ryzen is that it is a fail with AMD investors. This is exactly what a company on the rebound doesn't need after a new product release.

Yeaa the share price is back where its was the 17 february and before that 2007.
2007. Nearly 10 years ago.
It have gone up like 500% the last year.
A fail for investors?
Or was it a fail at trolling ?

 

[cytg111]

The one certain thing with Ryzen is that it is a fail with AMD investors. This is exactly what a company on the rebound doesn't need after a new product release.

I dont see a problem here? Looking further back it seems as volatile as always with the exception that with a little linear regression that is a good for one steep climb .. No way in beep you can conclude its levelling off by that graph alone.

 

[rgallant]

^ 100% agree.

Take what you need/want from all the information provided.

But don't yell "Fake!" at something you don't like. Or someone will most likely correct you.

re low res bench marks
why go low when you can go high res testing
whats missing in the low res bench marks are real world high end gaming imo

2 gtx 1080ti's in sli @ 4k is what is Now and in the future as gtx 1080ti [and maybe vega] become mid/high range cards upon next releases of higher end cards.
so gtx 1080 ti sli won't be a gpu bottle neck in today's games that use sli for testing and will load up the base system for a real look at gaming with the games we have today with today's and tomorrows cards.

but really zen needs some time just as every intel new ack. release needed

 

[Puffnstuff]

I dont see a problem here? Looking further back it seems as volatile as always with the exception that with a little linear regression that is a good for one steep climb .. No way in beep you can conclude its levelling off by that graph alone.

Or was it a fail at trolling ?

So a $648 million stock dump is okay?

So the first warning was not taken seriously ? Try for number 3 if you want a vacation
Trolling and off-topic are not allowed
Markfw
Anandtech Moderator

 

[Rifter]

It's more that some people are pushing their own use case as being all important. And that everyone should follow their assessment. The benchmarks that actually matter to *me* are the productivity ones for example. The 1080p ones don't hold any value for my own use case. But it doesn't bother me in the least if someone else's needs are different, and require them to look at different benchmarks being more important to them. The low rez twitch games probably don't care in the slightest about production or content creation benchmarks either.

Now, if people would just use the benchmarks that actually matter to them, and leave people alone who need different things this thread could move right along with some real information and benefit.

I agree, and im in the same boat. Bottom line is i just dont have the time to game that i used to, and when i do its at 1440p so a potato of a CPU would work for that res as you are GPU limited 100% of the time. What i do spend more time doing now is photo/video work in which ryzen absolutely kills intel, i can now build a full ryzen system for less than the cost of a equivalent intel CPU, and it performs as well as the intel in what i do.

Ive said this before in ryzen threads, if you just game and treat your PC like its a console get a 7700k and dont look back. If you actually use your PC as a PC and run multiple types apps on it then ryzen starts to meet your needs better than intel for a fraction of the price. Its up to each individual to decide what they need for themselves.

I dont understand why this is so hard for people to grasp that people use their PC's for a variety of tasks, this is a enthusiast forums we should know better than most the wide array of things that can be done with a PC.