AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[sm625]

Here are a bunch of benchmarks. These are taken directly from passmark PerformanceTest 9.0. Note: Actual clocks are not listed on these graphs. The true clocks are as follows:

The good news is that turbo for the Ryzen is apparently disabled.

Without a source, I am not going to allow these "benchmarks"
Please add a source.
Markfw
Anandtech Moderator

 

[Rngwn]

Here are a bunch of benchmarks. These are taken directly from passmark.

*snip*

Wait... Why is the 5960x did better than 6900k in everything? That just doesn't make sense or is it?

And Zen did actually worse than the Bulldozer in Physics and Prime... I don't even

 

[lolfail9001]

These are taken directly from passmark.

If only these had actual clocks to go with them. Because we are now supposed to believe that 9590 has twice the single thread performance of fx-8350, something that is actually bull.
But assuming that 5960X is at 4.4Ghz (from your Pentium sample), Ryzen actually has better integer performance than 5960X, how about that?
Assuming that Primes and Physics use 256-bit AVX, Ryzen's performance is also in line with what we know (half the throughput of Haswell and higher in SIMD).
The only new piece of info is general fp performance and from napkin math i see that it is basically equal with Haswell as well.
These are all very good benches, man, you could not do any better advertisement.

EDIT: Baselines provided, only thing remaining is to establish the hype train. Choo-choo, mo^Wpeople.

 

[.vodka]

For f...s sake man, what's the source? Where are you getting them from? What validates them? What are those baselines you're using? Are they overclocked results? It doesn't help when you say "they're taken directly from Passmark" and then link to a random image hosting site with a bunch of graphs that look like they're made on MS Paint. You've deceived us a few posts ago with a 4.4GHz G3258 not mentioning it was overclocked in your first post. I can continue listing irregularities.

Take your new single threaded picture. 5820k scoring 100 points behind a 7700k. Excuse me, I don't believe a, let's say, 4.5GHz 5820k is only 100 points behind a 4.5GHz stock turbo 7700k.

Physics is probably a very easily multi threadable benchmark. I don't believe for one second that 8c16t Ryzen, architecturally, unequivocally superior to 4M8T Piledriver, is scoring 2 points less. Not even with that CPU clocked to >5GHz.


Passmark may be a legit benchmark suite... but sorry, there are many unknowns here to take your results or whatever place you're getting them from as legit.


If you want us to take you seriously start providing links to the baselines in question so we can see the whole picture. Get us a link to that Ryzen baseline ASAP.


edit: baseline numbers for all the results, thanks. Will have a look.

 

[Doom2pro]

Wait... Why is the 5960x did better than 6900k in everything? That just doesn't make sense or is it?

And Zen did actually worse than the Bulldozer in Physics and Prime... I don't even

Those aren't average scores for those CPUs, they are individually uploaded baselines, by anyone... For all we know those systems weren't performing as best as they could if not better than normal (overclock), again seems cherry picked.

 

[lolfail9001]

Excuse me, I don't believe a, let's say, 4.5GHz 5820k is only 100 points behind a 4.5GHz stock turbo 7700k.

Passmark is very IPC improvement insensitive, 4.5Ghz 5820k could be exactly the same as 4.5Ghz stock turbo 7700k.

Physics is probably a very easily multi threadable benchmark.

Do we know if it uses AVX? It perfectly explains the numbers i, personally, see.

If you want us to take you seriously start providing links to the baselines in question so we can see the whole picture.

Links do not list their actual frequency anyways. I am only curious about where he takes these from.

 

[sandorski]

If everything goes according to rumour, in a month we will know.

 

[Doom2pro]

Links do not list their actual frequency anyways. I am only curious about where he takes these from.

Well the images show the baseline ID numbers, but those are one time individual baselines not average scores for every CPU of that model, and from who uploaded that individual baseline, what clocks were they like you said, what memory configuration, what was running in the background, etc...

Hell I could underclock any CPU, gimp the memory and leave a bunch of stuff running in the background and then upload a baseline and use it to compare vs any other baseline.

 

[.vodka]

Passmark is very IPC improvement insensitive, 4.5Ghz 5820k could be exactly the same as 4.5Ghz stock turbo 7700k.

Then it's not a good measure of single threaded performance if it isn't sensitive to IPC increases as a result of architectural improvements. The other parts of the benchmark suite could be useful, though. Even with that into consideration, Ryzen is scoring quite well here vs Skylake.

Do we know if it uses AVX? It perfectly explains the numbers i, personally, see.

Good point. If it uses AVX then we could be seeing the result of AMD using 128bit FMA units and buses vs Intel's full fat 256bit units and buses.

Links do not list their actual frequency anyways. I am only curious about where he takes these from.

Well, we may not know frequencies of those baselines, but at least knowing where he's taking these from could really help.

 

[lolfail9001]

Well the images show the baseline ID numbers, but those are one time individual baselines not average scores for every CPU of that model, and from who uploaded that individual baseline, what clocks were they like you said, what memory configuration, what was running in the background, etc...

Well, the guy just did provide actual baseline list. I am fully on hype train from this moment on, Ryzen is a bomb.

 

[sm625]

Those aren't average scores for those CPUs, they are individually uploaded baselines, by anyone... For all we know those systems weren't performing as best as they could, again seems cherry picked.

It is cherry picked. The only way to produce these graphs is to pick baselines from a list of thousands. But anyone can do this! The software is free.

 

[Doom2pro]

The good news is that turbo for the Ryzen is apparently disabled.

Shades of all the Engineering Samples.

 

[.vodka]

<redacted "benchmarks without a source">
Markfw
Anandtech Moderator

Thanks. So you're using a 4.7GHz 5960x, a 4.2GHz 6900k, a 4.8GHz 5820k, and a 5GHz 7700k as baselines.

So.... 3.4GHz Ryzen is putting up those numbers vs that competition?

Holy crap.

Or passmark doesn't know how to read turbo on Ryzen so it's reporting as N/A and it's actually boosting to 3.8GHz depending on the workload. Or the rumors that turbo is broken are true and that feature will get enabled soon as the last remaining bugs are ironed out.

 

[Doom2pro]

It is cherry picked. The only way to produce these graphs is to pick baselines from a list of thousands. But anyone can do this! The software is free.

And did you pick those baselines based on well represented means that match what most people are going to score with respective hardware or did you just randomly pick anyone's baseline for comparison?

 

[lolfail9001]

And did you pick those baselines based on well represented means that match what most people are going to score with respective hardware or did you just randomly pick anyone's baseline for comparison?

Randomly, considering that this 4.7Ghz 5960X does as well as his 4.4Ghz Pentium.

 

[IllogicalGlory]

Yeah, if that's at 3.4GHz, then it's faster per-clock than Haswell (3.7GHz boost i5-4590), but only very slightly (1.8%).

The multithread results clearly have OC'ed Haswell chips in them, given that the 5960X beats the 6900K by a fair amount.

if we assume that the 6900K is stock, then a 3.8/4.2GHz Ryzen should be pretty far ahead of a 5960X out of the box. It would be almost exactly equal to a 6900K. (16521/14807 * 3.4/3.8) = 6900K / Ryzen now * Ryzen now / Ryzen at launch = 6900K / Ryzen at launch = 99.8%.

If that's a 4.2GHz 6900K, then Ryzen is a freaking monster, at least not considering the poor specialized results. Again, assuming it's not someone with a 6900K playing games with us.

 

[sm625]

And did you pick those baselines based on well represented means that match what most people are going to score with respective hardware or did you just randomly pick anyone's baseline for comparison?

I picked the 3rd fastest from the last month. It would take some effort to sift through the baselines to find a good mean/median for each SKU. I hope someone will do it, but it's not going to be me.

 

[lobz]

sm625, after so many years of reading, I had to register to write this down, how incredibly epic of a fail this was, what you've done with these benchmarks.
according to all the numbers you've provided - against all your intentions - Ryzen is one of the greatest achievements of AMD to date, being financially constrained, on an inferior process with smaller cores, well.... WOW at these scores!!! Yep, even if it's actually @ 3.8

 

[Doom2pro]

I picked the 3rd fastest from the last month. It would take some effort to sift through the baselines to find a good mean/median for each SKU. I hope someone will do it, but it's not going to be me.

Well there's your problem

 

[Doom2pro]

sm625, after so many years of reading, I had to register to write this down, how incredibly epic of a fail this was, what you've done with these benchmarks.
according to all the numbers you've provided - against all your intentions - Ryzen is one of the greatest achievements of AMD to date, being financially constrained, on an inferior process with smaller cores, well.... WOW at these scores!!! Yep, even if it's actually @ 3.8

I'm sitting here with my beer and just enjoying this, but wow to pull a multiple year lurker out of the void, that's pretty impressive

 

[.vodka]

If Ryzen packs this kind of a punch at 3.4GHz or even considering 3.8GHz, with the leaked SKUs with turbo or base clocks well into the 3.7-4GHz range at the rumored $300-500 range, then this Intel note posted by Kyle Bennett @ HardOCP

is that much more hilarious now. If it wasn't obvious enough when it surfaced, it's full damage control, with the added touch of people jumping out of the burning ship after seeing those results. Bloodbath it is, at least for Intel's precious monopolistic fat margins. Apart from that, this completely validates AMD's results at the new horizon event with a 3.4GHz fixed Ryzen equaling a stock turbo enabled 3.2/3.7GHz 6900k. What would happen if we were to put up that Ryzen baseline vs stock Intel processors? Ouch. We know the number of that baseline... once it shows up in their database, we can validate.

Gotta love competition, and what a shoestring budget with a few bright engineers left unconstrained can do for a product. AMD wanted that lucrative server market? Yup, that 0% share they hold right now will go up quite quickly... Not only that, imagine these cores in the APUs powering future consoles. Gaming future sure is bright, all that CPU power will do wonders!


It's nice to catch some fresh air after all these many years. The leaks can't be stopped now, only 18 days to go...

 

[lolfail9001]

Yup, that 0% share they hold right now will go up quite quickly...

Eh, their HPC share is not going to grow with such AVX performance (that statement is independent from these benchmarks because we do not know if they use AVX).

 

[Asterox]

<redacted "benchmarks without a source">
Markfw
Anandtech moderator


All this fact and all photos, you should have noted and post in your first Passmark bench post.So you deliberately created unnecessary confusion.You're lucky that I'm not a moderator on this forum, you deserve at least 10 days ban no doubt.

you are not a mod, don't try and act like one.
Markfw
Anandtech Moderator

 

[Magic Hate Ball]

All this fact and all photos, you should have noted and post in your first Passmark bench post.So you deliberately created unnecessary confusion.You're lucky that I'm not a moderator on this forum, you deserve at least 10 days ban no doubt.

 

[JDG1980]

Eh, their HPC share is not going to grow with such AVX performance (that statement is independent from these benchmarks because we do not know if they use AVX).

HPC is only a small part of the server market. For most server loads, AVX performance is virtually irrelevant. Downplaying it in Zen was a smart move since it lets AMD keep die size down, thus allowing them to beat Intel on price while still maintaining healthy profit margins.

AMD probably figures that HPC and other massively parallel loads are best handled by GPUs. With Vega, they're going to try to whittle away at Nvidia's dominance here. The biggest challenge will be CUDA; thanks to that, even Intel with Xeon Phi hasn't made many inroads.