Chinese site shows extensive benchmarks results of AMD Ryzen

[vorticalbox]

1 hour ago, MageTank said:

You cannot use the words "at least" when talking about overclocking. The sample size is too low. I don't care if MSI overclocked to 5ghz, we don't know anything yet. 

agreed. My 8350 will do 5ghz with 1.47 volts, I've seen people not about to get past 4.5 on 1.5 volts.

[Guest Darth Revan]

6 minutes ago, done12many2 said:

 

The majority of stuff currently available.  Single-threaded impacts perceived user interface a great deal more than multi-threaded.  My 5950x at 4.6 GHz gets smoked by my 7700k @ 5.3 GHz in almost everything, but the very few tasks that are effective at taking advantage of all cores at one time.  It's cool to see the 5960x crush tasks when all cores are firing, but I don't see it happen too often at all.

 

Things in the area of multi-threading are improving, but it is painfully slow.  So much so that an 8 core processor that I've had for almost 2 years is still barely utilized to it's potential. 

 

 

So basically the problem is that developers are too lazy to make their games or programs work with multi-core.

[MageTank]

6 minutes ago, done12many2 said:

 

The majority of stuff currently available.  Single-threaded impacts perceived user interface a great deal more than multi-threaded.  My 5950x at 4.6 GHz gets smoked by my 7700k @ 5.3 GHz in almost everything, but the very few tasks that are effective at taking advantage of all cores at one time.  It's cool to see the 5960x crush tasks when all cores are firing, but I don't see it happen too often at all.

 

Things in the area of multi-threading are improving, but it is painfully slow.  So much so that an 8 core processor that I've had for almost 2 years is still barely utilized to it's potential.  

 

For day to day use and for most gaming, single-threaded performance should remain an area to focus on.  

 

 

It's just going to be a matter of time and folks will see that TDP is just a play on words.  Anyone who paid enough attention to the AMD press event where they showed power draw during testing could see that at load they were essentially drawing the same power.  The total system draw at idle was lower during those examples, but as soon as load hit they look like twins.  

People often forget the T in TDP stands for "thermal". They automatically think TDP equates to actual power consumption, but it does not. Sometimes it's close, but others, it's off by a long shot. It's also certainly not the max a chip can pull. As shown above, I can pull 130w (123w in that specific screenshot after only 3 mins of 48k FFT) with relative ease. That's a massive 40% (give or take) difference if you are actually comparing power consumption to TDP. This is also me running stock clocks AND undervolted. Imagine if you actually overclocked, or threw more volts at it. I have friends that push 180w on their 4790k's. 

[79wjd]

Just now, Darth Revan said:

So basically the problem is that developers are too lazy to make their games or programs work with multi-core.

That and at the end of the day, certain tasks can only be parallelized so much. A lot of things can't be parallelized as you're dependent on information that hasn't yet been computed.

[Thony]

AMD looking real strong ! 

[Cela1]

Can we just stop with the leaks please?

 

...Seriously, just wait 3 days....

[Coaxialgamer]

I'm more interested by this tbh :

 

I did a bit of napkin math to try and approximate the clock speed of this chip .

Assuming a linear increase in clock speeds and thread count vs performance ( which doesn't really happen , but bear with me )

- Using the leaked 1600x scores ( assuming these are accurate of course) we get a score of 12544 across 6c/12t at 3.56ghz

-We increase that by 25% ( to account for the extra cores/ threads of an 8 core chip ) . That gives us around 15680 points on an 8 core / 16thread chip at 3.56ghz .

-We then do 19874 / 15680 = 1.267 , or a 26.7% increase in frequency ( again , i know it doesn't QUITE work that way) .

-A 26.7% increase on 3.56ghz puts us around 4.5ghz , similar ( albeit at the higher end ) to what rumors are saying . In practice though , it probably needs to be clocked higher to reach this performance , meaning this is probably on water or even phase change/ln2 if previous OC reports are reporting .

 

But if this is at 4.5ghz , single core scores are impressive considering the 7700k also runs at 4.5ghz on a  single core .

 

[vorticalbox]

1 hour ago, goodtofufriday said:

To my understanding these are all OC'd results.

then why include a non-oc cpu?

[ivan134]

27 minutes ago, Cela1 said:

Can we just stop with the leaks please?

 

...Seriously, just wait 3 days....

You're not my real dad.

[Master Disaster]

1 hour ago, Ryan_Vickers said:

I just saw the posts that came in between mine above and the ones half way down the first page which didn't show up for some reason.  My post was based on the assumption that the 7700k in this chart was at stock.  I now see  it was at 4.9 GHz.  This just got even more interesting.

Did you misread something? The 7700K in the chart is at stock, the one at 4.9Ghz scores 2480pts which is more than the 1700X.

 

http://valid.x86.fr/bench/de1va2/1

[vanished]

2 minutes ago, Master Disaster said:

Did you misread something? The 7700K in the chart is at stock, the one at 4.9Ghz scores 2400pts which is more than the 1700X.

Oh, ok. ... that makes a lot more sense actually.

 

Well, in some ways... in others it raises more questions:

 

I assume the stock 7700k in the chart - the one that scored 2301 - was running at 4.2 GHz.  You're saying that a boost to 4.9 GHz (16.7% more) only brings that up to 2400 (4.3% more)?  Even if we assume the 7700k was at its max boost of 4.5 GHz in this test (highly unlikely), that's still a 8.9% boost - roughly twice the measured increase in performance.  That does not make sense.

[Master Disaster]

1 minute ago, Ryan_Vickers said:

Oh, ok. ... that makes a lot more sense actually.

 

Well, in some ways... in others it raises more questions:

 

I assume the stock 7700k in the chart - the one that scored 2301 - was running at 4.2 GHz.  You're saying that a boost to 4.9 GHz (16.7% most) only brings that up to 2400 (4.3% more)?  Even if we assume the 7700k was at its max boost of 4.5 GHz in this test (highly unlikely), that's still a 8.9% boost - roughly twice the measured increase in performance.  That does not make sense.

Its 2480pts to be exact

 

http://valid.x86.fr/bench/de1va2/1

 

What's interesting is the 1700X still beats the OCed 7700K in MTP though.

[vanished]

Just now, Master Disaster said:

Its 2480pts to be exact

 

http://valid.x86.fr/bench/de1va2/1

ok, again, 2480 / 2301 = 7.8%

 

4.9 / 4.2 = 16.7%

 

It just doesn't add up.  The score should have improved to more than 2480

[MageTank]

20 minutes ago, Ryan_Vickers said:

ok, again, 2480 / 2301 = 7.8%

 

4.9 / 4.2 = 16.7%

 

It just doesn't add up.  The score should have improved to more than 2480

This is assuming the test scales linearly with clock speed, is it not? Can anyone confirm that a 10% increase in clock speed directly correlates to a 10% increase in performance? I could test this, but my ram is starting to flake out on me at the moment, and I am busy benching it (will skew the results heavily). 

[vanished]

14 minutes ago, MageTank said:

This is assuming the test scales linearly with clock speed, is it not? Can anyone confirm that a 10% increase in clock speed directly correlates to a 10% increase in performance? I could test this, but my ram is starting to flake out on me at the moment, and I am busy benching it (will skew the results heavily). 

Yes, this is assuming that, but I see no reason why this wouldn't hold up unless you started pushing into fringe scenarios where other stuff like cache speed and quantity starts significantly interfering 

[vanished]

59 minutes ago, Coaxialgamer said:

I'm more interested by this tbh :

*snip*

 

I also arrived at 4.5 GHz for my guess but I did it a much more simple way.  Looking at the AMD slide comparing the 6900k and 1800X matching each other exactly in single threaded, and knowing that was likely a case of 3.7 GHz on the 6900k and 4 GHz on RyZen, I figured that to basically match the 7700k in single threaded the way they did, RyZen would have to once again be ~300 MHz faster than the 7700k's 4.2 GHz base speed.

[Coaxialgamer]

4 minutes ago, Ryan_Vickers said:

I also arrived at 4.5 GHz for my guess but I did it a much more simple way.  Looking at the AMD slide comparing the 6900k and 1800X matching each other exactly in single threaded, and knowing that was likely a case of 3.7 GHz on the 6900k and 4 GHz on RyZen, I figured that to basically match the 7700k in single threaded the way they did, RyZen would have to once again be ~300 MHz faster than the 7700k's 4.2 GHz base speed.

i also though about doing that , but considering the nature of IPC ( integer , FP , physics , AVX etc )  and how zen compares to broadwell , i thought it would be better just to compare it to a chip with the same architecture .

[AnonymousGuy]

OC'd vs. non-OC'd is completely fucking useless.  No one buys a 7700K and then leaves it stock.

[Brehohn]

Why is everyone getting so bent out of shape

 

Here's my take on this. 

 

Isn't it strange that when you click "Back To Validation" that it just takes you to the damn menu instead of the person's benchmark? How could a person even control that? Also, WHO CARES?? If it's beating the damn stock i7 in single threaded being overclocked, THEN GREAT. When is the last time we've had this happen? A LONG TIME AGO. It's a VALID benchmark. If it wasn't valid, CPU-Z would have deleted or denied it. Multi-threaded is amazing. Single threaded is a great feat. 

 

End of story.

[MageTank]

6 minutes ago, Brehohn said:

Why is everyone getting so bent out of shape

 

Here's my take on this. 

 

Isn't it strange that when you click "Back To Validation" that it just takes you to the damn menu instead of the person's benchmark? How could a person even control that? Also, WHO CARES?? If it's beating the damn stock i7 in single threaded being overclocked, THEN GREAT. When is the last time we've had this happen? A LONG TIME AGO. It's a VALID benchmark. If it wasn't valid, CPU-Z would have deleted or denied it. Multi-threaded is amazing. Single threaded is a great feat. 

 

End of story.

The validation link was likely pulled by AMD requesting for it to be pulled. Either way, we won't really know for certain until we actually get our hands on these CPU's. 

[MageTank]

28 minutes ago, Ryan_Vickers said:

Yes, this is assuming that, but I see no reason why this wouldn't hold up unless you started pushing into fringe scenarios where other stuff like cache speed and quantity starts significantly interfering 

Just canceled my ram tests to test this (had to load XMP for these tests, which is blasphemy to a man like myself). 4.2ghz scored 2110 SC, 9081 MC. At 4.5ghz, I scored 2261 SC, 9699 MC. Now, 4.5ghz is exactly 7.15% higher than 4.2ghz. 2261 is exactly 7.15% higher than 2110. I would say linear scaling is real in this benchmark. Interestingly enough, the multi-core benchmark doesn't scale as perfectly, as I only gained 6.8% this time around, but close enough. 

 

This makes this result even more interesting if it's real, and should aid in the previous "paper math" calculations. 

[vanished]

Just now, MageTank said:

 

Nice, thanks a bunch!  

[Simon771]

Better than 10 core from Intel?

Better in single core performance than i7 7700k?

 

I'm sorry, but I just won't belive this. I know how this goes ... I get my hopes up, and then when it's released my dreams are crushed.

[goodtofufriday]

3 minutes ago, Simon771 said:

Better than 10 core from Intel?

Better in single core performance than i7 7700k?

 

I'm sorry, but I just won't belive this. I know how this goes ... I get my hopes up, and then when it's released my dreams are crushed.

Its a Validated score, can't be faked. The real question is what cooling was used to achieve these scores.

[Simon771]

Just now, goodtofufriday said:

Its a Validated score, can't be faked. The real question is what cooling was used to achieve these scores.

Must have been LN2 and clocked at 5,0GHz or above.