Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

Page 11 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

TemjinGold

Diamond Member
Dec 16, 2006
3,050
65
91
That reminds me, given Ryzen is +52% IPC, what Intel IPC does that actually put it at? Haswell? Skylake? Something else?
 

MaxDepth

Diamond Member
Jun 12, 2001
8,758
43
91
So heatsinks aside... if the thermal ceiling is not going to eek out that much more in performance, then would a locked 1700 be a better value than a 1700x?
 

Absolute0

Senior member
Nov 9, 2005
714
21
81
You are correct if you are talking all-core turbo. That's a lot of heat to dissipate with a tiny die surface.

On the other hand, there are rumors of per-core OC controls from AMD, so I am optimistic that I can make lightly threaded apps (like some games) competitive with Skylake/Kaby Lake.

Intriguing but different clockspeeds on different cores sounds a bit wonky to me. Like, go into the BIOS and redo your overclock every time you want to game? Unless they are talking windows overclocking utility, which would be much more useful. Make it turbo 4C on game launch :).

As far as 5.1-5.2 GHz CB15 on LN2, it seems to be on par with Broadwell-E.
The 8-core CB15 results are surprising, matching 5960X scores @ 5.8-6 GHz @ 2350-2450 points. :')

I would imagine that 4.2-4.4 GHz on all cores should be the target for 1800X which has been my prediction all along but I guess we will see next week.

I wonder if the rumors of Zen not having a coldbug are true, I was expecting closer to 6 GHz on LN2 @ full cold?

Guessing even 4.4 Ghz 8C fully stable will be limited to good chip + watercooling. As mentioned even intel 8C have a hard time clocking high. Here I think we see poor voltage scaling. Which will likely increaes with Zen+ and Zen++ as the Si process matures (let's hope).

What gets interesting with this is that in 2018 when AMD is transitioning to Zen+ and maybe bringing clocks up past 4 Ghz, Intel should be launching 10 nm Cannonlake and they had better be using the 10nm process to pack 8C on there while maintaining the (up to 4.5 Ghz) clocks we've seen from 4C kaby lake.

We all know Intel's cadence is out of whack, I think Kaby lake was supposed to be the 10nm tick (delayed bc 10nm low yields) and cannonlake the 10nm tock, so, hope the uarchitecture change is there (or, else they are going to blow it early with coffee lake, idk, so much speculation).
 

plopke

Senior member
Jan 26, 2010
238
74
101
fubXp0M.png


Preorder prices in Belgium , so that is 21% VAT included and in general Belgium is one of the more expensive countries in the euro zone to buy hardware.
 

Insomniator

Diamond Member
Oct 23, 2002
6,294
171
106
Stupid question -- these don't support DDR3 right? I have 32GB.....

Also, anyone know when mATX boards are being released?
 

AtenRa

Lifer
Feb 2, 2009
14,001
3,357
136
So IPC is slower than Broadwell in Cinebench R15.

Core i7 6900K ST score 162cb at 3.7GHz (max ST clock)
Ryzen R7 1800X score 162cb at 4.1GHz XFR (Max ST clock)

2nksbk6.jpg
 
  • Like
Reactions: Drazick

CatMerc

Golden Member
Jul 16, 2016
1,114
1,149
136
Well, they are not submitted to hwbot, so i guess so.
https://vid.me/ZVcf

Found video mirror, look at the 3 minute mark or so.

So IPC is slower than Broadwell in Cinebench R15.

Core i7 6900K ST score 162cb at 3.7GHz (max ST clock)
Ryzen R7 1800X score 162cb at 4.1GHz XFR (Max ST clock)

2nksbk6.jpg
She did say results would be better with a better cooler, which might mean XFR wasn't actually boosting it in this case (though I doubt it).

This does imply AMD's SMT implementation is superior like rumored.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,149
136
XFR was never stated to work for multithreaded workloads, actually.
Also, Broadwell almost certainly had 4Ghz single core boost thing going on, because this result is greater than usual 3.7Ghz Broadwell result. So yes, Zen is a smidge behind in IPC in Cinebench.
https://youtu.be/1v44wWAOHn8?t=12m

This comment from Lisa seems to /possibly/ imply XFR was not in action here.
 
Mar 10, 2006
11,715
2,012
126
XFR was never stated to work for multithreaded workloads, actually.
Also, Broadwell almost certainly had 4Ghz single core boost thing going on, because this result is greater than usual 3.7Ghz Broadwell result. So yes, Zen is a smidge behind in IPC in Cinebench.

The Broadwell single core boost is Turbo Boost Max 3.0 which requires a lot of work to set up. I would wager that they just used the true turbo boost 2.0 speeds.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,149
136
Wait a minute, guys, i think i just noticed something really weird.

Why the hell does Zen have almost 5 BILLION (4.8 to be exact) transistors?
For reference, it is almost the same transistor density as Polaris, and much much larger than Broadwell-E's.
That was by far the most shocking discovery for me.
The two CCX's should take 2.8 billion transistors together. Where the hell are the other 2 billion?
 

Headfoot

Diamond Member
Feb 28, 2008
4,444
641
126
MicroCenter is marking up the list price by $30 - $50 and then putting them on sale for $30 - $50 off.:mad:

They need a serious slapdown for this.
Dang MicroCenter hasnt played that kind of game in the past... disappointing