Printable View
Is it just me or is Charlie a bit of an AMD fanboy? I cringed upon reading his 'Inqpression' on 4x4, he used absolutely NO benchmarks (I wonder why) and blabbered on for 2 pages about how Kentsfield is FSB limited, and that 4x4 would outrun Kentsfield in multithreaded games (with no benchmarks whatsoever to back up that claim).Quote:
Originally Posted by Fred_Pohl
he does have a rather valid point, that Operating system does not support Numa and will effect 2P system performance by 15-25%Quote:
Originally Posted by Epsilon84
you are really starting to annoy meQuote:
Originally Posted by Fred_Pohl
if i read this post its clearly you dont understand anything about Zram and AMD
4x4 got kicked hard aswell in Vista, XP64 and 2003. So its not a valid point and didnt make the change hoped for in performance. and 15-25% for NUMA or not? LOL!Quote:
Originally Posted by nn_step
Theinq is basicly useless in terms of tech info. Charlie and Fuad Fud is the worst people around and its all about tabloid news. They been wrong more than right, and you gotta love the best OC platform. Going from 680i to RD600. Tho the champ is P965. And Fuad with his 4GB will only work on Vista and told us Vista changed everything. Tho the problem at that time was chipset limitations and PCI resources being mapped under 4GB. No OS can fix that. And when asked for prof, his "technicians" had just reinstalled the box.Quote:
Originally Posted by Epsilon84
And all the product bias they have, its simply not worth reading. And we all now much sucky Kentsfield/Clovertown obviously is with the "limited" FSB :rolleyes:
Or with Charlie and his Kentsfield will OC like crap and its FSB OC potential sucks. Tho something like 1600FSB over the official 1066FSB in his eyes was poor.
4x4 is a platform not a CPU. The CPU isnt the single important thing in a system anymore.Quote:
Originally Posted by Shintai
Dual Socket CPUs (4 cores now, 8 later), Quad Video, terrabites of data, cool look.
And then, for the CPU component, the only benches that matters for 4x4 are the multithreaded & high threading situations.
This is what it was designed for, to handle huge threading situations, and it does this better than Kentsfield, QX6700 etc, 5160, etc.
For example, the 3.2 FX74 beats the 3.0ghz 5160 in W2003 X64 (or Vista64) in rendering (one of the few areas of modern apps that
are efficiently multithreaded) and also, in general conditions, it scales much better as the number of threads increases.
it is very limitted but don't expect to find it's limits in game testings or SuperPI parties.Quote:
Kentsfield/Clovertown obviously is with the "limited" FSB
How many apps even multithreaded does it beat it at? And FX74 is 3.0. Maya, mental ray and 3DS Max Kentsfield is still faster. And I dont see many professionals rendering in cinebench...rendering is also one of the least memory bandwidth applications.Quote:
Originally Posted by alayashu
http://www.extremetech.com/article2/...2065498,00.asp
And NUMA+Vista+64bit
http://babelfish.altavista.com/babel...o-core_10.html
maya, max, xsi /w mentalray || vray || any other modern renderers are exactly theQuote:
Originally Posted by Shintai
apps where 4x4 surpasses the QX6700 (ofc on X64 /w NUMA support).
Better than in Cinebench actually.
also, as a side note, their Max bench uses the Scanline renderer wich is very
poorly optimised for multiple cores (and actualy is very rarely used in 2006 CG
world, the same for LW Renderer, both of them are 90' outdated engines)
and aswell they did the test under XP32.
Undiscussable the Dual FXs are better in modern rendering than QX,
and the same in heavy threading situations. Ofcourse under X64 /w NUMA.
Benchmarks dont agree with you. :fact:Quote:
Originally Posted by alayashu
the benches you posted are relevant for the editor's professionalism.Quote:
Originally Posted by Shintai
No one in modern times uses X32 anymore in rendering area
and max scanline or LW renderer or POV Ray are 90' engines that
very few use today (mainly cause of scene compatibility issues with
newer 3rd party renderers).
Do yourself a favor and browse for some mentalray/vray/renderman
tests under X64 w2003 if you don't believe me. Opterons / FXs are
slightly faster in every modern renderer than their Woodcrest/Kentsfield
counterparts. I say it as a lighting artist and not as fanboy as you
seem to consider every one that says FXs are better in certain areas.
Why dont you post some? Or perhaps you cant find any that gives the result you want? :rolleyes:
Because thats not the result we got at www.afilm.com but what do we know, we only make movies.
http://www.hexus.net/content/item.php?item=6256&page=7
Dual Woodcrest 3.0ghz Vs. Dual Opteron 2.6ghz
The last two (Mental Ray Standalone and Pro/Engineer) are 64 bits.
Um, On vista X64 and X86. Compared to XP X23800 no OC and my conroe... both are fast on vista period!
umm, then probably you're involved in something totally different than rendering,Quote:
Originally Posted by Shintai
as what you want to prove is laughable in rendering arena :P
as for own tests, I wont be able to post in these 2 weeks, the studio is closed during holidays.
But anyway, PetNorth already proved what it is to prove.
Merry Christmas
Some rendering 64-bit benches under Vista:
http://www.hwupgrade.it/articoli/cpu...o-core_10.html
basically showed that AMD is pretty close in low Bandwidth renders and takes a soaring lead in high Bandwidth renders.
It depends on what you call "to be close". FX-74 has higher frequency (but still losing) and it is far-far-far away from Kentsfiel in terms of performance-per-watt.Quote:
Originally Posted by nn_step
I do tend to annoy AMD fanboys with inconvenient facts but I honestly don't care. As for AMD using ZRAM cache in 2006, that certainly wasn't my idea. You'll have to ask The Inq about that ridiculous rumor. While you're at it, please ask them what happened to that miraculous RHTT technology that promised to allow a dual core AMD CPU to split single-threaded code into two threads and process it 25% faster with nothing more than a microcode update and OS patch. If you ask very nicely, they might also tell you how to convert lead into gold in your bathtub...Quote:
Originally Posted by GoThr3k
at the same time a Via C7-m at 1w and 2Ghz completely OWNS kentsfield in terms of Performance per watt. :fact:Quote:
Originally Posted by kl0012
So what is your point?
The two top ends are within a handful of points of each-other.
on a more ontopic note, why must EVERY SINGLE AMD thread talk about Intel?
Quote:
Originally Posted by kl0012
So what's new? It's a 90nm chip, you know. :stick: It would be strange if it was different.
Why are we talking about FX74 in a K8L thread?
There is no such CPU.Quote:
Originally Posted by nn_step
Wrong!Quote:
completely OWNS kentsfield in terms of Performance per watt. :fact:
I think you got the point.Quote:
So what is your point?
Why are people still trying to find some positive points in 4x4. Imho it is competely useless platform at least for the next 2 quarters.
1) yes there is, ask Via for a 65nm sampleQuote:
Originally Posted by kl0012
2) at less than 1/89 the power, it would only need 1/89 the performance to have the same performance/watt. Since performance in that benchmark doesn't scale perfectly, It would be impossible for Kentsfield to win.
3) Don't be so quick to judge. It is still a wonderful value for a good 2P system.
Topic: K8L
Quote:
Originally Posted by Fred_Pohl
Yeah, and intel fanboys annoy me too.. :cool:
Let me take that back 98% of intel fanboys annoy me.
~Mike
This is getting off topic but then the topic is moot anyway since there will be no K8L ES chips this December. AMD fanboys occasionally annoy me but usually they just amuse me with their fanciful rumor-based wishes that never seem to come true. I can't think of a single AMD fanboy rumor-wish that came true this year. :fact:Quote:
Originally Posted by arisythila