AMD's dual-core performance boost
Posted by malebolgia on 06 October 2004 - 01:34 · 30 comments & 2488 views
About the Author
Full name: Unknown
Forum name:
malebolgia
Contact: via forum PM
Submissions: Normal · Headline
Full name: Unknown
Forum name:
malebolgiaContact: via forum PM
Submissions: Normal · Headline
Slashdot It!
Submit to reddit
Submit to blinklist
Bookmark on del.icio.us
Add to furl
Share on Facebook
Add to Windows Live
Add to Google- Advertisement
-
-
#1 Posted by crackhbob on 06 Oct 2004 - 01:36
- Can't wait til we have dual, dual core cpu motherboards floatin around
-
#2 Posted by sonob on 06 Oct 2004 - 03:14
- This will be sweet I hope the make desktop versions!!
-
(5 replies)
#3 Posted by goatsniffer on 06 Oct 2004 - 03:25
- Sounds like glorified Intel Hyper-Threading
-
#3.1 Posted by ev0| on 06 Oct 2004 - 03:48
- Glorified ? It's real dual CPU dude, Not some half ass fake **** like HT. Don't worry, Intel too has seen the light and they're making dual core too. Maybe you should learn a bit more before posting around here.
-
#3.3 Posted by SVT on 06 Oct 2004 - 12:28
- goatsniffer, that was one of the most ignorant posts I've ever seen on Neowin.
-
#3.4 Posted by GamblerFEXonlin on 06 Oct 2004 - 14:01
- You're all idiots!
A HyperThreading enabled CPU is more pleasant to use and can sometimes get a 50% or more ? performance boost with a small amount of extra transistors.
A dual-core CPU could have HT too, HT just means being able to use two features of the same CPU at the same time. Using more transistors at a time that is.
If all you guys are workstation users with industry programs I can see why you are so fascinated, if you are more into games you can pipe down now. Games benefit more from a faster CPU or specialized hardware like 3D accellerators and SIMD extentions (e.g SSE2).
http://www.hardocp.com/article.html?art=NjAyLDY=
The one thing worth mentioning here that does come into doing your daily work with this CPU is multitasking and the lack of support for a HyperThreading-like implementation. When working with the AthlonFX under heavy loads with more than one program doing actual work at the same time, I do not get the same fluidity that I do with my Pentium 4 HyperThreading system. Now this of course is totally subjective, and not everyone is going to be encoding DivX movies while they work as an example. But if you are a bit beyond the normal desktop user, the lack of HyperThreading is something to keep in mind. -
#3.5 Posted by Coolme on 06 Oct 2004 - 14:52
QUOTE You're all idiots!
A HyperThreading enabled CPU is more pleasant to use and can sometimes get a 50% or more ? performance boost with a small amount of extra transistors.
A dual-core CPU could have HT too, HT just means being able to use two features of the same CPU at the same time. Using more transistors at a time that is.
If all you guys are workstation users with industry programs I can see why you are so fascinated, if you are more into games you can pipe down now. Games benefit more from a faster CPU or specialized hardware like 3D accellerators and SIMD extentions (e.g SSE2).
First off....I don't think anyone here is doubting that using HT on dual core processors would improve performance on SOME APPLICATIONS (Other applications may actually load SLOWER)
The topic here is that goatsniffer's comment "Sounds like glorified Intel Hyper-Threading". Which I assue that he's implying that have dual-core on one die (Two processors on one chip) is like "glorified Hyper-Threading", in which is not.... because of it's overall performance on applications is up to 50% faster then just one one-core processor and does not have any decrease in any application. Unlike HT, which brings about a maximum of 40-50% increase and sometimes DECREASE in performance.
-
#4 Posted by bush on 06 Oct 2004 - 04:10
- kewl. me likes.
-
(1 reply)
#5 Posted by Dark2020 on 06 Oct 2004 - 04:13
- dual core coming to the fx processors or not
-
#5.1 Posted by slapnuts_ox on 06 Oct 2004 - 10:43
- yes its coming to both socket 939 and socket 940.
-
(1 reply)
#6 Posted by hardgiant on 06 Oct 2004 - 05:18
- AWESOME
AWESOME
AWESOME
AWESOME
AWESOME
AWESOME
AWESOME
AWESOME
AWESOME
AWESOME
-
#7 Posted by Lowdown on 06 Oct 2004 - 07:33
- Wow, 30 to 55 percent. That's great. For awhile going to the next model processor in speed wasn't worth the upgrade. With both Intel and AMD looking at increasing processing power that much it will defenetly be worth upgrading. Technology has been increasing at leaps and bounds but there has been times when the differences between processor models wasn't that much.
-
(1 reply)
#8 Posted by BetaguyGZT on 06 Oct 2004 - 09:07
- Lets see....current Opteron runs at 3.2
New Opteron runs at (if we go the +55%) somewhere between 4.8 to 5.1. OC that bad boy to 6.0 easily if you have some good cooling (+20% OC).
*we wants teh precioussss!!!* -
#8.1 Posted by slapnuts_ox on 06 Oct 2004 - 10:46
- wrong...current opterons run at 2.4Ghz not 3.2Ghz. The fastest Athlon 64 is 2.4Ghz with 1MB l2 cache and 1000Mhz Hypertransport for a 4000+. Don't confuse the performance rating with clock speed. Also if you read the article then you'd notice they said a 600Mhz - 1000Mhz speed drop so since the FX is 2.6Ghz then we can assume that the slowest dual core chips offered will be 1.6Ghz and the fastest ones will be 2Ghz for the first initial release. Im sure they will eventually increase the speeds to 2.6Ghz per core as yeilds increase and maybe another die shrink.
-
(1 reply)
#9 Posted by ynohtna on 06 Oct 2004 - 14:26
- I can't wait for dual core Intels with HT enabled. 2 physical + 2HT ones = 4! um wait, how's windows licensing work again?
-
(1 reply)
#10 Posted by Coolme on 06 Oct 2004 - 14:58
- Good work AMD!! Although 50% performance increase verses one-core processors is not that much (Considering that there is actually 2 processors there) I am very impressed that AMD got the dual-core chip to fit in the same slot, and have a same amount of power being percipitated (compared to single-core chips). I hope that AMD will release a dual-core chip version that has more efficientcy (from 50%) even if it means creating a new socket for it.
-
#11 Posted by kronik on 06 Oct 2004 - 16:08
- excuse my n00bness but is this any better than the new intel Extreme Edition CPUs with 1000+FSB, i.e. is this a better performance enhancement (for games in particular) than increased FSB?
-
#12 Posted by Jstphish on 06 Oct 2004 - 16:30
- Sounds to me like they (AMD and Intel) can't make a faster single processor so they are resorting to dual core. I don't think there is anything bad with that but the marketing makes it sound like this is such a great idea; it may be, but be realistic and don't make everyone out for idiots. Just to let folks know, the only reason they are going to dual core is because the they are having a very hard time making a faster single processor.
-
(2 replies)
#13 Posted by Ifoow on 06 Oct 2004 - 16:57
- Would be cool if they just can't get any faster with processor speed. But ofcourse that's not going to happen because al chip manufactururers have a secret agreement to slowly boost processor speed in order to gain maximum profits. *******s! Same thing with internet they are holding the speed back because they fear the consequences and now they can charge more.
-
#13.1 Posted by noyb on 06 Oct 2004 - 17:24
- Nice theory, and was this deal brokered by the MIB or the Freemasons. The whole industry hit the wall not just the x86 guys but others like IBM are facing the same problems.
-
(3 replies)
#14 Posted by ynohtna on 06 Oct 2004 - 17:33
- What's wrong with the idea of dual processors for better multi-tasking? I think all this 10,15,20% faster means nothing if I still have to wait and do things one at a time. Bring on Multi-core!
-
#14.1 Posted by Kushan on 06 Oct 2004 - 21:02
- Well dual CPU's vs. Dual Core CPU's is all about cost benefits when you think about it.
Dual core would be very similar to a Dual CPU, except the cost would be cheaper, it wouldn't need as big a heatsink/fan (and it would only need one) since it generates the same amount of heat and the performance benefit would be about the same.
As for doing things one at a time, imagine what a Dual Core CPU with HT would be like, it would be like having 4 CPU's in one, how awesome is that? -
#14.2 Posted by Coolme on 06 Oct 2004 - 21:50
- ....Also, not all applications takes advantage of the dual processors, but they do for dual-core processors. And the link between the two processors in the dual processors are a lot slower than the link for dual-core processors.
-
#15 Posted by dhitb on 07 Oct 2004 - 03:49
- SMP-aware, 64-bit software + dual core Hammer = yummy
The future looks better than it did around the release of Prescott.
In addition, AMD said the dual-core Opteron chips will be about the same size and produce the same amount of waste heat as current single-core Opterons, Kevin McGrath, manager of the Opteron architecture, said in a speech at the Fall Processor Forum here. That means the new chips will fit into the existing server designs.
Dual-core chips put two processing engines on the same slice of silicon, a way of letting chips do more work. The change is made possible by moves to more advanced manufacturing processes with smaller features that permit more circuitry to be squeezed into a given area.
What's new ?
Added lost extended registers unlocking code in RAM configuration detection routine for NVIDA GeForceFX and newer boards.
Hour is now displayed correctly on time markers in hardware monitoring log file viewer.
Fixed driver-level color correction for non-default rotation angles under the Detonator 42.10 and newer drivers.
Added ForceWare 66.31, 66.29, 66.32, 66.51, 66.70, 66.72 and 66.81 drivers support.
Updated databases for Detonator and ForceWare driver families. Due to more complex database build automation process used to generate database build scripts in this version, the databases contain many new system registry entries, missed in database build script in the previous versions.
Updated NVStrap antiprotection and ForceWareAntriprotector patch scripts for ForceWare 66.31, 66.29, 66.32, 66.51, 66.70, 66.72 and 66.81 drivers.
Revised antialiasing modes list for the ForceWare drivers.
Changed ForceWareFSAAModes format. Now system requirements are specified using GPU family ID rather than PCI DeviceID.
Updated Catalyst 4.9, 4.9 hotfix and 8.07 beta certified SoftR9x00 and ATIOverclockingAntiprotection patch scripts.
Added RV410 support. Now all R420 specific features including clock frequency control, temperature and fan speed monitoring / control are also available on RADEON X700 series. Thanks to Andrew Worobiew for testing RivaTuner with ATI RADEON X700XT.
Added VID monitoring for NV40 GPU based display adapters. Now GeForce 6800/GT/Ultra owners may also see either raw VID data or the corresponding voltages in hardware monitoring window.
Improved VID interpretator for "Core VID" hardware monitoring data sources:
Now besides NV30/NV35/NV38 specific "ISL6569 with hardwired VID0 / VID1" VID interpretation mode, RivaTuner's database also contains "1.1V + 0.1V / 0.3V loop", "1.1V + 0.3V / 0.2V loop" and "1.1V + 0.1V / 0.2V loop" interpretation modes allowing to translate raw VID data to the corresponding voltage on all NV40 based boards.
Added "Autoselect" button to the "VID interpretation" dialog. This button forces RivaTuner to compare voltage table stored in VGA BIOS with all VID interpretators available in the database and automatically select matched one.
Added interpretation preview window to the "VID interpretation" dialog. Now RivaTuner displays all available VID values and the corresponding voltages for selected intepretator.
Greatly improved built-in patch script engine:
Added BIOSChecksumGenerator RTS post-installation plug-in module, allowing you to restore BIOS checksum after patching a file with RivaTuner's patch script engine.
Added BIOSChecksumGenerator stub patch script, allowing you to restore VGA BIOS checksum in image file without making any additional changes in it.
Improved patch script format. Now the DstFile and BakFile fields can contain %filename% and %ext% macro strings. This feature is useful when source file name contains wildcards and is unknown a priori.
Added NV40BIOSUnitsMaskEliminator patch script, allowing you to remove software pixel / vertex units mask from VGA BIOS image file.
New patch script formats support. Now besides previously available native patch scripts, applying changes to a file, RivaTuner's patch script engine also supports so called runtime scripts, which can be applied directly to a driver loaded in memory. Runtime patch scripts are useful for patching runtime decoded drivers, protected against native patch scripts. Furthermore besides native-only and runtime-only patch scripts, patch script engine also supports so called universal patch scripts, which can be installed in both modes.
Added ATIOverclockingAntiprotectionRuntime patch script, based upon newly introduced runtime patching technology. This script is an example of universal patch script, which can be applied to both driver miniport binary file in the distributive and directly to the miniport driver, loaded in memory.
Improved NVIDIA VGA BIOS formats support:
Now BIT structure based VGA BIOS images with no BMP structure are fully supported by RivaTuner. This allows displaying "NVIDIA VGA BIOS information" diagnostic report category on GeForce 6600 display adapters family, which have no BMP structure in BIOS.
Added BIT tokens list interpretator. Now locations of tokens defining offsets to performance and voltage tables are calculated by translating and scanning whole BIT tokens list and no longer referenced using fixed offset from the beginning of BIT header.
Improved low-level graphics subsystem diagnostic report module:
Now NVIDIA VGA BIOS version is read from the BIT rather than from the BMP when both structures are available in VGA BIOS.
Added dozen of new NVIDA / Intel / SiS / AMD northbridges to RivaTuner's hardware database.
"GPU units mask" line in "NVIDIA VGA BIOS information" diagnostic report category has been renamed to "SW units mask". Software units mask representation has been also changed to make it more plain for beginners. Now RivaTuner displays "none" instead of "16x1,6vp" when there are no software masked units detected in VGA BIOS, and shows which pixel / vertex units exactly are disabled if software units mask exists in VGA BIOS.
Added "HW units mask" line to "NVIDIA specific display adapter information" diagnostic report category. This line displays "none" if the GPU doesn't have hardware masked pixel / vertex units, otherwise it shows which pixel / vertex units exactly are masked with hardware straps.
New NVStrap v1.6 driver brings you more nice exclusive features:
Now the NVStrap driver configuration tab contains "Allow enabling masked units" option, allowing the driver to force the graphics processor to activate all pixel / vertex units, even if they are hardware strapped as damaged. Please be extremely careful and use this option at your own risk. Enabling hardware masked units, which have not passed hardware quality testing, may cause unpredictable results and even permanently damage your graphics hardware.
Improved "Custom graphics processor configuration" dialog layout. Now you may also see enabled/disabled state of each pixel/vertex unit in separate column. New "HW masked" column shows you which pixel / vertex units are hardware strapped as defected and protected from enabling by default by hardware mask to prevent system instability.
Added NVStrap driver version tracking wizard. This module checks version of the currently installed NVStrap driver each time you open NVStrap configuration tab and offers you to replace outdated version with new one if you have not reinstalled the driver manually after upgrading RivaTuner version.
New NVInfo v1.4 bundled DOS utility. Now NVInfo provides RivaTuner-styled command line interface for R/W access to NVIDIA graphics processor's registers. Take a note, that NVInfo correctly supports queued R/W commands like RivaTuner's command line interface.
Due to users' requests, unofficial support for integrated NVIDIA graphics processors is returned back.
Due to numerous requests from third party developers, some components of RivaTuner's SDK are no longer available via personal email request only and will be included in distributive since this version. Now RivaTuner's publicly available SDK includes RTHMSharedMemory VC++ sample, allowing third party tools to access RivaTuner's hardware monitoring statistics in realtime.
Improved compatibility with some slow flashrom chips used on NVIDIA display adapters.
New Easter eggs. Added built-in device driver dump utility, allowing you to dump loaded driver memory and save it to the file for subsequent analysis. This utility is useful for analyzing runtime decrypted drivers as well as for creating runtime patch scripts and tracking differences caused by other tool patching driver in runtime.
Improved installer. Now RivaTunerEx.sys driver is automatically unloaded from memory during installation to ensure proper driver operation.
Added about 10 new overclocking and NV40 softmodding related questions and answers to FAQ.
Minor UI changes and improvements.
RIVATUNER IS SUPPLIED "AS IS". THE AUTHOR ASSUMES NO LIABILITY FOR DAMAGES, DIRECT OR CONSEQUENTIAL, WHICH MAY RESULT FROM THE USE OF RIVATUNER This program should be used at your own risk. We are not responsible for anything that happens to you or your computer after using this utility.