Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[unseenmorbidity]

AM4 Mainboard VRM Circuitry List

https://gfycat.com/DopeyLoathsomeBlueandgoldmackaw

 

[IEC]

AM4 Mainboard VRM Circuitry List

EDIT:

Strange. 13 seconds on stopwatch ~ 10 seconds in Windows

Ryzen or Windows 10 Timing Issue?

Lol, the cheapest MSI motheboard ($80) uses the same VRMs as their top-end ($300) - just fewer phases. They must have a boatload of those Niko Semi PKs

 

[unseenmorbidity]

Lol, the cheapest MSI motheboard ($80) uses the same VRMs as their top-end ($300) - just fewer phases. They must have a boatload of those Niko Semi PKs

lol, I didn't notice that. Good catch! That Titanium pricetag is looking more and more ridiculous.

 

[IEC]

lol, I didn't notice that. Good catch! That Titanium pricetag is looking more and more ridiculous.

I'm of the opinion that while they aren't as good efficiency or quality-wise as some of the very nice VRMs used by (e.g. Asus C6H/Asrock Taichi), they are more than adequate for 99%+ of users.

They BIOS lock my vcore on the $80 board to 1.4V and you hit diminishing returns on OC before that voltage anyways. You can easily do 1.3-1.35V for a stable 24/7 overclock and never notice the cheaper components.

 

[Rifter]

I'm of the opinion that while they aren't as good efficiency or quality-wise as some of the very nice VRMs used by (e.g. Asus C6H/Asrock Taichi), they are more than adequate for 99%+ of users.

They BIOS lock my vcore on the $80 board to 1.4V and you hit diminishing returns on OC before that voltage anyways. You can easily do 1.3-1.35V for a stable 24/7 overclock and never notice the cheaper components.

The efficiency isnt the issue its the fact that they run so hot they tend to burn out after a few years if you push them to hard.

 

[IEC]

The efficiency isnt the issue its the fact that they run so hot they tend to burn out after a few years if you push them to hard.

Well, technically the efficiency *is* the issue as less efficient components = more heat produced for same current. But as long as you're below about 95C on the VRMs I wouldn't worry much given that it takes sustained temperatures above that point to decrease component lifespan. For what it's worth, they use some of the same VRMs on Z170/Z270 boards that see overclocks with no real quantifiable difference for end users. (AFAIK)

 

[sushukka]

Maybe already an "old" review of Ryzen Linux benchmarks:
https://www.phoronix.com/scan.php?page=article&item=ryzen-1800x-linux&num=1
As globally more and more workload is running on Linux, this is again good news for Ryzen. Can't wait to see Naples in action.

 

[imported_jjj]

Have there been any updates to Deus Ex or maybe it's a matter of settings or GPU drivers? They test with a GTX1080.

 

[unseenmorbidity]

The efficiency isnt the issue its the fact that they run so hot they tend to burn out after a few years if you push them to hard.

That reminds me, here are some thermal imaging reviews I found.


Tweaktown Taichi Review: Including Thermal imagining and circuitry breakdown


Tweaktown Crosshair Review: Including Thermal imagining and circuitry breakdown


MSI Titanium

Gigabyte G5

Asus Crosshair

Read more at http://www.legitreviews.com/thermal...gigabyte-and-msi_193006/5#V7tVMWvxzWMoJ4Py.99

 

[IEC]

Lol, a 1700 at stock clocks on a test bench. That sets a floor maybe, but we are more interested in a ceiling.

Do a 1800X @ 4GHz 1.4V and let's turn up the heat to a 28C ambient. Then see how it does.

 

[unseenmorbidity]

Lol, a 1700 at stock clocks on a test bench. That sets a floor maybe, but we are more interested in a ceiling.

Do a 1800X @ 4GHz 1.4V and let's turn up the heat to a 28C ambient. Then see how it does.

The Tweaktown reviews used 4GHz, iirc. Though they only did the crosshair and the Taichi thus far.

EDIT: Actually, they used auto-voltage OC'ing...

 

[IEC]

JayzTwoCents did an OC review and got his 1800X to 4.1GHz and DDR4 running at 3200 CL14:
https://youtu.be/I71ov8iiDP8?t=76

He didn't know about the 20°C temp offset though...

 

[ultima_trev]

I think it is all bandwidth.
Higher speed RAM would help even more with the scores.
Here is my best run in a PC I built for a customer:
http://www.3dmark.com/fs/12079421

Your system has a higher clocked GPU, higher clocked CPU, but slower clocked RAM, and that offsets it.
My personal observation is that physics test loves bandwidth.
The 1700X was overclocked using Ryzen Master. I could ave gone to 4.0GHz in that chip, but the voltage required was insane (1.425V) so the 3.9Ghz at 1.3875V was a nicer balance.

Appreciate it, this is exactly the kind of data I was looking for.

 

[krumme]

Toms reviews 1700x and meassures power consumption under full load.
The 1800x uses less than a 7700k.


http://www.tomshardware.com/reviews/amd-ryzen-7-1700x-review,4987-8.html

This is going to be a big mess for Intel...not good.

 

[Mopetar]

Interesting chart, but should normalize for performance. I might not care about an extra 20W if the 7700K gets an extra 20 FPS.

Interesting nonetheless though.

 

[krumme]

Interesting chart, but should normalize for performance. I might not care about an extra 20W if the 7700K gets an extra 20 FPS.

Interesting nonetheless though.

In dx9 games right? I know where you are going but serious. Under full load in bf1 an 7700k will have frametimes that runs under like 40fps while a 1800 never gets below the 70.

The damn point is. under full load where a 1800x have a throughput that is like 60% more than a 7700k it uses slightly less energy. To me this looks like the cpu have a node advantage. It doesnt. It is just crazy efficient. And it goes for all loads to idle.

Bring this arch to 3 ghz levels and it looks to me a major part of Intel server line is seriously threatened.

This is bar none the most important metric and Intel simply doesnt look good.

 

[Accord99]

Toms reviews 1700x and meassures power consumption under full load.
The 1800x uses less than a 7700k.

In two tests which aren't true CPU tests. And the curious thing is that they used Autocad 2015 (assuming it's not a typo) for their power tests even though they used 2016 in their performance tests. And at least for 2016, the 7700K outscores the Ryzens significantly.

The damn point is. under full load where a 1800x have a throughput that is like 60% more than a 7700k it uses slightly less energy.

You can't tell from their measurements because they didn't do a power consumption reading using any such benchmarks where the Ryzen does have such a throughput advantage.

 

[krumme]

We also have a a 4c vega apu comming to mobile. Where is Intel i7 2c line here?It will look insane. A major part of their portfolio is soon under severe pressure.

That zen is highly compettitive even in desktop doesnt bode well for Intel. Zen on desktop is like 0.001% of the challenge they have from this arch.

 

[formulav8]

That zen is highly compettitive even in desktop doesnt bode well for Intel.

Looking top to bottom, Intel will Not be able to sustain their current profit margin/volume in most segments once AMD starts running on all cylinder's.

 

[ElFenix]

i don't think i've ever found any real satisfactory measurement of power energy consumption. how i'd do it:
a) fixed task to make it as apples to apples as possible. so, something like encoding a file. or, playing a game for 30 minutes on a loop with a fixed fps. telling me how much power is used in a benchmark run is useless. i don't play games seeing how fast a benchmark run can be completed, i play games until the match is over or until the story is over, and those things happen independently of how fast the frames can be drawn.
b) report the energy in kilowatt-hours consumed. i pay my power bill based on that. giving me peak wattage is completely useless. average would be ok if they used fixed tasks, but 1) i'm not sure they do; and 2) i'm not entirely certain how they reach average power consumption, since i'm guessing a bunch of their equipment reports in watt-hours (which is the proper unit anyway). i know, it should just be a simple mathematical identity. but since i haven't seen any explanation of how this stuff is done from pretty much anywhere, i'm suspicious.

 

[imported_jjj]

Summit Ridge seems to be the most comfortable up to 1V - so up to 3.3-3.4GHz.
It can do some serious damage in laptop and server.
It's almost annoying that they don't have a 45W octa core for laptop.

 

[sushukka]

In dx9 games right? I know where you are going but serious. Under full load in bf1 an 7700k will have frametimes that runs under like 40fps while a 1800 never gets below the 70.

The damn point is. under full load where a 1800x have a throughput that is like 60% more than a 7700k it uses slightly less energy. To me this looks like the cpu have a node advantage. It doesnt. It is just crazy efficient. And it goes for all loads to idle.

Bring this arch to 3 ghz levels and it looks to me a major part of Intel server line is seriously threatened.

This is bar none the most important metric and Intel simply doesnt look good.

Finally someone else here also addresses the Ryzen's position in server market. Breaking in takes some time sure, but power consumption is one of the major cost factors in data centers around the globe. Not only you need to distribute that energy, but you have to cool it down too. Having thousands of servers using tens of watts less each will for sure affect the business calculations when choosing the hardware. And the difference widens when the core amount rises. Intel's highest core amount per socket is now 22, price 4110$, TDP 145, but very low base clock 2.2GHz (highest over 8core Xeon base clock currently 2.6GHz). It's very interesting to see what the AMD's 32core Naple's brings to the table. At least more bandwidth, more max memory and 128threads vs Intel's 88 per one server. If AMD succeed here, their liquidity problems should be vanished for a while. As Naples is basically a pumped up Ryzen they have much more room for pricing on server side (Intel topping currently with that 4110$/CPU).

Of course someone would say that there is also Intel Xeon Phi, but price again higher and different market segment (machine learning etc, where also GPU based alternatives).

 

[Malogeek]

What? That dang thing still isn't out? I've been patiently waiting for this, pretty much ignoring the slow progress, assuming it would be ready if I just stop paying attention...that was about 2 months ago.

Well the team stated very early they were looking at March for an initial release.

 

[krumme]

Sure assessing power consumption is darn difficult. And will always be dependant on the specific situation.
A way to look at it is this:

Prime autocad metro idle all consistent points the same direction.
Bd is very bad at the same assessments.

Toms usually is quite good at this power stuff. Unfortunately they dont give us the details.
But so what. Look at the results. We know the stilts results and asumptions for lower freq of zen that will be applied outside of desktop.
When we get to 2.8 that is server and notebook base like freq it looks to me this arch is just far superior to intel for efficiency for many loads.
Its a guess but the data we have mostly point this way.

 

[DrMrLordX]

It really needs to be considered as 1:1 ratio as the fabric is running at the speed of the memory controller, but the DIMMs are double data rate.

The DIMMS are actually running at 8x data rate. DDR4-3200 runs at 400 MHz. The I/O bus clock is 1600 MHz, as is the mem controller clock and the data fabric clock.