[Guru3d]AMDs 8 core Zen compared against Intel's 5960X - Page 8 - Overclock.net

Forum Jump: 
Reply
 
Thread Tools
post #71 of 135 Old 08-13-2016, 11:43 AM
retarded guru
 
sumitlian's Avatar
 
Join Date: Jun 2011
Location: India
Posts: 3,212
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 312
Quote:
Originally Posted by prjindigo View Post

So the 5960X reports 3.2 on the clock, was it running at 3.2 in the test? Nope. Betcha the Zen stayed right at fixed clock speed the whole time. Kinda puts these garbage reviews to task, doesn't it?


Further, the info I have is that the 8/16 Zen right now is locked at 2.8GHz and doesn't boost either. There was another joke of a review that was trying to compare Zen to a 4770 series CPU... hands down the single most powerful simple gaming i7 on the market at stock clocks. There's all sorts of spurious garbage going on in this review as well as the other that make them both useless. The 5960X is a very very mature CPU with very very mature drivers on a very very mature board. We're past the 3rd generation of LGA2011, the Zen is gonna be on a board with new drivers as well. But the problem I have here is the 5960x is a 3.0 to 3.5 processor... why is it reporting 3.2GHz?

And not to be too blunt about it... who gives a flying flaming farting damn WHAT the CPU rendered graphics rate is? Is there someone out there moronic enough to NOT buy a graphics card who's trying to use a firewire connector over HDMI to display his gameplay? AotS has to know what to do with a CPU in order to render through it, that's special custom code and it looks like what we're seeing is simply AotS driving the Zen 8/16 like a BullDonkey 4/8 and displaying the improvements of Zen over the original bogging-down problems inherent in the SMT-stripped PlinkoDoohicky family.


There is ZERO chance that AotS is driving the Zen processor correctly and a 99% chance that it's driving it like a standard FX chip. So the only Apples to Apples on that chart is the Zen vs 8350. In that it's also software render of a game we might as well discuss a chart comparing how well an oven and a dishwasher do laundry.

Pardon me if I didn't understand everything.
Going with you last lines, can it be due to the Intel compiler's non genuine Intel parts are kicking in ? redface.gif

sumitlian is online now  
Sponsored Links
Advertisement
 
post #72 of 135 Old 08-13-2016, 11:45 AM
New to Overclock.net
 
Blameless's Avatar
 
Join Date: Feb 2008
Posts: 29,121
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 3129
Quote:
Originally Posted by Asmodian View Post

The CPU benchmark doesn't have to be completely CPU limited, they used the SAME GPU and drivers for the 5960X test.

Which doesn't rule out effects of an immature platform or memory bottleneck.
Quote:
Originally Posted by Asmodian View Post

Same with the game leveraging more than eight threads or not; who cares?

Anyone trying to compare it to Vishera because while both the Zen and the FX-8370 both have eight cores, the former has SMT while the latter does not.
Quote:
Originally Posted by Asmodian View Post

The AotS CPU bench is very representative of CPU performance in AotS, which all that can be measured using AotS.

Rare is the benchmark that is only representative of itself.
Quote:
Originally Posted by Asmodian View Post

we aren't sure the benchmark is scheduling threads on Zen correctly (though if it needs special tuning by software developers compared to Intel CPUs it would be a big negative)

At the very least, the OS needs to know what logical CPUs are attached to which physical cores to schedule things correctly. This isn't a certainty unless the OS can accurately identify the CPU. It took a while for MS to patch the Windows scheduler for both Intel's HT and AMD's CMT/Modules.

Absent specific info, the OS or app will generally assume all the logical cores are the same and might be putting two demanding threads on the same physical core, which could greatly distort performance impressions.

At the very least, pre-release tests should feature manual affinity testing to rule out major scheduler issues (because near release, these will be patched out).

...rightful liberty is unobstructed action according to our will within limits drawn around us by the equal rights of others. I do not add 'within the limits of the law,' because law is often but the tyrant's will, and always so when it violates the right of an individual. -- Thomas Jefferson
Blameless is online now  
post #73 of 135 Old 08-13-2016, 11:48 AM
 
warm's Avatar
 
Join Date: Mar 2013
Posts: 223
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 4
i was hoping amd to smash intel next year so we could see some 6-8 core i5's, but seems like that's not happening anytime soon ;/

help
warm is offline  
Sponsored Links
Advertisement
 
post #74 of 135 Old 08-13-2016, 12:07 PM
WaterCooler
 
Asmodian's Avatar
 
Join Date: Aug 2009
Location: San Jose, California
Posts: 2,069
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 173
Quote:
Originally Posted by Blameless View Post

Rare is the benchmark that is only representative of itself.

You were the one speculating whether or not the AotS CPU benchmark was representative... as you say, it probably is.
Quote:
Originally Posted by Blameless View Post

At the very least, the OS needs to know what logical CPUs are attached to which physical cores to schedule things correctly. This isn't a certainty unless the OS can accurately identify the CPU. It took a while for MS to patch the Windows scheduler for both Intel's HT and AMD's CMT/Modules.

Absent specific info, the OS or app will generally assume all the logical cores are the same and might be putting two demanding threads on the same physical core, which could greatly distort performance impressions.

At the very least, pre-release tests should feature manual affinity testing to rule out major scheduler issues (because near release, these will be patched out).

This is true, maybe Windows' scheduling needs to be tuned for Zen. We will have to wait for more benchmarks.

However, given this very limited leak, it is looking like matching Haswell is well out of range for Zen. We cannot be sure but the first hint is negative, not positive. frown.gif
Asmodian is offline  
post #75 of 135 Old 08-13-2016, 12:59 PM - Thread Starter
 
iLeakStuff's Avatar
 
Join Date: Jun 2015
Posts: 3,685
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 216
Quote:
Originally Posted by kylzer View Post

Zens intention was to never to take the performance crown

It was to be competitive for the price and build a good platform (am4) to expand upon.

This is exactly the strategy AMD used with Bulldozer because the architecture was so poor against Intel CPUs.

The result was a company almost going bankrupt.

Sounds like a stupid thing to try again. Especially if they try the "more cores for less money than Intel" approach
Very few general and mainstream applications and software need more than 2 or 4 cores. That includes the workload for majority of users.
Lacking in IPC but trying to make up for it in cores and threads could be a way in to the server market and regain marketshare there, but will fail mostly elsewhere.

Could be that Intel is gunning for the professional market. But what do I know
iLeakStuff is offline  
post #76 of 135 Old 08-13-2016, 01:12 PM
*cough* Stock *cough*
 
Marios145's Avatar
 
Join Date: May 2011
Posts: 344
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 70
Is it a 65W or 95W part, desktop or mobile, OEM or retail?
Do CPU-NB-HTT run at final speeds?(running them <2Ghz impacts performance
Did they use high freq ddr4 or 2133?
Is it running dual channel or single?
Did the engineering sample run at 2.8 or 3.2 during the test? AMD processors only use max turbo for single-thread workloads
Is aots capable of scheduling work for physical cores FIRST, THEN virtual cores on an unknown cpu?

Does aots have 100% scaling for every core you throw at it?
Does Zen arch have 95%+ scaling in multi-threading?
Single-core IPC is impossible to predict from a multi-threaded benchmark.
Take cinebench for example, you can get 7.00 MT score from an 8core if you just assume perfect scaling you would expect ~0.9 ST score, but actual ST score is 1.1


The only thing you can safely say about this sample is that it's way better than piledriver, which would be a given anyway
Marios145 is offline  
post #77 of 135 Old 08-13-2016, 01:13 PM
 
Piraal's Avatar
 
Join Date: Feb 2015
Posts: 144
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 3
Quote:
Originally Posted by outofmyheadyo View Post

So it cant even compete with a few years old intel lineup? What is the exact point of zen then?
Pretty sure AMD has some crazy scheme to make money. Even if these numbers are true, not competing with the top end Haswell-E doesn't mean not-profitable.
Piraal is offline  
post #78 of 135 Old 08-13-2016, 02:16 PM
New to Overclock.net
 
Blameless's Avatar
 
Join Date: Feb 2008
Posts: 29,121
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 3129
Quote:
Originally Posted by Asmodian View Post

You were the one speculating whether or not the AotS CPU benchmark was representative... as you say, it probably is.

I'm still speculating and I'm still not sure. It seems like it should be representative of overall CPU performance...up until a certain number of threads and as long as the GPU is fast enough.

Since I'm not going to spend 40 dollars on a game I'd probably never actually play, nor am I keen on installing Windows 10 on any of my decent systems, I have to rely on second hand tests for my AotS info, and very few of them go into any real depth in their benches.
Quote:
Originally Posted by iLeakStuff View Post

This is exactly the strategy AMD used with Bulldozer because the architecture was so poor against Intel CPUs.

The result was a company almost going bankrupt.

Sounds like a stupid thing to try again.

Bulldozer and it's first few successors were also relatively expensive to make and incredibly inefficient, which kept them from being good server or mobile options as well as drastically reducing profitability on the desktop.

Even if Zen winds up in a similar position performance wise (and even the more pessimistic interpretations of this leak don't suggest so dire a situation) the part can still be profitable and highly viable, as long as other mistakes aren't repeated.

Ideally, Zen would directly compete with Intel's best, but that's always been a highly optimistic and improbable scenario. To help AMD, Zen just needs to have a significant niche and be profitable; it doesn't need to dethrone Intel's raw performance or IPC kings.

...rightful liberty is unobstructed action according to our will within limits drawn around us by the equal rights of others. I do not add 'within the limits of the law,' because law is often but the tyrant's will, and always so when it violates the right of an individual. -- Thomas Jefferson
Blameless is online now  
post #79 of 135 Old 08-13-2016, 02:44 PM
Mobo Master
 
nakano2k1's Avatar
 
Join Date: Sep 2011
Location: Montréal, QC
Posts: 830
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 85
Intel's R&D budget will be just below 12 billion dollars for 2016. AMD on the other hand is spending just a little north of 1 billion. The fact that they're even coming out with something remotely competitive is impressive.
SuperZan likes this.

nakano2k1 is offline  
post #80 of 135 Old 08-13-2016, 04:55 PM
New to Overclock.net
 
SuperZan's Avatar
 
Join Date: Sep 2015
Location: Latveria
Posts: 3,593
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 641
Quote:
Originally Posted by nakano2k1 View Post

Intel's R&D budget will be just below 12 billion dollars for 2016. AMD on the other hand is spending just a little north of 1 billion. The fact that they're even coming out with something remotely competitive is impressive.

 

Oft-overlooked in these discussions. In their battles with Nvidia and Intel the name of the game for AMD is market-share recovery. That's how they win back mind-share and thus long-term profitability. They also need improved professional performance and so need architectures and designs that parley well into those markets. Aiming to service the mainstream retail and professional markets is what will regain market share for AMD. They know that they haven't got the budget to come over the top of Nvidia and Intel. How on earth would it make sense to once again attempt rough enthusiast-class parity a la Fury X when they can achieve performance targets for mainstream users at a better price-point than Intel or Nvidia? We enthusiasts are the minority and the only scenario in which it would make sense for AMD to go broke beating Intel on a tiny fraction of overall sales is in the mind of a person who wants a couple hundred quid off their 6950x.

 

I predict that Zen's mainstream desktop processors will be roughly competitive with Haswell mainstream i5's and i7's whilst their new server variants will win back at least a bit of market-share on the respective backs of solid performance and aggressive marketing. HEDT realm may be challenged by Zen+ if Zen ends up bringing in the cash-flow required to invest in that market.


"You are individuals. You are small and you think in small terms." - Seven of Nine, Tertiary Adjunct of Unimatrix 01
SuperZan is online now  
Reply

Quick Reply

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off