AMD Ryzen 8C/16T Cinebench R15 & Fritz Chess Scores

This was already shown as a fake 'news' story. and at least one aggregator has since removed it from their front page news. Of course if all the site cares about is 'engagements' and advertising dollars they won't remove it.

Personally I have no stick in the AMD v Intel game. I really could care less what CPU is in most of my PCs as long as I know it will do the job I need it to do. But, that said, I enjoy a climate of healthy competition as it might lead to breaking Intel's current cycle of releasing 'new' chips with only miniscule gains in performance.
 
What I don't understand is why anyone would give credence to a leak that doesn't even show the CPU info. If I posted a cropped cinabench score with NO additional information, would it spark a flame war too? He could have easily showed the system info but chose to crop it, why? I was reading in hopes of finding some sort of confirmation...but nothing.

Not worth the discussion guys. Move along.
 
Okay so lets assume a bad position and pose a question, if AMD delivers performance like a SB/IB in respect of IPC.

1) Does that keep AMD competitive?
2) Does it keep AMD in business over the next 4 years or so?

Then if we assume a positive outcome, say AMD matches Haswell/Devils Canyon IPC.

1) How does that affect the market?
 
I believe it. I was pretty let down by Bulldozer / pile driver and ive seen pretty weak ipc updates to FM2+ as ive had 8120, 8320s, llano, 5800, 6800, 7870k -- ill get a zen too, but im just not expecting to be impressed this time around.
 
After playing around with some memory configurations on my Haswell HTPC I was able to find that timings, SPD's and with a friends X79 found that memory setup can add increments to Cinebench scores from 15-60 odd points. Granted that the i7 5960X was tested in its best environment fastest configuration and posted 1400 flat shows that with AMD still needing to set clocks, set turbo, set the native Memory support, all in all look more like AMD will achieve Haswell/Devils Canyon IPC's.
 
Broadwell and Skylake is only about 3% and 5% faster than Haswell in Cinebench IPC.

You shouldn't set your expectations that high.

I'm going to spent $250 USD on a Broadwell-EP 14 core Xeon with all core Turbo of 2.5Ghz. It should be really close to a 3Ghz Sandy in IPS. I think I'll only regret my purchase if Zen is Haswell level and can OC to over 4.5Ghz.
 
Broadwell and Skylake is only about 3% and 5% faster than Haswell in Cinebench IPC.

You shouldn't set your expectations that high.

I'm going to spent $250 USD on a Broadwell-EP 14 core Xeon with all core Turbo of 2.5Ghz. It should be really close to a 3Ghz Sandy in IPS. I think I'll only regret my purchase if Zen is Haswell level and can OC to over 4.5Ghz.

Actually it is 10-12% if you take a 6900K at 1565 vs a 5960X at 1400
 
Just some baseline scores on CB15

http://wccftech.com/intel-core-i7-5...hip-processor-benchmarked-overclocked-40-ghz/

Intel-Core-i7-5960X-MaxxMem-Performance-635x357.png



i7 5960X scores 1398

Since the source is Coolaler I am comfortable with the results of his benches.

1398*3/3.5 = 1198 - base clock IPC over multi cores

Clock for Clock:

i7 5960X: 1398*3.1/3.5 = 1238

Ryzen at this stage in its ES state is inside 3-4% of Haswell EP

Broadwell and Skylake are a bit to far away but Haswell was the more realistic goald to achieve

i7 6900K scores 1556

i7 6900K @3.1Ghz: 1556*3.1/3.7 = 1303

Ryzen is about 8-9% slower clock for clock than a Broadwell EP
 
Just some baseline scores on CB15

http://wccftech.com/intel-core-i7-5...hip-processor-benchmarked-overclocked-40-ghz/

Intel-Core-i7-5960X-MaxxMem-Performance-635x357.png




i7 5960X scores 1398

Since the source is Coolaler I am comfortable with the results of his benches.

1398*3/3.5 = 1198 - base clock IPC over multi cores

Clock for Clock:

i7 5960X: 1398*3.1/3.5 = 1238

Ryzen at this stage in its ES state is inside 3-4% of Haswell EP

Broadwell and Skylake are a bit to far away but Haswell was the more realistic goald to achieve

i7 6900K scores 1556

i7 6900K @3.1Ghz: 1556*3.1/3.7 = 1303

Ryzen is about 8-9% slower clock for clock than a Broadwell EP
that's oc'd to 4GHz though. its still to early, nobody has one in hand.
 
All I care about is how does it play games and can it encode on the fly while gaming at the same time.
 
that's oc'd to 4GHz though. its still to early, nobody has one in hand.
Intel-Core-i7-5960X-MaxxMem-Performance-635x357.png


I saw them bench it to 1398 though if you do the math at 4Ghz 1508*3.5/3 = 1314 and that changes things a lot though lets assume 1398 that is still very impressive to be wtihin a tweak of Haswell EP
 
2f7b6a40_Uncore47-2400mem.PNG


To scale it back to base clock 1944*3/4.8 = 1215

Scaled to 3.1Ghz 1944*3.1/4.8 = 1255

Scaled to 3.4Ghz 1944*3.4/4.8 = 1377 (ZEN ES 1303)

Still roughly about 4% or so. With some tweaks they can close that up a couple percentage points.
 
Getting very pedantic about results and false results especially the once places like Anandtech have on their database that are blatent lies.

Just a bunch of images:

1) Coolaler http://wccftech.com/intel-core-i7-5...hip-processor-benchmarked-overclocked-40-ghz/

Intel-Core-i7-5960X-MaxxMem-Performance.png

The CPU is running 3.3Ghz with DDR4 2133 to achieve 1270 so its base clock IPC is 1154

2) Emiliano B's HWBot http://hwbot.org/submission/2644437_emiliano_b_cinebench___r15_core_i7_5960x_1717_cb/

1261356.jpg

i7 5960X with DDR4 2133 1717*3/4.4 = 1170


3) Hippogriff's HWBOT http://hwbot.org/submission/2688953_hippogriff_cinebench___r15_core_i7_5960x_1746_cb/

1297784.jpg

5960X @4.5Ghz with DDR4 2400 (16 clocks) 1744*3/4.5 = 1164

4) Atisoc0936 HWBOT http://hwbot.org/submission/3157242_atisoc0936_cinebench___r15_core_i7_5960x_1763_cb/

1604405.jpg

5960X @4.5Ghz with DDR4 2400 (15 clocks) 1763*3/4.5 = 1175

5) Clavingers HWBOT http://hwbot.org/submission/3072850_claviger_cinebench___r15_xeon_e5_1680_v2_1701_cb/

1552990.jpg

This was a bit different, this was an Ivy Bridge Xeon E5 1680V2 (3-3.9Ghz) 25MB cache running at 4.7Ghz

1701*3/4.7 = 1085

If we assume the baseline of Zen is 1188 that means that Zen is clock for clock on par or faster than Haswell
 
Why dont you test around at review sites, they all lies? Also dont try and use different clocks for your compare at a lower clock. Uncore speed etc may not have changed for example.
upload_2016-12-22_12-52-3.png

cinebench-multi.gif


Anandtech numbers are fine.

And from here we know it can boost up to 3.3Ghz with 8 cores.
http://www.intel.com/content/www/us/en/support/processors/000006652.html

So 3.4Ghz Zen is 1188 and 3.3Ghz Haswell is 1300+.

Not faster clock for clock is it. And CB15 is another relatively light to medium load.

If you go through all the leaked Zen benchmarks. AOTS, GB, CB etc. You find that it ends up nowhere near Haswell, but rather near Sandy Bridge.
 
Last edited:
No CPU Z is invalidated, there is no way of actually determining those clocks and scores

So you take the 1324 as an example.

take one of my base clock IPC number say the 1175 one and you 1175 x 3.4 / 3 = 1331 ie: 1324 is not baseclock it is at least 3.4Ghz score

Magic right.

I used HWBot because there is validation and you can't thumb suck validation.
 
No CPU Z is invalidated, there is no way of actually determining those clocks and scores

So you take the 1324 as an example.

take one of my base clock IPC number say the 1175 one and you 1175 x 3.4 / 3 = 1331 ie: 1324 is not baseclock it is at least 3.4Ghz score

Magic right.

I used HWBot because there is validation and you can't thumb suck validation.

Where do you get your calculations from?

Are you trying to say hardware sites like techreport, anandtech and hardwarecanucks got no idea what they are doing? I dont think so.
 
Where do you get your calculations from?

Are you trying to say hardware sites like techreport, anandtech and hardwarecanucks got no idea what they are doing? I dont think so.

The numbers are from HWBOt which is an official benchmark and overclock platform, in order to post your results for the league you need to validate them, to validate them you need a screenshot with CPU-Z and other system information. I took their official results and did the math by converting the authenticated score timed by the base clock of 3 ghz and divided by the overclocked speed to get the base line score for a 5960X and it ranges from 1150-1180 depending on the RAM speed
 
The numbers are from HWBOt which is an official benchmark and overclock platform, in order to post your results for the league you need to validate them, to validate them you need a screenshot with CPU-Z and other system information. I took their official results and did the math by converting the authenticated score timed by the base clock of 3 ghz and divided by the overclocked speed to get the base line score for a 5960X and it ranges from 1150-1180 depending on the RAM speed

Even going by your own random hwbot submissions. A 3.3Ghz 5960X scores 1270. Way above Zens 1188 at 3.4Ghz.

Then you can keep making your extrapolated homemade numbers, but its not making them any more right.
 
Even going by your own random hwbot submissions. A 3.3Ghz 5960X scores 1270. Way above Zens 1188 at 3.4Ghz.

Then you can keep making your extrapolated homemade numbers, but its not making them any more right.

1) there is no authentication that it was a Zen and Zen has no working turbos, I know a person by the name of ShockG a writer/tester/overclocker and the information that he had about Zen which went viral stated that turbos don't work on Zen, it doesn't have a standard turbo but is designed to use XFR based turbo, which means ultimately whatever cooling you have, the turbo will be much higher. The 1188 was more likely a 2.8/3.2 Ghz score at max turbo given that there was hardly enough to clarify which ES chip it was if it was not even photoshoped.

2) 1270 at 3.3Ghz reverted to baseclock is 1154 base clock IPC 1270*3.4/3.3 = 1308
 
1) there is no authentication that it was a Zen and Zen has no working turbos, I know a person by the name of ShockG a writer/tester/overclocker and the information that he had about Zen which went viral stated that turbos don't work on Zen, it doesn't have a standard turbo but is designed to use XFR based turbo, which means ultimately whatever cooling you have, the turbo will be much higher. The 1188 was more likely a 2.8/3.2 Ghz score at max turbo given that there was hardly enough to clarify which ES chip it was if it was not even photoshoped.

2) 1270 at 3.3Ghz reverted to baseclock is 1154 base clock IPC 1270*3.4/3.3 = 1308

So you just make guesses and extrapolate as you see fit.

Here is another fun one for you:
http://hwbot.org/submission/3007772_yt.pinguin.pl_cinebench___r15_core_i7_5960x_1494_cb

3.3Ghz 5960X verified with 1494 in CB15.

Maybe you should stick with review sites and stop with your homemade numbers. ;)
 
You make no sense. As with all your calculations.

So now the higher score was gotten with an even lower clock? Brilliant!

If you lower the baseline it will trick CB into thinking your much larger boost is performance over base clock which is lower.

ie: 1188 x 3.4 / 2.2 = 1836 and 1188* 3.4 / 3 = 1346

I am trying to get Cinebench to read my i5 4460 underclocked but it is always showing 3.2 which is odd.
 
If you lower the baseline it will trick CB into thinking your much larger boost is performance over base clock which is lower.

ie: 1188 x 3.4 / 2.2 = 1836 and 1188* 3.4 / 3 = 1346

I am trying to get Cinebench to read my i5 4460 underclocked but it is always showing 3.2 which is odd.

Just stop the excuses. Really.

2.5Ghz or 3.3Ghz doesn't matter. It scored 1494. If anything it would have to be ABOVE those clocks.

In short, your homemade math doesn't check out. Your reliance on hwbot over review sites didn't turn out either. What's next, Naples ran at 1Ghz too? :D
 
Just stop the excuses. Really.

2.5Ghz or 3.3Ghz doesn't matter. It scored 1494. If anything it would have to be ABOVE those clocks.

In short, your homemade math doesn't check out. Your reliance on hwbot over review sites didn't turn out either. What's next, Naples ran at 1Ghz too? :D

Maybe you should try explain how an underclocked 5960X can produce as score higher than 500 others and why he has no points awarded despite his submission. HWBot likely disqualified the result for some altering he did.
 
1507569.jpg


These guys like changing baseclocks anyway it is still easy to convert:

1802 x 3.7 / 4.6 = 1446

1446 * 3 / 3.7 = 1175 @ 3ghz

1364926.jpg


1808 * 3 / 4.5 = 1205 which is more or less what the baseclock IPC is
 
I thought we already agreed that the Zen numbers are fake.

Why are you speculating based on fake results to begin with?

You are just wasting your time.
 
I thought we already agreed that the Zen numbers are fake.

Why are you speculating based on fake results to begin with?

You are just wasting your time.

we are not, we are finding the real ipc of a 5960X which seems 1150-1200ish
 
Congratulations you have discovered the true performance of the i7 5960X.

Now what?

If the leaked Zen is 1188 @ 3.1Ghz then it is about on par with Haswell clock for clock which to me at least is great and for the market it is a good outcome if that is true.
 
I think I figured out the YT Pinguin lower clock score:

1496 is essentially a IPC score with a 800MHZ overclock, the baseline IPC at 2.5Ghz is 1133 if you take a 1175 base score at 3Ghz and add a 800 overclock to work out that score you get 1175 x 3.8 / 3 = 1488

In short the score which is higher is because there is a much higher frequency difference between 3.3 and 2.5 (800) vs 3.3 and 3 (300)

Zen ES 1188 * 3.9/3.1 = 1495 with a 800Mhz over base clock.

Cinebench just calculates the IPC difference between max clock and base clock, you can then reverse the numbers depending on whether you calculating Baseclock IPC or IPC at overclock.

All the numbers corrolate mathematically, there is a window of 1133-1205 ranging from 2.5Ghz to 3.1Ghz it is not that difficult to figure out.
 
we are not, we are finding the real ipc of a 5960X which seems 1150-1200ish

1150-1200 isn't IPC either. And we know from reviews what it scores in CB15 MT. And that's 1320-1340ish no matter how much you try to reject this.

IPC for a CPU depends on the workload as well and how it handles that exact workload.
 
Last edited:
And my 5960X @ 4.6 consistantly gets a better score than the guy above - DRAM speed/timings, cache/uncore OC also play a part in the Cinebench score. Trying to extrapolate IPC from a couple of Cinebench submissions with vastly different configurations really isn't going to unlock the secrets to Ryzen.
 
1150-1200 isn't IPC either. And we know from reviews what it scores in CB15 MT. And that's 1320-1340ish no matter how much you try to reject this.

IPC for a CPU depends on the workload as well and how it handles that exact workload.

Of course it is going to give a score of 1350ish because 1175*3.5/3 = 1350ish, the score is being computed to max boost frequency, so at 3.5 ie: 500mhz OC over base it is producing a score of 1350 but at 3Ghz, it is only 1175.

And my 5960X @ 4.6 consistantly gets a better score than the guy above - DRAM speed/timings, cache/uncore OC also play a part in the Cinebench score. Trying to extrapolate IPC from a couple of Cinebench submissions with vastly different configurations really isn't going to unlock the secrets to Ryzen.

Yes, configuration matters a lot, timings etc does play a role. Can you screenshot a CB15 with CPU Z?
 
Of course it is going to give a score of 1350ish because 1175*3.5/3 = 1350ish, the score is being computed to max boost frequency, so at 3.5 ie: 500mhz OC over base it is producing a score of 1350 but at 3Ghz, it is only 1175.

You just make up things as you go dont you? Now its running 3.5Ghz? Perhaps you should stop trying to extrapolate and to try working backwards in scores in an attempt to get your predetermined results. Because your posts keeps contradicting themselves, not to mention your "math".
 
Of course it is going to give a score of 1350ish because 1175*3.5/3 = 1350ish, the score is being computed to max boost frequency, so at 3.5 ie: 500mhz OC over base it is producing a score of 1350 but at 3Ghz, it is only 1175.



Yes, configuration matters a lot, timings etc does play a role. Can you screenshot a CB15 with CPU Z?

Yeah, but not until I get home. I might be able to do it at lunch. And I'll leave the processor loaded so CPU-Z shows the actual clocks.
 
Last edited:
Lol, Cinebench scores argument. You guys must be bored or hoping AMD rebadged the 5960x or some shit. Its an entirely new architecture and its not going to be a power house at everything. It won't even have AVX 2 from what was said in the past. If it clocks well and competes in what matters and has Ivybridge IPC, it will be a good seller.
 
  • Like
Reactions: Zuul
like this
Back
Top