AMD puts on the ritz with six-core Opteron demo

[Sharktooth]

With Intel's Nehalem-based Xeons gathering like a storm on the horizon, AMD today gave the first working demonstration of its potential counterpunch: a six-core Opteron processor code-named "Istanbul." Istanbul is a fairly straightforward upgrade over current 'Shanghai' Opterons: a 45nm processor with 6MB of L3 cache that fits into a Socket F-style motherboards, only with six cores rather than four. As a result, the upcoming Istanbul-based Opterons will serve as drop-in upgrades for existing Socket F systems. The chips will take advantage of the same 2P, 4P, and 8P infrastructure as today's Opterons, with HyperTransport and two channels of DDR2 memory per socket.
Read more... (http://www.techreport.com/articles.x/16448)

[Blue_MiSfit]

Interesting stuff.

I wonder how 12 cores performs in an x264 benchmark? Versus an 8 core (16 logical) nehalem server?

~Misfit

[Shinigami-Sama]

this is what they should've done when C2Qs were released

once again AMD is a year behind the curve :(

[Sharktooth]

yep. that year behind was/is the real price to pay for the ATi acquisition.

[Shinigami-Sama]

yep. that year behind was/is the real price to pay for the ATi acquisition.

that and horrible driver support
but that's finally catching up...

[swaaye]

ATI acquisition wouldn't have had any effect on CPUs. CPUs are in development for ~5 years usually. ATI acquisition was an attempt at diversification, perhaps because they knew the CPU front was going to get ugly soon. AMD wants to turn into Intel and have their whole platform covered like Intel does (chipsets, IGPs, CPUs). They also must have seen this slowly evolving GPGPU future as well.


but that's finally catching up...
Oh get over the ATI drivers-are-bad nonsense already. NV and ATI have been pretty much at parity for years now. Probably since like Radeon 9800. The monthly Catalyst program is on its 9th year!

Let's take a moment to remember NV's little cheating years in there with GeForce FX with them clawing for every extra bit of framerate by forcing lower precision and hiding visual details. The last couple of years have been rather ugly for them too. They have been struggling with BSODs pretty bad. I think it's partly down to semi-defective cards that, when pushed, go unstable. My bro is on this 3rd 8800GT. My 8800GTX has been a champ though.

[Sharktooth]

ATI acquisition wouldn't have had any effect on CPUs. CPUs are in development for ~5 years usually. ATI acquisition was an attempt at diversification, perhaps because they knew the CPU front was going to get ugly soon. AMD wants to turn into Intel and have their whole platform covered like Intel does (chipsets, IGPs, CPUs). They also must have seen this slowly evolving GPGPU future as well.
it had since AMD invested a lot of money to acquire ATI and they had to slow everything down and paradoxally push the phenom out when it was not competitive at all.
afterwards other factors came into play and the situation is getting even worse they had to "sell" their fabs and create the Foundry Company.
however, the world economic situation is not helping at all...

[deekey777]

x264 HD video encoding (http://techreport.com/articles.x/17005/11)
Is it fast or slow?
These scores come from the newer, faster version 0.59.819 of the x264 executable.

Pretty old?

[MfA]

It's all slow ...

Hardware encoding is worse than useless for both NVIDIA and ATI, which one is least useless is pretty academic. Maybe if we get GPU like parts which can also do 8 and 16 bit ops (faster than single precision floating point) and which can handle fine grained branching they will start to outperform CPUs, but not for the moment.

[Sharktooth]

@MfA, the link deekey777 posted is referring to the new 6 cores Opteron.

@deekey777, it's slow on the first pass (dunno why...) and pretty fast on the second pass. however, no words on settings (expecially on the number of threads used), so i would not consider that test at all.

[deekey777]

@MfA, the link deekey777 posted is referring to the new 6 cores Opteron.

@deekey777, it's slow on the first pass (dunno why...) and pretty fast on the second pass. however, no words on settings (expecially on the number of threads used), so i would not consider that test at all.

http://www.techarp.com/showarticle.aspx?artno=520
--level 4.1 --keyint 24 --min-keyint 1 --ref 3 --mixed-refs --bframes 3 --b-rdo --bime --weightb --direct auto --subme 6 --analyse p8x8,b8x8,i4x4,i8x8 --8x8dct --ipratio 1.1 --pbratio 1.1 --vbv-bufsize 30000 --vbv-maxrate 38000 --qcomp 0.5 --me umh --threads auto --thread-input --sar 1:1 --progress --no-psnr --no-ssim --output "../run1-720p.mp4" "test-720p.avs" --mvrange 511 --aud --nal-hrd
-----
threads auto - Sets the number of threads that x264 will use to 1.5 times your number of processor cores. Therefore, a dual core chip would use 3 threads; a quad core chip would use 6 threads, etc.

[Sharktooth]

didnt realize they used that test. then there's definatly something weird going on.
probably a newer x264 build will behave in a different way though.

[Manao]

The first pass doesn't scale well at all, because of the fast encoding settings. So a quad core 3 GHz will perform almost twice as fast as a octo core 1.5 GHz on the first pass, but the speed will be the same on the second pass.

As for the build, it is a bit old and predates phenom optimizations in x264.

[simonhowson]

As for the build, it is a bit old and predates phenom optimizations in x264.
Don't those optimisations help Penryn and i7 CPUs too?

[Manao]

Well, in the 300 or so revisions since that build, there have been several optimizations targetted at all kind of CPU. Some of these optimizations were specific to Phenom (those made for SSE4a), some were specific to penryn and i7 (SSE4), some helped all CPUs, but not necessarily to the same extent. In the end, you can't really extrapolate how much each CPU has been speeded up since that time, so the only relevant comparison would have to be made with a much more recent build.

[deekey777]

An user of the german 3DCenter Forums has modified the x264 Benchmark 2.0, so it uses x264 v0.59.1101 (I hope, everything is ok).
http://www.forum-3dcenter.org/vbulletin/showthread.php?t=446675

[deekey777]

A Glance into the Future: Six-Core AMD Istanbul in a Desktop Platform (http://www.xbitlabs.com/articles/cpu/display/amd-istanbul.html)

x264 numbers: http://www.xbitlabs.com/articles/cpu/display/amd-istanbul_8.html#sect0
Is it good or bad?

[saint-francis]

Doesn't compete with an i7 920? I'll pass. When we see mainstream 6 core 32 nm CPU's from Intel I'll bite.

[IgorC]

I think that benchmark doesn't reflect reality.

x264's numbers of threads:
Core i7: 8 threads*3/2=12 threads.
Istanbul: 6threads*3/2=9 threads.

2 microprocessors should be tested at the same number of virtual threads (i.e. 12).

[prOnorama]

Doesn't compete with an i7 920? I'll pass. When we see mainstream 6 core 32 nm CPU's from Intel I'll bite.

It does compete, sort of (all virtually of course), but hey it's good to keep in mind Intel's profits are higher than AMD's turnover so unfortunately there's not a very healthy and equal competition (think of research budgets) in the CPU market like there's now in the GPU market between ATI (AMD) and nVidia.

Of course Intel could obliterate AMD but they won't because of trust laws (like they might haver to split up the company if they'd obtain monopoly of the consumer CPU market- mind you they are so big and strong they are very close to that monopoly already)