new AMD benchmarks

[TAFKAT]

Originally Posted by jcschild

got to say this is the first time i have seen a new processor be slower than an old one.

First for AMD, but Intel pulled this stunt , remember the first Netburst Willamette PIV's, they were garbage, the earlier PIII's killed them... :-)

Man, AMD have dropped the ball big time , and of course now with the pressure off Intel, they'll take their sweet time rolling out the single socket Octocore... :-(

Those results are laughable , the original i7 920 that is now 3 years old, has 1/2 the cores at a far lower clock speed wipes their clock.. !!

[PaPi61]

This is bad news indeed. Competition is what keeps prices reasonable. If AMD can no longer keep up with Intel, the latter can start charging whatever they want.

[Mege]

Considering how bad recent benchmarks have been for Bulldozer, I'm surprised that the 6100 was only a little behind the 1090T at the 32 Buffer. This looks like a huge misstep for AMD (at least marketing-wise). You can look at the 6100 as a 3-core chip with an extra integer processing unit in each core (in which case pricing is way out of line). There's not too much relief in sight as there is only a 10-core scheduled for release as the next top chip on the Bulldozer platform, unless there's a drastic price drop.

[Finnish]

Sandy bridge, allah aghbar....

[mattiasnyc]

Originally Posted by jcschild

got to say this is the first time i have seen a new processor be slower than an old one.

Scott
ADK

Yup, all the benchmarks I saw basically indicated the same general tendency, and it ain't pretty.

Some guy did a bunch of benchmarks where he deactivated one core per module so the remaining cores didn't have to share L2 and float, and the results were instantly better.

Incredibly strange design in my opinion.

Only thing stranger is their pricing. If they had dropped the prices by about $50 or so then at least one could have said that it was an uninspiring update that sits where it should sit. But as it is now it's just plain... bizarre....

[q_h]

AMD FX 6100 vs 1090T ? Why not FX 4100 ?

AMD FX 8150 - 8120 - 6100 and 4100 performance review

[Mege]

I thought about making that point too but the FX-6100 and 1090T are similar in price and both are considered 6-core processors (although the 1090T is closer to the top of its range in the previous product lineup).
Price-wise, the FX-8150 is comparable to Intel's i7-i2600 while the FX-8120 would go up against the i5-i2500K. The FX-4100 would compare to the Phenom II x 4 965 or the i3-i2105

[tarnationsauce]

[VenVile]

As a long time AMD supporter and user, I am incredibly disappointed with this launch (after all the hype) and their shuddy PR work trying to dim the flames they started. From site to site, benchmark to benchmark, the new FX line consistently fails. It's new architecture, and it does have some pros to it, but it simply isn't very efficient at all. Five year old tech is completely trumping it in nearly everything. I really wanted a repeat of when AMD launched their first FX line, the Athlon 64; they wiped the floor with Intel at that period of time. But alas, that time has come and gone. It's a sad day for the PC consumer indeed. Competition is good...and we don't have that now. Their 2nd Gen Bulldozers better get up to scratch, or else...

[GoldMember]

#1. RXC test is stupid...
use T-Racks 3

#2. AMD Opteron 6272 is the flag ship of AMD.

#3. cpubenchmark.net rules.
AMD is the fastest SMP cpu in the world.
PassMark Software - CPU Benchmarks - Multiple CPU Systems

the price of a single 6272 is $500usd. aprox.
purchasing FX-8120 @ $190usd. is wasting time for a serious application, it's just a game processor.

i purchased an i7 970 last week because i had an empty x58 board.
but if i didn't had that board, i would have purchased the Operon 6272.
was a very hard choise to make...
wasting/selling cheap a top of the line x58 board to purchase a new compatible board for the 6272.
or puchasing an expensive 1st gen i7 cpu to avoid wasting that board.
price was similar between i7 970 vs. Operon 6272.
12 HT cores vs. 16 cores.
3.2ghz vs 2.1ghz.
9,864 vs. 10,245 PassMark Intel vs AMD CPU Benchmarks - High End points.
non-expandable vs. MP expandable with supermicro
Supermicro Products | Aplus | Motherboards | H8QGL-6F
or
Supermicro Products | Aplus | Motherboards | H8QGL-iF
also i have RME hdsp9632 & Lynx AES16 pci cards...i need a startech pcie-to-pci expander.

top of the line Opteron is the 6284SE, but according to cpu-world.com a faster 6286SE is planned.

CPUs always have a power mid-range level where price & perfomance is best.
i7 970 for 1366 / x58
Opteron 6272 for G34 1944
arround 10k points, for ~$500usd. aprox.
FX-8120 ... 8,169 points or 9,517 overclocked @ $190usd. is a non-expandable bargain.
purchasing any different cpu is wasting money or time, unless you have critical mission applications & unlimited budget.
like HPC clusters.

AMD also has ECC capability, only Xeon have that.
ultra important if you have a RAMDrive or PT10 to load & edit audio projects.

6272 is the best price vs performance CPU in the world for SMP applications. the 6284SE is even better & the 6286SE should be even better.
Bouncing in "Cubase/Nuendo/Sonar/Reaper/etc...", unless you have UAD, TC Powercore, or GPU plugins like gpuimpulseverb.

for bad coded Uniprocessor applications, an Overclocked Intel xtreme i7 3960x should be better.
16k points with 6x or less cores arround ~$1000usd.
PassMark Overclock Cpu Database Graph

[ivansc]

I have always fought shy of opterons and xenons, but maybe next time around it would be worth pricing up a system.
Any real downside to using these industrial grade cpus for audio/midi applications?

Those numbers are pretty interesting though.
Thanks too for the link.

[jcschild]

[QUOTE=GoldMember;8183262]#1. RXC test is stupid...
use T-Racks 3

#2. AMD Opteron 6272 is the flag ship of AMD.

#3. cpubenchmark.net rules.
AMD is the fastest SMP cpu in the world.
PassMark Software - CPU Benchmarks - Multiple CPU Systems

absolutely hillarious somehow a test FOR AUDIO is stupid but a synthetic test for cpu is ok...

yeah what ever you say buddy.. they dont come anywhere close. to intel..
anything you say to justify your sad pruchase..

Scott
ADK

[Bassmankr]

Goldmember, that linked Passmark CPU Overclock database results seems unrealistic in that it doesn't reflect the real world. Many of the top end chips just don't overclock well or even overclock at all if locked or the motherboard won't let them. Additionaly that graph handicaps CPU's that DO OVERCLOCK WELL such as the reasonably priced 3770K and 2700k to just a 3.5 Ghz overclock (both of which I've seen recently priced for only $260), which is laughable as even with autoclocking done by some of the motherboards you get at least 4.3 Ghz with them on air and a cheap decent cooler (under $40). You should just post your rig using Dawbench as that is specific to what we use here and will be an APPLES TO APPLES comparison. All other benchmarks are fairly meaningless for our purposes. One has to remember you can still get over 100 tracks with plugs and low latency with an old single core box that uses the right components and is tweaked right.

[Syncamorea]

Originally Posted by Bassmankr

One has to remember you can still get over 100 tracks with plugs and low latency with an old single core box that uses the right components and is tweaked right.

Good point. When people start drooling about pushing a big OC on a current generation Intel quad or hex, the thing I ask is: do you need that power? Certainly, some studios do, especially for video editing, 3D modeling, etc. On the audio side, a stock clock 2700K will handle quite a load.

[TAFKAT]

Originally Posted by GoldMember

#1. RXC test is stupid...
use T-Racks 3

Is the T-Racks better optimised for AMD's Multi Dualcore/shared FP architecture that trips over itself in any Real World application than ReaFX ReaXComp, WaveArts MD5, URS Channel Strip Pro, Elysia mPressor , Softube TLA-100A, Softube Classic Channel ( some Waves and Slate Digital coming soon ) ?

Does it somehow circumvent the Windows Task Scheduler and use some magic algo that would give precedent to AMD's cock eyed vision of SMP ?

Got any numbers ?

Also, lets not forget the VI test using Kontakt 4.

[GoldMember]

Originally Posted by Bassmankr

Goldmember, that linked Passmark CPU Overclock database results seems unrealistic in that it doesn't reflect the real world. Many of the top end chips just don't overclock well or even overclock at all if locked or the motherboard won't let them. Additionaly that graph handicaps CPU's that DO OVERCLOCK WELL such as the reasonably priced 3770K and 2700k to just a 3.5 Ghz overclock (both of which I've seen recently priced for only $260), which is laughable as even with autoclocking done by some of the motherboards you get at least 4.3 Ghz with them on air and a cheap decent cooler (under $40). You should just post your rig using Dawbench as that is specific to what we use here and will be an APPLES TO APPLES comparison. All other benchmarks are fairly meaningless for our purposes. One has to remember you can still get over 100 tracks with plugs and low latency with an old single core box that uses the right components and is tweaked right.

to know more about each specific test...
you need to download & install the passmark software
then Download other results to compare side by side to your results.
the Web page is a reference only... more details are available.

overclocking a 100% stable system is tricky.
i can make my i7 920 overclock to 4ghz EASY from 2.67ghz = +50%, loads windows/linux/osx, opens windows, does maths, write letters, browses internet, everything works ok for a few hours, but doing intensive CPU & Memory tasks, is where it fails.
why? memory.
at 3.6ghz is rock solid stable...

there is a very good software callled SystemStabilityTester, it's an improoved version of HyperPi for Windows that is also an improved version of SuperPi.
http://systester.sourceforge.net/downloads.php
Hyper PI 0.99b (Super PI Mod front-end) Download - EXTREME Overclocking

most cpus should overclock +35%
only top of the line / extreme version boards&cpus allow to fully overclock.
not all can do it.

temperature also affects overclocking and life ... over 65°C

i get 48°C each core at 100% cpu load watercooled. 2.67ghz.
60°C at 3.6ghz.
70°C at 4ghz WATERCOOLED!

anything over 65°C means shorter life.

[GoldMember]

Originally Posted by TAFKAT

Is the T-Racks better optimised for AMD's Multi Dualcore/shared FP architecture that trips over itself in any Real World application than ReaFX ReaXComp, WaveArts MD5, URS Channel Strip Pro, Elysia mPressor , Softube TLA-100A, Softube Classic Channel ( some Waves and Slate Digital coming soon ) ?

Does it somehow circumvent the Windows Task Scheduler and use some magic algo that would give precedent to AMD's cock eyed vision of SMP ?

Got any numbers ?

Also, lets not forget the VI test using Kontakt 4.

Totally the opposite... CPU intensive, GPU intensive.
in cubase there is "Cubase/Nuendo CPU Meter"
but T-Racks is so intensive that breaks the 1:1 ratio of the real CPU meter in Task Manager vs. Cubase CPU meter.
to something like 100% Cubase but 20% Intel W7 x64 Task Manager.
That's why is better.
it's impossibe today to load 300x T-Racks 3 with 8x 670 compressors each "with output gain", passing white noise audio without clicks or pops.
at 96khz even harder.

MDA test tone generator---->T-Racks 3.x--->T-Racks 3.x--->T-Racks 3.x--->T-Racks 3.x--->T-Racks 3.x--->T-Racks 3.x--->audio without clicks or pops.

unless you have a 64-core AMD with Lynx AES16 or RME HDSPe AES sound card.



download the demo... you tell me the numbers.
T-RackS 3 Deluxe - Modular High-End Mastering/Mixing Suite of Dynamics and EQ processors

[GoldMember]

Originally Posted by jcschild

absolutely hillarious somehow a test FOR AUDIO is stupid but a synthetic test for cpu is ok...

yeah what ever you say buddy.. they dont come anywhere close. to intel..
anything you say to justify your sad pruchase..

Scott
ADK

you say intel is better, but when i purchased intel it's a sad purchase?
learn to think jajajajajaja

Intel horses are Faster, but...
AMD has more horses, LOL that can pull more tasks at the same time.

[Syncamorea]

Originally Posted by GoldMember

Intel horses are Faster, but...
AMD has more horses, LOL that can pull more tasks at the same time.

I wouldn't go bettin' the farm on that, pardner...

[TAFKAT]

Originally Posted by GoldMember

Totally the opposite... CPU intensive, GPU intensive.

What ?

in cubase there is "Cubase/Nuendo CPU Meter"
but T-Racks is so intensive that breaks the 1:1 ratio of the real CPU meter in Task Manager vs. Cubase CPU meter.
to something like 100% Cubase but 20% Intel W7 x64 Task Manager.

The Cubase Meter is not a CPU Meter its an ASIO Meter, there are numerous variables involved including the respective buffer size and driver efficiency. You have no idea of the mechanics involved by the sound of it.

A 64 Core AMD eh.., I take it you have one or have access to one ?

download the demo... you tell me the numbers.

Why would I bother , you are the one making all of the claims , the onus of proof is on you , not me.

Lets see these all conquering Bulldozer numbers.. !!!

[FLYINGJAY]

[QUOTE=jcschild;8184530]

Originally Posted by GoldMember

#1. RXC test is stupid...
use T-Racks 3

#2. AMD Opteron 6272 is the flag ship of AMD.

#3. cpubenchmark.net rules.
AMD is the fastest SMP cpu in the world.
PassMark Software - CPU Benchmarks - Multiple CPU Systems

absolutely hillarious somehow a test FOR AUDIO is stupid but a synthetic test for cpu is ok...

yeah what ever you say buddy.. they dont come anywhere close. to intel..
anything you say to justify your sad pruchase..

Scott
ADK

Agreed!

Sent from my PC36100 using Tapatalk

[dubrichie]

GoldMuppet clearly has no idea what the feck he is talking about.

Let's not feed the troll, eh?

[ivansc]

(grin) Considering ALL we are "discussing" here (if that is the right word) is a couple of computer cpus, you guys are awful mean....

Just saying.

[jcschild]

ummm no not really.. when you start a post with this

#1. RXC test is stupid...

[TAFKAT]

FWIW :

A few years back on the Nuendo Forum we had a member who posted a thread about a dual AMD Magny-Cour system ( 2 x 12 Physical cores : 24 Cores ) and how it was a game changer, easily demolishing the then current Dual Xeons and the king of the hill single Intel of the time - 980X . The O.P has an aversion to DAWbench and preferred to use his own in house benching, which is all cool, however they are not in the public domain and needed to be duplicated manually by anyone who wanted to qualify the results.

In short, claims of 200-400% improvement over Dual Xeon 5600 Hex systems, and totally destroyed the 980X by around 700% by my calcs.

One of my clients with a single Xeon X3520 ( i7 920 ) Quad @ 2.66 duplicated the test and not only beat the posted 980X result by 300 % but managed around 50 % what the 24 Core AMD system , so 1/6 the cores , 1/2 the performance using the O.P's own benchmark, No DAWbench in site.

BTW: Thread Title was changed later... LOL

Does this directly correlate to the Bulldozers, well not entirely, but its a good indication that throwing more AMD cores at a DAW will not necessarily result in quantum improvement as GoldMember is trying to assert.

There is also a few reports at AnandTech of the Magny-Cours v Xeon 5600 and Bulldozer 6276 v Xeon E5 using more conventional non audio benchmarks which GM seems to like to point to. In both cases the Xeon with 1/2 the cores fed its arse back to the AMD chip. Second report also shows M-C v B-D.

[jcschild]

bwahahaha i remember that... are you sure that wasnt in this forum?

Bless his heart.. it was a good attempt at advertising if nothing else.. i wonder if he ever sold any.. he is on my facebook may have to ask..

emailing you actually..

[TAFKAT]

Originally Posted by jcschild

bwahahaha i remember that... are you sure that wasnt in this forum?.

Its on the old Archived Nuendo Forum, link is in the post... :-)

[GoldMember]

Originally Posted by TAFKAT

The Cubase Meter is not a CPU Meter its an ASIO Meter, there are numerous variables involved including the respective buffer size and driver efficiency. You have no idea of the mechanics involved by the sound of it.

URS Channel Strip Pro
Windows Task Manager CPU Meter vs. Cubase CPU/ASIO Meter has 1:1 ratio,.
AMD Opteron 8xxx vs. Intel Xeon E54xx plugin tests.

[TAFKAT]

Originally Posted by GoldMember

URS Channel Strip Pro
Windows Task Manager CPU Meter vs. Cubase CPU/ASIO Meter has 1:1 ratio,.
AMD Opteron 8xxx vs. Intel Xeon E54xx plugin tests.

Whats your point ?

Thats Protools , not Cubase and there is no indication of the system CPU loadings in the sreenshot !

There are instances when the ASIO/Performance meter can mimic the TM when using efficient ASIO/Core Audio drivers at numerous latencies , depending on hardware, but throw in a linear phase plugin or even a plugin with high inherent latency, and that balance can and will shift very quickly.

Also scroll down a bit further on that linked thread and see a sceenshot I posted 4 years ago showing Sonar coughing up a lung when one of its internal threads maxed , which did not correlate on TM.

The various DAW's use different metrics to indicate what their respective performance meters are showing , thats why gauging scaling performance by meters is near useless most of the time.

Seriously Mate, give it a rest..