Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

Page 80 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Magic Hate Ball

Senior member
Feb 2, 2017
290
250
96
I accidentally got him in trouble, but was not fully liable, the mod of the forum we discussed forgot to make the link untraceable and it was traced to source, he got a reprimand from AMD. Basically his information to me is me asking questions IE: if my calculation is correct, Ryzen at 3.7Ghz scores 1602 and at 4.1Ghz scores ~1750 give and take .15% margin of error depending on setup. and he would respond by say, I can confirm that is 100% correct.

He has had to correct information regularly but that is due to AMD firmware updates, initial samples had horrible turbo boost issues to the point it was just more sane to turn off. Lets say I trust him, and thankfully he didn't get removed from AMD's testing list and thankfully he has been able to help sort the fake from teh real. I think there may be one ore two secret surprises to come, but we can accept it is far and beyond what Bulldozer was. In fact I have practically forgotten about that AMD.

I have somewhat of a fancrush on Lisa Su, I will admit I didn't like the decision at first but I got it wrong.

You realize by confirming that this is the same guy leaking you information, in a previously documented case with AMD, that you might get him in trouble again?
 

itsmydamnation

Platinum Member
Feb 6, 2011
2,743
3,075
136
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.
What are you talking about? are you clueless?
Have you missed the ENTIRE DESIGN METHODOLOGY AROUND ZEPPELIN???

How many broadwell-E's can i link together to create a 32core 8 channel memory 128 x pci-e v3 monster. 1 flexible SOC that is competing against 4 different dies from intel. Yes that die size is a dangerous problem for amd :rolleyes:.

Its an easy 25% of the die for GMI + associated infrastructure ( like those 2 massive sram patches outside the CCX's etc) that does nothing for the consumer zen but at the same time unlike the GPU in kabylake it doesn't matter if its defective.
 

OrangeKhrush

Senior member
Feb 11, 2017
220
343
96
You realize by confirming that this is the same guy leaking you information, in a previously documented case with AMD, that you might get him in trouble again?

The good thing is nobody believes anyone on a forum and that is really how it should be, I could sit and tell you how you are gong to die, it means jack. Having been in the legal profession there are ways of asking that confirms what you know without the other party making disclosure.
 
  • Like
Reactions: lightmanek and IEC

looncraz

Senior member
Sep 12, 2011
722
1,651
136
I don't know why you are making fun of anybody who said that Ryzen had Sandy Bridge IPC. Based on AMD's 40% claim, that's where it would have landed. There is no way for anybody who was making predictions before these announcements to know that AMD was sandbagging and actually got 52%.

My top guess was 49%. Two weeks shy of one year ago:

est-final.png
 

OrangeKhrush

Senior member
Feb 11, 2017
220
343
96
If we are talking about Juanrga and his Sandy IPC, he gets knocked because he was forewarned that AMD sandbagged for obvious reasons, 40 was a target but nobody just aims for a target, this created a situation where Intel believed like Juanrga in that so much they dropped the ball on any variable. Then to still argue that after being proven wrong is just downright worthy of internet mockery.
 

hotstocks

Member
Jun 20, 2008
81
26
91
Ok, so Asus doesn't have any 32gb kits listed for 3200 mem. So what is going to work 4 x 8gb 3200 sticks or 2 x 16 3200 sticks ? I would think that 2 X 16 would be much easier on the mem controller and work, but it is scary that Asus doesn't list any 16gb sticks at 3200...Maybe you are limited to 16gb at 3200, which would suck because I am NOT building a new rig with 16gb of ram.
 

lolfail9001

Golden Member
Sep 9, 2016
1,056
353
96
What are you talking about? are you clueless?
I am talking about die size, man. By the way, you have already ate your words on PassMark latency report, didn't you?
Have you missed the ENTIRE DESIGN METHODOLOGY AROUND ZEPPELIN???
"Let's make a small core and LARGE ARSE interconnect that takes as much as space as four cores and 8 mb of L3". That sort of philosophy? Because i don't remember the second part. You have a point though, i sometimes forget that Zeppelin is just like HEDT is just a server reject.
How many broadwell-E's can i link together to create a 32core 8 channel memory 128 x pci-e v3 monster. 1 flexible SOC that is competing against 4 different dies from intel.
Obligatory reminder that these supposed "monsters" are only useful for scale-outs.
but at the same time unlike the GPU in kabylake it doesn't matter if its defective.
You forget that Intel salvages defective GPUs just fine as E3 Xeons.
 

Riek

Senior member
Dec 16, 2008
409
14
76
Sure, it is actually the hardest to think about, but whatever floats your boat. And i was talking about 1700 vs 7700k, so you are looking at wrong numbers in the first place.

What if i told you that it is not a 50% advantage but 50% disadvantage? You can check with trivial multiplication that it is not Watts/EncodeTime (because then it turns out i7 was consuming 1250 watts. If it is EncodeTime/Watts, then it turns out 7700k was consuming less power (you divide 30% higher number by 100% higher number). The only options remaining would be that it is somehow 1/(Watts*EncodeTime) or Watts*EncodeTime, in which case it still makes no sense Of course there is a chance it is simply joules/encode time, but then the number is clearly too small for both.


Looks like a diehard AMD fan from his earlier comments. But definitely like the last person with any sort of inside info, especially considering his tendency to jump on conspiracies and usage of rx480 crossfire.


FYI: i rectified my post in less than a minute, so try not to be condecending will you.


how the calculations happened is our guess, only thing i'm seeing is that the perfromance difference is smaller than the perf/W difference.

Simply put if I can do 2 times your job in the same hour. but I only did 1.5times your job, i wouldn't need that one hour.


So if my W/perf is bigger than the perf difference, my power consumption should be lower (not higher).
 

OrangeKhrush

Senior member
Feb 11, 2017
220
343
96
apparantly very good results in gaming with a 1700X, enthusiast level 3200Mhz CL14 used on all platforms to make it fair and it is going to make some people cry, others amazed, and some just going to get their bounce on.
 
  • Like
Reactions: lightmanek

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
Yep I believe this is Media Bios effect :D. Jayztwocents is, just like CPC HW, very salty. It can be even seen in his videos from the last week.

edit:
Found this on Sisoft site:
http://ranker.sisoftware.net/show_r...e6d5ecdfecdbeec8ba87b791f491ac9cbac9f4cc&l=en
68ns latency to the memory

http://ranker.sisoftware.net/show_run.php?q=c2ffcdf9dfbedfe2d1e9ddedcbb984b492f792af9fb9caf7cf&l=en
16ns latency to the memory

Same sticks, same amount

lol 16ns memory wat ?! o_0

If this turns out to be true, i think many are in for a rude awaking in regards to latency performance :)
 

inf64

Diamond Member
Mar 11, 2011
3,685
3,957
136
This is a quote from livebriand @ SA forum:
livebriand said:
There's a lot of issues affecting results right now. SMT harms performance in some games, BIOS in general is really bad but being updated daily. Windows also needs to be set to "high performance" power mode. As Fott just aluded to, you can relax about memory performance.
wink.gif


It's just the kind of stuff you might expect with a new arch, but makes me wonder just how much of a monster Ryzen will be after all the issues are ironed out.

Any weird results you see are likely down to the above issues so keep that in mind, especially as we only found out about much of it a couple of days ago. I still don't have my sample due to a bunch of unlikely mishaps and I won't be getting it in time for the reviews tomorrow now. Bit of a shame as I'd already put a crapload of work into this, new equipment etc, but ************ happens.

Early adopters bugs in effect. Will get ironed out eventually. Hopefully soon.
 

CentroX

Senior member
Apr 3, 2016
351
152
116
I feel bad for those who preordered asus prime pro. Releasedate seem to be march 17th lol.
 

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136
lol 16ns memory wat ?! o_0

If this turns out to be true, i think many are in for a rude awaking in regards to latency performance :)

That does sound almost too good to be true, but then again so much surrounding zen seemed that way at some point.

Assuming it is true, where do the goalposts get moved then? You'd think they're about out of places by now. :p
 

hotstocks

Member
Jun 20, 2008
81
26
91
OrangeKrush,
So you are saying it is safe to buy a 16gb X2 =32 GB 3200 CL 14 memory kit and it will work on a good board at those speeds and latency, like Asus crosshair hero vi? Or will you have to relax the settings for two 16gb sticks and better off saving money with cas 15 or 16? Or go 3400 or 3600 cas 16 or 18?