Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[looncraz]

That ^^ CB r15 ST score for 6900K @ 'stock" of 185(!!!)pts has to be the highest "stock" score of all time in history of BDW-E reviews. Kudos for finding such a rare gem!
Anandtech lists the score of the same SKU as 154 ( @ 3.7Ghz Turbo 2.0) in their database, but hey that is puny Anandtech.

BTW Linus reran the benchmark after the presentation and 1800X @ stock scored 1627pts FYI. So AMD actually sandbagged a bit their own score.

That's actually the OpenGL score, not ST score.

 

[french toast]

Nice excuse...Amd sandbagged the i7 6900k score on purpose to make their own cpu's look better...im sure they boosted theirs aswell..

Now Ryzen doesnt look that good is it? yeah they have lower price but it isnt faster then what we already have...

Where was that graph from? What cooling/ram did it have?
AMD did not sandbag i7 6900k at all, it seems AMD used intels ref cooler? and used turbo boost 3 for single thread when most reviews of 6900k didn't use turbo boost 3 @4ghz including Anandtech (used a water cooler which boosted MT scores).
The only thing different AMD did was use dual channel memory instead of quad-which according to what ive read doesn't make a difference in cinebench, It actually made intel look better in the perf/watt measurement if anything.
Edit, Yea thats open GL, missed that.

 

[inf64]

That's actually the OpenGL score, not ST score.

Good catch looncraz, I missed that.

The score AMD quoted for SPEC INT 2006 isn't that impressive. It looks below Haswell.

From RWT http://www.realworldtech.com/forum/?threadid=165321&curpostid=165356:

Haswell score with gcc -O2 compilation: 35.6 @3.8GHz source

3.4Ghz Haswell would get 31.85, basically the same as Zen at 3.4Ghz (1.1% difference)

Also this post shows SKL score:

Here are some official and subset int_2006 scores for comparison:

AMD A10-7850K i7-6700K E3-1280 v5
base: 30.34 70.75 71.21
peak: 31.06 72.84 73.83


If we remove 462.libquantum when calculating the scores:

AMD A10-7850K i7-6700K E3-1280 v5
base: 18.89 37.39 37.69
peak: 19.34 38.49 38.83


37.4 at 4.0/4.2GHz for i7-6700K + icc + base optimization flags

So SKL @ 4Ghz scores 37.4 meaning @ 3.4Ghz it should score : 37.4*3.4/4=31.79 or basically on par with Zen and Haswell, not much of a difference. I used 4Ghz as best case scenario for SKL, not its ST turbo clock.

 

[looncraz]

The score AMD quoted for SPEC INT 2006 isn't that impressive. It looks below Haswell.

Do you have a score for the 6900K?? Because I can't find anything.

Nevermind

 

[Nothingness]

From RWT http://www.realworldtech.com/forum/?threadid=165321&curpostid=165356:

3.4Ghz Haswell would get 31.85, basically the same as Zen at 3.4Ghz (1.1% difference)

Also this post shows SKL score:

So SKL @ 4Ghz scores 37.4 meaning @ 3.4Ghz it should score : 37.4*3.4/4=31.79 or basically on par with Zen and Haswell, not much of a difference. I used 4Ghz as best case scenario for SKL, not its ST turbo clock.

So Ryzen at about 9.4/GHz. The data I have show 4790 @ 10.4/GHz in 64-bit and 10.2/GHz in 32-bit. Both with gcc 5.2 and -O3 (IIRC). Not exactly the same compiler revision and not the same flag, but I don't expect more than 10% difference.

I hope someone will make a fair Intel vs AMD comparison with the same compiler and the same set of flags!

 

[inf64]

That's actually the OpenGL score, not ST score.

So Ryzen at about 9.4/GHz. The data I have show 4790 @ 10.4/GHz in 64-bit and 10.2/GHz in 32-bit. Both with gcc 5.2 and -O3 (IIRC). Not exactly the same compiler revision and not the same flag, but I don't expect more than 10% difference.

I hope someone will make a fair Intel vs AMD comparison with the same compiler and the same set of flags!

Yep AMD used version 4.6 with O2 flag. Someone will do it , I think Stilt for sure.

 

[blublub]

Just saying,

Cinebench R15

Ryzen R7 1800X - 1601 (Amd Own benchmarks) >at least a little bit of salt..

Intel I7 6900k - 1578

That's 23 points more for 50% less money - I would call that a win, for both performance and perf/dollar.

But yeah, go ahead and buy a 7900k for 1100$ and I meanwhile buy 2 CPUs for the same money and have a LOT more horsepower.

 

[Valantar]

That's 23 points more for 50% less money - I would call that a win, for both performance and perf/dollar.

But yeah, go ahead and buy a 7900k for 1100$ and I meanwhile buy 2 CPUs for the same money and have a LOT more horsepower.

Exactly. This makes high performance editing rigs/workstations actually accessible. I can build a complete Ryzen 1700X PC for roughly the price of a 6900K and a motherboard, with near identical performance and efficiency. That's a game changer.

 

[Nereus77]

The AnandTech review can't come soon enough. Hopefully the CPUs will have a nice OC headroom, especially the 4 and 6 core variants. They'd be absolute bargain gaming CPUs.

 

[Topweasel]

Exactly. This makes high performance editing rigs/workstations actually accessible. I can build a complete Ryzen 1700X PC for roughly the price of a 6900K and a motherboard, with near identical performance and efficiency. That's a game changer.

That's my opinion on this. I have been running a SB 3930K for several years now. I wanted to gut my machine and get a few more cores. Now I could build a new computer minus video for nearly the same price as a Mobo, CPU, Memory upgrade. It's a pretty bold step here from AMD. On the Enthusiast chips from Intel they are comparable in IPC and have a major cost advantage.

As they move down the stack they might start losing slightly in IPC but gain back in Cores. The 4c/8T Ryzens at nearly i3 prices becomes a tempting low cost solution.

This next week is going to be torture waiting for reviews and actual performance numbers.



Sent from my Pixel XL using Tapatalk

 

[bjt2]

~4.2, 4.3-4.4 generally requires water to not throttle.

You are still convinced XFR is worth a damn? Also, telling the right voltage is only relevant for the power savings/longevity, the last things you care about in LN2 conditions.

If you use a lower Vcore, you can up the clock a bit, if you are exaggerating in the first place... Moreover 5.2GHz on Ln2 was obtained with all core loaded. What about max clock sustained with only one core? Probabily more, but with fixed oc you are stuck to max clock with all core loaded... With XFR no. You can have 5.2 all core and maybe 5.3-5.4 on one core...

 

[lolfail9001]

If you use a lower Vcore, you can up the clock a bit

If you use a lower Vcore, you sit at lower Fmax, that's fairly straightforward when your CPU sits at under 100 Kelvins.

What about max clock sustained with only one core?

100-300Mhz higher, if you look at SuperPI for 6900k/5960X.

With XFR no. You can have 5.2 all core and maybe 5.3-5.4 on one core...

We have literally 0 evidence XFR works like that, ya know. And yes, it is theoretically possible, but you do understand that in this case the real issue with stability at these frequencies is SMT, not core count, don't you?

 

[OrangeKhrush]

Nice excuse...Amd sandbagged the i7 6900k score on purpose to make their own cpu's look better...im sure they boosted theirs aswell..

Now Ryzen doesnt look that good is it? yeah they have lower price but it isnt faster then what we already have...

That must be over 5Ghz on most, I have a 5960X at 4.7 and good luck getting 200+ single thread, or your source is meh

 

[OrangeKhrush]

Do you want to actual backup your continued accusations with any evidence ?


cuz you know
http://images.anandtech.com/graphs/graph10337/81823.png
https://www.vortez.net/articles_pages/intel_broadwell_e_core_i7_6900k_6950x_review,12.html
http://www.sweclockers.com/test/221...-i7-6850k-och-i7-6800k-familjen-broadwell-e/4
http://www.hardwarecanucks.com/foru...l-broadwell-e-i7-6950x-i7-6900k-review-5.html
http://www.legitreviews.com/intel-core-i7-6900k-processor-review_191040/7

All these lairs, when one source is completely different to all the rest and given that AMD give exact system specifications in the damn slide deck maybe you have gone for the erroneous one because it fits your rear guard action?



edit: ROFL, those red scores aren't ST CPU they are openGL............ LOL!!!!! look at 6700k @ 4.8ghz...... AWESOME OWN GOAL!!!!!

Nice spot, I was wondering why my 5960X gets nowhere near 200+ scores at 4.7Ghz.

 

[PhonakV30]

4Ghz Boost on all cores? if yes then awesome!!!

 

[lolfail9001]

Nice spot, I was wondering why my 5960X gets nowhere near 200+ scores at 4.7Ghz.

Well, looks like your source on OC was on-point. About everything except the 1T command rate, of course.

 

[OrangeKhrush]

4Ghz Boost on all cores? if yes then awesome!!!

Nope only single core

 

[OrangeKhrush]

Well, looks like your source on OC was on-point. About everything except the 1T command rate, of course.

That is relatively minor and it may have been generic to Asus validation boards and/or sample chips which he did state to me he did not have turbo effectively working until about a week ago when AMD released a final firmware update.

 

[guskline]

moonbogg, last night I was able to pre-order both a R7 1800x and an Asus ROG Crossfire VI from Newegg.

 

[OrangeKhrush]

Remember the old 1188 cinebench leak that was dismissed, well it was the very first ES 2.8ghz ACT

1537 * 2.8 / 3.7 = 1183 ~ 1188

 

[inf64]

Looks like 1600X is having exact same clocks as 1800X with 33% less cores. Not bad at all especially if they price it at 260-280$ range. ST performance @ 4Ghz or 4.1Ghz(XFR) combined with ACT at ~3.7Ghz should be really good for games. If it OCs to ~4.2Ghz (~16% above base) it will be best perf./dollar SKU IMO.

 

[OrangeKhrush]

Looks like 1600X is having exact same clocks as 1800X with 33% less cores. Not bad at all especially if they price it at 260-280$ range. ST performance @ 4Ghz or 4.1Ghz(XFR) combined with ACT at ~3.7Ghz should be really good for games. If it OCs to ~4.2Ghz (~16% above base) it will be best perf./dollar SKU IMO.

1600X and 1400X are sweetspots

 

[inf64]

1600X and 1400X are sweetspots

The whole lineup looks really good, we have to just wait for real reviews. For the price Ryzen is unbeatable. There will be workloads where both 7700K and 6900K could be faster or more efficient but overall , the average performance and resulting perf./$ value of the whole package is likely going to be very impressive.

 

[CatMerc]

1600X and 1400X are sweetspots

Definitely. The moment I saw 3.6/4.0 for 1600X, I realized that this is going to be my CPU.

 

[lolfail9001]

And here i am thinking that the real sweetspot is r7 1700. Though in hindsight it is like full 100 bucks for 2 more cores.