Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[inf64]

Fritz nerds already consider it one, just lurk the Chinese forums.

I'm curious why do you keep mentioning Fritz. 1800X scored how much according to chiphell? 22700pts?
How much 6900K gets @ stock (3.5Ghz boost)? I found a few results and the are just a few percents higher than 1800X's result, around 22800-22900.

edit:
http://www.mykancolle.com/?post=1380
1700 @ 4Ghz and 12T setting in Fritzbench @ 20804.

 

[Atari2600]

Great perf/W, but i hate the fact that 1700 actually consumes more power than 7700k.

Say what now?!?

 

[OrangeKhrush]

Zen die and IHS

Regarding this persian "test" Zen has over 100ns RAM with the non new agesa AMD rumor-wise fixed at those last days accoring to some reps at ASUS/others. I think it can have a huge impact in those gametests.

This is true, a source inside ASUS told me that there was a roll out of late firmware that fixed the bios and memory detection. ASUS have guaranteed 3800Mhz on the Crosshair.

 

[tamz_msc]

This is true, a source inside ASUS told me that there was a roll out of late firmware that fixed the bios and memory detection. ASUS have guaranteed 3800Mhz on the Crosshair.

Just hope that it is in time for the reviews.

 

[french toast]

This is true, a source inside ASUS told me that there was a roll out of late firmware that fixed the bios and memory detection. ASUS have guaranteed 3800Mhz on the Crosshair.

what does this mean for latency?

 

[OrangeKhrush]

Say what now?!?

This is like comparing the farmtruck and all the lamborghini's it stomps over a 1/4 mile, or why a supercharged pro-mod is heavier on gas than moms run to the shop weekend car.

 

[PotatoWithEarsOnSide]

1700 @3.2GHz was 1.067v
1700 @4.0Ghz was 1.342v

None of that 1.875v from earlier CPU-Z screenshots.

 

[OrangeKhrush]

what does this mean for latency?

I don't know yet, he is unlikely to respond in a while as he is in the Republic of China, and I assume it is early morning.

 

[tamz_msc]

Memory QVL for Crosshair VI is out - almost every GSkill Trident Z 3200MHz CL14 is supported.

 

[unseenmorbidity]

I'm curious why do you keep mentioning Fritz. 1800X scored how much according to chiphell? 22700pts?
How much 6900K gets @ stock (3.5Ghz boost)? I found a few results and the are just a few percents higher than 1800X's result, around 22800-22900.

edit:
http://www.mykancolle.com/?post=1380
1700 @ 4Ghz and 12T setting in Fritzbench @ 20804.

Because the only relevant benchmarks are the ones AMD loses. You should ignore all the others!

 

[lolfail9001]

Say what now?!?

They give you overall result and result per watt. Do the division and look at wattages you produce. In POV difference is like 20 watts more on 1700, in handbrake it's 10.

None of that 1.875v from earlier CPU-Z screenshots.

That 1.875 is on LN2 and pushing well over 5Ghz, open your eyes.

This is like comparing the farmtruck and all the lamborghini's it stomps over a 1/4 mile, or why a supercharged pro-mod is heavier on gas than moms run to the shop weekend car.

Nop, i am comparing truck and charged hatchback and wonder why manufacturers claim identical fuel consumption, but truck consumes more. Turns out truck was made by VW. I mean, AMD.

Because the only relevant benchmarks are the ones AMD loses. Ignore all the others!

Why, i am sure if you use Handbrake and Cinebench all the time, then Ryzen is amazing CPU. I don't.

 

[PrivateCeralion]

Delidding Ryzen: Is AMD Ryzen 7 soldered or not?
https://www.youtube.com/watch?v=lOZbK3tP7EU

 

[unseenmorbidity]

 

[OrangeKhrush]

They give you overall result and result per watt. Do the division and look at wattages you produce. In POV difference is like 20 watts more on 1700, in handbrake it's 10.

That 1.875 is on LN2 and pushing well over 5Ghz, open your eyes.

Nop, i am comparing truck and charged hatchback and wonder why manufacturers claim identical fuel consumption, but truck consumes more. Turns out truck was made by VW. I mean, AMD.

Why, i am sure if you use Handbrake and Cinebench all the time, then Ryzen is amazing CPU. I don't.

i7 7700K is binned at what? 77TDP but its like 140W if not more, when did TDP become power consumption? last I checked it was just thermal displacement.

 

[Glo.]

Anyone care to calculate the die size?

 

[Atari2600]

They give you overall result and result per watt. Do the division and look at wattages you produce. In POV difference is like 20 watts more on 1700, in handbrake it's 10.

I blame the sckools.


The fact that the AMD is both (POVray & handbrake):
1. Performing more calculations per second
2. Consuming less watts per second

Means the AMD will finish first and consume less Joules to that point while doing so. The Intel will go on to consume even more energy till it also completes the task!

 

[Bouowmx]

watts per second

Pedant: You mean just watts.

 

[nathanddrews]

So should I just go for a base 1700 and cheapest board if I want to use it for a VM/encode/transcode server?

 

[lolfail9001]

2. Consuming less watts per second

It is not watts per second, it is seconds encoded per watt of average power consumption during those seconds . Entirely useless metric but what can you do.

Pedant: You mean just watts.

Nope, it is literally written " Watts/sec of encode time" on the slide. But they actually meant the reverse of that. 10/10 marketing, AMD.

i7 7700K is binned at what? 77TDP but its like 140W if not more, when did TDP become power consumption? last I checked it was just thermal displacement.

7700k is 91 or 95W TDP and consumes less than that in most workloads. And in case of CPUs, TDP IS power consumption in some manufactured scenario.

 

[inf64]

Hmm a new "cheat" BIOS popped up on "others" OS selection (FR) page for Crosshair VI.
https://www.asus.com/fr/Motherboards/ROG-CROSSHAIR-VI-HERO/HelpDesk_Download/

CPC HW calls it a "cheat" BIOS for some reason. It is supplied at the last minute by Asus and AMD. Likely rYzes the scores hence CPC calls it "cheat". They are so salty it is crazy

edit: Performance bias option

 

[.vodka]

Anyone care to calculate the die size?

Let's use photoshop's vanishing point

We know the AM4 package measures 40*40mm. One side is correct, one as calculated by VP has a little error of 0.38mm. Let's correct that... I'm missing my coffee so this could be horribly wrong, but still

40.38mm -> 40mm
21.96mm -> x = 21.75mm

Corrected: 21.75*9.51 = 206.84mm^2
Non corrected: 21.96*9.51 = 208.83mm^2

Rounding... 9.5mm * 22mm ?

There's still some margin of error here to consider (perspective although VP tries to correct for, some solder over the die corners), but that's pretty close to the rumored 190-193 mm^2. In fact there's an article over at semiaccurate that talks about Ryzen's die size, but it requires a subscription to be read.

There was some talk that AMD's Ryzen die shot is actually shrunk a bit, the aspect ratio threw someone off in some forum vs the 44mm^2 number for the CCX... I don't remember where.

 

[IEC]

So... don't delid.

der8auer is a world-class overclocker and hardware modder and he just killed 2/3 of his chips trying to delid.

Good news:
All chips soldered, no need to delid unless you are extreme LN2 overclocker.

 

[raghu78]

https://twitter.com/BitsAndChipsEng/status/837005253211992064

Take it for what its worth. bits and chips tweet on the IMC. Something tells me the real performance results are only going to be out tomorrow. I think some of the poor results we are seeing might be a combination of older BIOS and/or ES CPU / MB. Hopefully there are more pleasant surprises from Ryzen tomorrow.

 

[lolfail9001]

Rounding... 9.5mm * 22mm ?

Sounds about right, outside of one issue: this is actually getting dangerously close to the size of BDW-EP LCC. That has more of like... everything. With definitely smaller core and even smaller CCX, that makes no sense.

Take it for what its worth. bits and chips tweet on the IMC. Something tells me the real performance results are only going to be out tomorrow. I think some of the poor results we are seeing might be a combination of older BIOS and/or ES CPU / MB. Hopefully there are more pleasant surprises from Ryzen tomorrow.

This gif fits both memory bandwidth report and memory latency report by all the software around.

 

[Riek]

They give you overall result and result per watt. Do the division and look at wattages you produce. In POV difference is like 20 watts more on 1700, in handbrake it's 10.
.

I understand the sanity, but to be honest things don't add up according to me... but maybe it is to late to think about math for me...

I took the easiest one to think about --> handbrake

Ryzen needs 4minutes33 seconds in handbrake
I7 needs 6minutes27

That would make Ryzen ~30% faster. (2minutes on 6minutes), yet it has a 50% advange against the i7 in perf/W (1.68 vs 3.23 => 1.68*2 = 3.36)

Simply said that would mean that if both machines would take 100 seconds, i7 would consume almost twice as much right?

So how can somthing be 30% faster, having twice the perf/W and yet consume more?

My simple believes would be that if they both consumed 100W, AMD would have had a faster time. (aka 50% or 3m13). Wouldn't that bring AMD to ~70W instead?


edit: it is to late for me , i looked at the wrong numbers :') (compared against 6800K)


Against i7 7700K it is the following for handbrake:
27% faster in time.
36% more perf/W

Not as dramatic as the other, but still I would expect a lower power consumption and not a higher.