Leaked AMD Ryzen Benchmarks?

lolfail9001 said:
Was in the video with LN2 run that went private. Or perhaps in linus one too. Did state outright, that r7 1700 (no-x) ran at 3Ghz in Handbrake demo against 7700k. Besides the leaks only work if my statement about boost is true, otherwise they do not work at all. Like Shintai's post illustrates ^_^

It is, 5.14Ghz Zen beats 6Ghz Haswell-E in Cinebench.
Click to expand...
Just went back through Linus vid and the Handbrake also included Blender at same time, also was the 1700X.
The 1700 was shown with games at the end of his vid, where in BF1 the test compared the 1700 against the 6800K.

LN2, well that needs extensive monitoring to compare different CPUs and their core behaviour/performance.
Cheers
 
Shintai said:
A 6900K doesn't run all cores as stock at 3.7Ghz. And if it was optimal it would score around 1540 in MT. It does however run 3.7Ghz at ST boost. The boost 3.0 is driver based and needs manual setting.
Click to expand...
You accept recent Xeon CPUs can run all cores up to 10C at near full boost?
What are the HEDT based upon?
Cheers
 
CSI_PC said:
Just went back through Linus vid and the Handbrake also included Blender at same time, also was the 1700X.
Click to expand...
Oh, so it was not Linus one. Mirror of the vid i am talking about is above your post, though.
CSI_PC said:
The 1700 was shown with games at the end of his vid, where in BF1 the test compared the 1700 against the 6800K.
Click to expand...
He actually compared 3.6Ghz base clock QS against 6800k in that vid, but he was too dumb to notice the QS name.
CSI_PC said:
LN2, well that needs extensive monitoring to compare different CPUs and their core behaviour/performance.
Click to expand...
There was like 10 CB results, trust me, these guys can find the frequency wall on LN2 in 10 attempts. Mirror of vid i am talking about above.
 
CSI_PC said:
You accept recent Xeon CPUs can run all cores up to 10C at near full boost?
What are the HEDT based upon?
Cheers
Click to expand...

A 5960X for example runs all cores at 3.3Ghz.
http://www.intel.com/content/www/us/en/support/processors/000006652.html

So what Xeon did you talk about with full boost up to 10 cores? The single one that's the exception? But even that one had to drop 200mhz after 2 cores.

1-1080.1792621465.png
 
Shintai said:
10 cores out of what, 22? If you think the 6900K runs all cores as stock at 3.7Ghz prove it.

A 5960X for example runs all cores at 3.3Ghz.
Click to expand...

http://www.tomshardware.com/reviews/intel-core-i7-broadwell-e-6950x-6900k-6850k-6800k,4587-9.html
Again, let’s take a look at the individual core frequencies for the three configurations we tested. Despite a 3.2GHz base clock rate, Turbo Boost pushes the CPU to 3.7GHz in lightly threaded workloads. Our sample managed to maintain this frequency across all cores during the stress test. One core even reportedly hit 4GHz.
Click to expand...

You were saying?
 
Shintai said:
3.7Ghz on all cores requires a mobo to run the non stock multicore enhancement.
Click to expand...
That was not active:
Intel’s Core i7-6850K is only marginally faster than the -6800K. Again, its base clock rate is 3.6GHz and its maximum Turbo Boost frequency tops out at 3.8GHz, though we only saw our sample get up to 3.7GHz.
Click to expand...
Evidently, neither was driver :)
 
lolfail9001 said:
Oh, so it was not Linus one. Mirror of the vid i am talking about is above your post, though.

He actually compared 3.6Ghz base clock QS against 6800k in that vid, but he was too dumb to notice the QS name.

There was like 10 CB results, trust me, these guys can find the frequency wall on LN2 in 10 attempts. Mirror of vid i am talking about above.
Click to expand...
Yeah agreed.
Unfortunately the vid compared Intel 4/8 to the 8/16 and then shows the LN2 world record.
My point is when a couple here are trying to compare to Intel CPUs and correlate also with LN2, LN2 makes this very tricky as the silicon behaviour-performance can also change with such low temps (in fact it does), and should be treated as a separate aspect.
Thanks
 
Last edited:
lolfail9001 said:
That was not active:

Evidently, neither was driver :)
Click to expand...

Then it cant be true can it. If it hit 4Ghz without its either OCed or someone is drunk.

This is what is required.


XFR on the other hand works automatically without this added layer. Much more elegant in that regard.
 
Shintai said:
Then it cant be true can it. If it hit 4Ghz without its either OCed or someone is drunk.
Click to expand...
I don't own 6900k to verify it, ya know.
CSI_PC said:
Unfortunately the vid compared Intel 4/8 to the 8/16 and then shows the LN2 world record.
Click to expand...
Eh, the point of it was to establish clocks the 8 core was running in Handbrake, comparison itself is lame, i agree.
CSI_PC said:
My point is when a couple here are trying to compare to Intel CPUs and correlate also with LN2, LN2 makes this very tricky as the silicon behaviour-performance can also change with such low temps (in fact it does).
Click to expand...
It may change silicon performance, but i highly doubt it can change architecture throughput out of thin air. Besides, the irony of 5.14Ghz or something Ryzen beating 6Ghz Haswell-E is delicious.
 
Shintai said:
A 5960X for example runs all cores at 3.3Ghz.
http://www.intel.com/content/www/us/en/support/processors/000006652.html

So what Xeon did you talk about with full boost up to 10 cores? The single one that's the exception? But even that one had to drop 200mhz after 2 cores.

1-1080.1792621465.png
Click to expand...

Part of what I am thinking comes back to the motherboards and microcode in some examples for the past.
The exception you noted in the E7 and also in E5 (where drops by I think only 100MHz) shows can run at near max even without motherboard support.

A fair few of motherboards do support running the 6900K/etc enhanced (couple reviews noted all cores at or near to max boost frequency)
Also possibly thinking of Haswell Xeon parts can up to 10 cores full boost before dropping depending upon the microcode, there is a glitch that enables up to 15C to run at full boost if the motherboard and cooler can handle it but it has issues being a glitch.
Also the 5930 dropped 100mhz for all cores probably was in the back of my mind.

Anyway as Lolfial9001 showed it is possible to run all cores for the Broadwell-E at near boost max without OC, Tom's from what I remember were not the only one.
Cheers
 
Last edited:
So 1 thing I hope [H] explains is...I keep reading that they can clock as high as cooling can go.....So does this mean even though they are unlocked, there will be a point where the software stops overclocking? I mean this is [H] where they test to that limit and even make it crash....curious if the Software would prevent that.

Is it confirmed that there is no Bios overclocking with Ryzen? All Software?
 
Ryzen is looking good...if I were in the market for a new PC I'd probably go with a 1700X. Aside from gaming I do a fair amount of photo and video editing on my computer so the 8 cores would be awesome.
 
According to Mr. Gibbo (seemingly the mouthpiece) of OCUK, the 1700, 1700X and 1800X all have XFR, the X means only that they are binned for higher boost clocks.

Also, people expecting these chips to overclock a lot more should naturally temper their expectations, wait for reviews.

Now, the third link apparently refers to this post where the poster states that the X means something a little more.

Now, I used Google Translate, so could be wrong, but the poster is saying the X is something still under NDA.
 
  • Like
Reactions: muxr
like this
Why compare it to a processor that came out in 2014 and is 3 generations ago?
Apples to oranges sounds like to me.
 
^^ The Excel test video, the guy pooched it on the C/T count on the 1700, "8 core 8 thread" he said. :eek:

right around the 0:28 'ish mark.
 
SighTurtle said:
According to Mr. Gibbo (seemingly the mouthpiece) of OCUK, the 1700, 1700X and 1800X all have XFR, the X means only that they are binned for higher boost clocks.

Also, people expecting these chips to overclock a lot more should naturally temper their expectations, wait for reviews.

Now, the third link apparently refers to this post where the poster states that the X means something a little more.

Now, I used Google Translate, so could be wrong, but the poster is saying the X is something still under NDA.
Click to expand...

I think most are now expecting at best 4.4GHz and probably the 4.2GHz with the 1800X and great air cooling, which seems pretty reasonable to me if it can.
It will be interesting to see if the 1700X matches the 1800X with OC.
Cheers
 
Last edited:
CSI_PC said:
I am looking at the data just released.
1800X mult-threaded score 1,601
6900K multi-threaded score 1,474

1800X single threaded score 162
6900K single threaded score 162.
Click to expand...

Those AMD numbers don't agree much with reviews

81824.png

81823.png
 
juanrga said:
Those AMD numbers don't agree much with reviews

81824.png

81823.png
Click to expand...

I doubt the AMD machines used in the demo are identical to the ones in the Anandtech database, memory is one aspect that comes to mind.
But they were identical in the AMD event for both AMD Ryzen and Intel 6900K so no unfair advantage, and in fact the Single threaded score was higher for the 6900K at the AMD event.
The difference is not enough to really debate as an issue unless you know for sure the system-environment for both the Anandtech score and AMD event.
 
Last edited:
Damn AMD back at it again with them low low prices.

Honestly if this thing turns out to be half as good as it seems, Intels got to restructure the future lineup (i dont see them doing a full drop pricing until the next release).

BTW anyone looking a the Linus video and knowing how XFR/tb 3.0 works tell me the weird clock frequency noted on the BF1 demo specifically as he pulls up both task managers.

Framerate achieved was

Intel 88 AMD 93

Task manager from Intel was 3.06 clock vs AMDs TM showing 3.64 clock

mind you the game is minimized, but do we assume that since he minimized both game windows Intel turbo boost down clocked for efficiency or lack of thermal headroom while XFR remained at the given frequency due to cooling headroom/granularity (linus hints at)? Or do we assume that the Intel chip was clocked at 3.06 during in game and minimized game window and the AMD remained pegged at 3.64?
 

Q2 for Ryzen 5, Ryzen 3 in 2H 2017.

http://www.eetimes.com/document.asp?doc_id=1331380&page_number=2

The Ryzen 7 represents the first of several volleys expected this year and beyond. AMD has said it plans to roll out its first Zen-based chips for servers before June and notebook versions before the end of the year.
Click to expand...

I wonder if Project Scorpio affects timetables for the other chips beyond using GF/Samsung's manufacturing capacity.
 
CSI_PC said:
I doubt the AMD machines used in the demo are identical to the ones in the Anandtech database, memory is one aspect that comes to mind.
But they were identical in the AMD event for both AMD Ryzen and Intel 6900K so no unfair advantage, and in fact the Single threaded score was higher for the 6900K at the AMD event.
The difference is not enough to really debate as an issue unless you know for sure the system-environment for both the Anandtech score and AMD event.
Click to expand...

We have to wait to reviews for proper comparisons.
 
juanrga said:
We have to wait to reviews for proper comparisons.
Click to expand...
And yet ironically it was you who provided the Anandtech Cinebench result to say AMD's results are wrong.....
Even reviews will have some variance between sites as they are not all using identical machines, to repeat the variance in this case is NOT enough to infer as you did that AMD is doing something wrong unless you knew the exact environments of both AMD and Anandtech.
Cheers
 
Last edited:
Dumb to be R3, R5, and R7, vs Intel's i3, i5, i7. Wouldn't have made more sense to be R4, R6, and R8? It matches core count AND it's ONE BIGGER!!!!!11!!!
 
My brother's sister's gardener's cousin has a friend who knows a dude who's sister is dating the guy who landscapes the AMD campus. HE smokes pot with this janitor who's friends with one of the girls who dates the boyfriend of an engineer at nVidia who knows a guy who cleaned the gutters of an AMD engineer.

HE said: Zendozer.
 
AMD numbers on Intel performance simply do not agree with what i have been testing. I noticed that AMD crippled Intel platform by switching from quad memory channel to dual memory channel, also running Intel CPU at stock which make Ryzen look even worse knowing that intel parts run low clock speed and still achieve pretty much same numbers.

Here are some benches i did on my Xeon 10/[email protected]@3.5 Ghz in 3DMark....

http://www.3dmark.com/fs/11790011

http://www.3dmark.com/spy/1259762

It is extra 2 cores but also 800Mhz disadvantage over 1800x Ryzen and it will still come out better.
 
Last edited:
Shintai said:
Then it cant be true can it. If it hit 4Ghz without its either OCed or someone is drunk.

This is what is required.


XFR on the other hand works automatically without this added layer. Much more elegant in that regard.
Click to expand...
It is in the BIOS without the driver-application and will detect strongest core, as long as it is on a motherboard (and microcode) supporting this or has its firmware upgraded if older.
That app-driver enables you to have greater control and in it specifically force applications to a single core (makes it easier for end users) and importantly also allows to set the Max Boost 3.0 above the rated frequency spec given by Intel - other utilities hook into this and can set higher such as Asus AI Suite.
Without the app it is just more general and cannot exceed the Max Boost 3.0 Intel spec of the CPU.
Anyway like I said it comes down to the motherboard/microcode and most will have this function if supporting Broadwell fully, same way it is the motherboard that controls all cores up to but not beyond the max boost frequency spec for the CPU unless one actually OCs - context though are CPUs greater than 4-cores and so Intels HEDT/extreme and their expensive motherboards that push the features by default.
But then would one run a PC without going into the BIOS to check XMP profile and right settings for their decent memory.

Here is a classic example where they compare performance of the 6950X against the driver-application set to 4.6GHz single core and forcing Cinebench through it.
Notice that the scores mean without the Max Boost 3.0 driver-app it must be boosting near to the 4.0GHz by default as the scores would be further apart as the max turbo frequency for the 6950X is 3.5GHz.
They also comment the BIOS detected strongest core.
https://www.bit-tech.net/hardware/2016/05/31/intel-core-i7-6950x-broadwell-e-review/2
Their score pretty much aligns with other Cinebench results from other sites, suggesting Boost 3.0 is hitting 4GHz or close to it by default even without driver-utility; 158 default, 185 set to 4.6GHz with driver-utility.
Before anyone posts, the Cinebench score where it mentions Turbo Boost 3.0 is with the driver-utility and 4.6GHz, the other is without and defaults but it is still hitting near 4GHz.

Anyway the driver-app is more to make an end-users life easier/greater flexibility, and that bit-tech review score of 158 fits reasonably well trend scores of AMD event/other sites while also being around the score point for 4GHz relative to the 4.6GHz test result of 185.
Cheers
 
Last edited:
Shintai said:
From AMDs own CB15 bench, it seems to suggest that:
4.1Ghz Ryzen matches Broadwell-E at 3.7Ghz in ST
3.9Ghz Ryzen matches Broadwell-E at 3.5Ghz in MT

That's assuming 100Mhz XFR.

We will see how it turns out when we see reviews.
Click to expand...

Intel is faster per clock, and in all these benchmarks AMD forgot that Intel runs easily at 4.4Ghz
 
Luka said:
Intel is faster per clock, and in all these benchmarks AMD forgot that Intel runs easily at 4.4Ghz
Click to expand...

Doesn't matter when everything is moving toward multithreading. Even a Ryzen 1400X 4C/8T probably gives Intel quads a run for their money in MT, while being adequate in ST.
 
  • Like
Reactions: muxr
like this
Luka said:
AMD numbers on Intel performance simply do not agree with what i have been testing. I noticed that AMD crippled Intel platform by switching from quad memory channel to dual memory channel, also running Intel CPU at stock which make Ryzen look even worse knowing that intel parts run low clock speed and still achieve pretty much same numbers.

Here are some benches i did on my Xeon 10/[email protected]@3.5 Ghz in 3DMark....

http://www.3dmark.com/fs/11790011

http://www.3dmark.com/spy/1259762

It is extra 2 cores but also 800Mhz disadvantage over 1800x Ryzen and it will still come out better.
Click to expand...

Are any Time Spy CPU scores for Ryzen posted? I only saw FS
 
You must log in or register to reply here.
Back
Top