Dual Duron Mobo?

[xdshrx]

Well I'm working on building a computer right now. In two weeks or so Im ordering the Chieftec Aluminum SX1040. Ive been having some trouble finding a mobo. Can someone point me to a good Dual Duron Mobo that supports 1.1Ghz or higher and DDR Ram? Im going cheap on this setup so $200 is prob the max im willing to dish out.

 

[Peter]

Dual Durons is not supported, and would perform rather poorly anyway since only the Athlon MP processors have the extended cache coherency protocol (MOESI) - this is what makes them efficient in dual processing environment.

 

[xdshrx]

I'm a newbie with motherboards so sorry for the dumb questions. I've been trying to do as much research as possible.

Guess I'll be looking into another mobo. Any for suggestions for an AMD based mobo $200 r less?

 

[ScrapSilicon]

Originally posted by: xdshrx
I'm a newbie with motherboards so sorry for the dumb questions. I've been trying to do as much research as possible.

Guess I'll be looking into another mobo. Any for suggestions for an AMD based mobo $200 r less?

Epox 8k5a2+ for about $115(cnetpc.com) or so and an Athlon XP of some flavor(XP1600+ Aroia or Agoia for about $55@ newegg.com) with a Volcano 9 ($25 or so..) is a good $200 operation..

 

[Vette73]

Originally posted by: Peter
Dual Durons is not supported, and would perform rather poorly anyway since only the Athlon MP processors have the extended cache coherency protocol (MOESI) - this is what makes them efficient in dual processing environment.

Ahhh, you sir are WRONG

The only Diff. between a Duron and a Athlon MP is the MP has been tested to do SMP, has more L2 cache, and runs at 133Mhz. There is no other difference.
How do I know, I had 2 1Ghz durons in my 760MPX board running 100% stable, and I now have 2 Athlon XP's. So yes Durons will do SMP quite well

So yes the 760MPX boards will run dual durons as long at the last bridge of the L5 is connected.

If you want a dual Athlon/Duron setupjust ask me what you need to know and I will try to help.

 

[mechBgon]

xdshrx, also keep in mind that in many situations, a single 2000+ would outperform a dual 1600+, as an arbitrary example. Part of this is due to the fact that today's highest-performance AMD chipsets, from a CPU/RAM standpoint, are single-processor ones, and part of it is due to the fact that the second CPU is significantly beneficial mainly when running a multithreaded app or when there's an ongoing intensive process running in the background but you still need to run some other program in the foreground. Sorry if that's a bit muddled, I have to get on my way to work. Bottom line, if you want good gaming performance, get a fast single-CPU system.

 

[Peter]

Yes they do run SMP (since each CPU has its own bus here, any Athlon or Duron does), but according to AMD only the MP/XP core has the enhanced cache ownership protocol. Without it, there is much more cache coherency traffic, slowing the system down. That's why they added it in the first place.

 

[Vette73]

Originally posted by: Peter
Yes they do run SMP (since each CPU has its own bus here, any Athlon or Duron does), but according to AMD only the MP/XP core has the enhanced cache ownership protocol. Without it, there is much more cache coherency traffic, slowing the system down. That's why they added it in the first place.

Again the only diff. is the AMOUNT of cache, other than that it is the same core. This has been tested by several sites, it is just a marketing thing to get people to buy the more $$$ chips

And my Dual Durons were pretty fast and you can get 2 1 to 1.3Ghz Durons for around $80

Good start if you are on a budget. Go cheap on the chips at first if that helps get a better board that can be upgraded.

 

[Kazuo]

Wait, you can do a dual AthlonXP? I heard that SMP was somehow disabled! Holy crap! Forgive me for saying something so stupid, but I gots to get a piece of that! Video editing would fly!

 

[MithShrike]

Yes, it is quite beautiful to do video on a dual Athlon XP setup. My friend... well he's got that IWill board with two 2200+s on it and with the Radeon 9700... it's just plain beautiful to do videos and other such things .

 

[bozo1]

Wait, you can do a dual AthlonXP? I heard that SMP was somehow disabled!

SMP was disabled in XP chips manufactuered after March by an L5 bridge cut. You need to reconnect the bridge to get it to work.

 

[Kazuo]

Wow. I read that you don't even need to fill in the hole between the sides of the bridge- painting over it works.
Pretty neat. I may have to do this, when I am once again employed.

 

[xdshrx]

So basically i should just get a AMD Based mobo/cpu. I can get 2 1.1ghz chips for $79shipped w/out the motherboard. I working on a buget I just want whats best its hard making a decesion being a newbie!

 

[mechBgon]

What do you want to use the system for? The two situations where dual is good:

• When you have an ongoing task running in the background that would make a single-CPU machine sluggish. Real-world example: rendering in Caligari trueSpace, which drags the system to a crawl until the rendering is complete, more often than not. On a dualie, if the rendering is hogging one CPU, the other one's still available for you to do image editing, browse the Internet, etc.
• When you have a program that is designed to take advantage of two CPUs at the same time, in which case you can expect performance boosts ranging from 10% (Quake3 Arena, the only multithreaded game I know of for sure) to 65% (Caligari trueSpace 4.3 raycasting rendering) to ~100% (Distributed.net OGR-25 client).

Beyond these scenarios, a dual 1.1GHz Duron system is going to perform essentially like a single 1.1GHz Duron system. If you just want a typical gaming system, get a faster single CPU on a faster single-CPU motherboard, that's my opinion.

 

[Ziptar]

Originally posted by: Peter
Dual Durons is not supported, and would perform rather poorly anyway since only the Athlon MP processors have the extended cache coherency protocol (MOESI) - this is what makes them efficient in dual processing environment.

Ughhh.... I wish people would read a little before they post..

read this Firing Squad Dual Duron Performance Review

Sure, a dual Athlon MP system offers faster performance than a dual Duron MP system, but how much are you willing to spend for 1% to 8% improvement at the same clock speed? Double the price? Personally, we'd rather save our money for something else.

and this 2CPU.com Tyan Tiger MP (S2460) Review

you can't go wrong with the Tiger MP and a pair of AthlonMPs (or even 1ghz Durons).

and this xbit labs Dual Duron at Home

as we found out, in most applications there is hardly any difference between dual-Athlon and dual-Duron systems

I have run Dual Durons for over a year.... I just Upgraded to a pair of modded XP 2000+s. Go here to learn about the XP --> MP Mod

A Duron will run MP Any Duron 1Ghz and up is a Morgan core, same core (and instruction set!) as the Athlon XP / MP (palamino) just less cache. They will run SMP WITHOUT ANY MODIFICATION

The cheapest way to go AMD Dually is with a pair of Durons and a Tyan Tiger MP S2460 Motherboard pricegrabber from $159.00 or from $152 on pricewatch also keep in mind that it requiers registered ECC DDR memory, (although, some have reported successfully being able to use up to 2 sticks unbuffered, non-ECC) some of the newer 760MPX boards don't but they cost more. The Newer 760MPX chipset has 64bit/66mhz PCI bus and some have USB 2.0 (iwill, i think.). If you are new to AMD SMP then check out the 2cpu.com 760MP/X FAQ., also lurk the motherboard forum over at 2cpu.com, lots of AMD Dually info

On the Tiger MP S2460 If you set j52 to 1-2 when the rest are set for 100 Mhz fsb it will give you 115FSB. Just about any Duron will do that and that gives you a 230Mhz FSB.

As to Dual or No to Dual read This.

And just as an example:

I don't know what point this makes if any and how scientific it is. Since I was upgrading to dual XP 2000's anyway I figured I would try an experiment.

I ran UT2003 Demo Benchmarks on a single XP then Dual, Then Single XP w/ a Seti Command Line Client running and CPU Priority set to High, Then Dual XP w/ a Seti Command Line Client running and CPU Priority set to High. Here are the Benchies:

Dual Duron 1Ghz Running @115Mhz FSB = 1.15Ghz
dm-antalus
9.521903 / 20.402187 / 38.640507 fps
Score = 20.411419

Single XP 2000+
dm-antalus
6.744068 / 29.279009 / 61.872112 fps
Score = 29.298101

Dual XP 2000+
dm-antalus
9.435257 / 29.955162 / 64.283737 fps
Score = 29.976755

Single XP 2000+ w/ 1 Seti Command Line Client running w/CPU Priority set to High (100% CPU Utilization)

I wish I had some numbers but the Benchmark didn't even start after 10 minutes... I got impaitent....

Dual XP 2000+ w/ 1 Seti Command Line Client running w/CPU Priority set to High (100% CPU Utilization)
dm-antalus
9.341468 / 29.277737 / 64.079590 fps
Score = 29.299311

I don't know if this proves anything in the real world other than obvious creamy SMP goodness.

I did run all of the benchmarks for single and dual XP (w/o seti) and I found my scores higher if only slightly when running duals. I didn't experience the Dual CPU performance Hit I often see mentioned.

Rig:
Geforce2 GTS-V running at 215/330
Dual XP 2000's
1GB Crucial PC2100 ECC REG DDR (4x 256MB)
Tyan Tiger MP S2460
WinXP Pro

 

[Peter]

To clarify things. The enhanced cache ownership protocol increases performance quite a bit when both processors are working on the same set of data (as in databases, threaded rendering or image processing, these things), while it practically does nothing in favor of completely separate tasks running on either processor. So depending on what you benchmark, the perceived difference is anywhere from zero to dramatic.

Athlon MP silicon is no different from Athlon XP - just that one configuration bridge lasered to enable or disable SMP support, and the SMP factory test passed on MPs or failed/untested on XPs.

regards, Peter

 

[Ziptar]

Originally posted by: Peter
To clarify things. The enhanced cache ownership protocol increases performance quite a bit when both processors are working on the same set of data (as in databases, threaded rendering or image processing, these things), while it practically does nothing in favor of completely separate tasks running on either processor. So depending on what you benchmark, the perceived difference is anywhere from zero to dramatic.

Athlon MP silicon is no different from Athlon XP - just that one configuration bridge lasered to enable or disable SMP support, and the SMP factory test passed on MPs or failed/untested on XPs.

regards, Peter

Agreed, True SMP benefits only come into play in certain situations. If you read the reviews I posted above you will see that in true SMP app benchmarks the Duron/MP differences are minor on a per clock basis. Your statement in your first post is incorrect. Morgan core Durons do possess the MOESI instruction set, as does the XP. MOESI is not exclusive to the MP. Durons 1Ghz and up are no different than the MP/XP Palamino cores other than in respect to cache size and rated FSB.

 

[mechBgon]

UT2003 Build UT2003_Build_[2002-09-13_17.31]
Windows 98 4.10 (Build: 67766446)
AuthenticAMD PentiumPro-class processor @ 1750 MHz with 255MB RAM
NVIDIA GeForce2 GTS/GeForce2 Pro (2832)

dm-antalus?spectatoronly=true?numbots=12?quickstart=true -benchmark -seconds=77 -exec=..\Benchmark\Stuff\botmatchexec.txt

12.728155 / 73.497978 / 128.902283 fps rand[13909]
Score = 73.539993

That's what I got with a KT333-based board and the same exact video card, the GF2GTS-V (not overclocked), using Win98SE. The CPU here is a 1600+ at 166/166. I don't know why the large disparity in score versus the Tyan dualie... I was expecting KT333 to show a little more muscle per MHz, but not 3 times as much. Must be something different about how it's being run on my system (resolution, detail or ???).

 

[Ziptar]

Originally posted by: mechBgon

UT2003 Build UT2003_Build_[2002-09-13_17.31]
Windows 98 4.10 (Build: 67766446)
AuthenticAMD PentiumPro-class processor @ 1750 MHz with 255MB RAM
NVIDIA GeForce2 GTS/GeForce2 Pro (2832)

dm-antalus?spectatoronly=true?numbots=12?quickstart=true -benchmark -seconds=77 -exec=..\Benchmark\Stuff\botmatchexec.txt

12.728155 / 73.497978 / 128.902283 fps rand[13909]
Score = 73.539993

That's what I got with a KT333-based board and the same exact video card, the GF2GTS-V (not overclocked), using Win98SE. The CPU here is a 1600+ at 166/166. I don't know why the large disparity in score versus the Tyan dualie... I was expecting KT333 to show a little more muscle per MHz, but not 3 times as much. Must be something different about how it's being run on my system (resolution, detail or ???).

Possibly, That is interesting though.. I thought my scores were good I am running XP Pro, The res could be the problem what res you at?? I have mine set to 1024 x 768 x 32 w/ openGL on and using detonator 40.41 drivers. LMK. Are you running your XP at stock speed?

 

[mechBgon]

Does the batch-file launcher adopt whatever settings the user ordinarily uses in the game? That would definitely explain things. I'll go try it again after setting it for your settings. Do you set all your detail levels to the highest, with decals stay and stuff like that?

 

[mechBgon]

Well, using DirectX (since I don't know how to switch to OpenGL), with all the visual tweaks turned on and maxed out, at 1024 x 768 x 32, it comes out like this:

UT2003 Build UT2003_Build_[2002-09-13_17.31]
Windows 98 4.10 (Build: 67766446)
AuthenticAMD PentiumPro-class processor @ 1750 MHz with 255MB RAM
NVIDIA GeForce2 GTS/GeForce2 Pro (2832)

dm-antalus?spectatoronly=true?numbots=12?quickstart=true -benchmark -seconds=77 -exec=..\Benchmark\Stuff\botmatchexec.txt

4.437691 / 22.014990 / 53.157375 fps rand[12845]
Score = 22.019033

I guess I'll add that I think the video card is holding up the CPU. Heck, the video card holds up the CPU at 640 x 480, wish I could afford a video card upgrade! Also, this botmatch was completely different than the last... are they randomly generated, is that the significance of the rand[12845] line?

 

[Ziptar]

Yeah, Settings all the way up with decals. Those numbers look a little more in line. You had me worried..
To turn to OpenGL, open this file :

UT2003Demo>System>UT2003.ini

In the third paragraph, you have :

[Engine.Engine]
RenderDevice=D3DDrv.D3DRenderDevice
;RenderDevice=Engine.NullRenderDevice
;RenderDevice=OpenGLDrv.OpenGLRenderDevice

You have to replace the ";" to get this :

[Engine.Engine]
;RenderDevice=D3DDrv.D3DRenderDevice
;RenderDevice=Engine.NullRenderDevice
RenderDevice=OpenGLDrv.OpenGLRenderDevice

and you're done

 

[mechBgon]

Cool, let me give that a whirl and I'll post the result for you

 

[mechBgon]

Originally posted by: mechBgon
Cool, let me give that a whirl and I'll post the result for you

...or not. The screen is filled with randomly-flying textured junk which obscures most of the scene completely.

 

[Peter]

Agreed, True SMP benefits only come into play in certain situations. If you read the reviews I posted above you will see that in true SMP app benchmarks the Duron/MP differences are minor on a per clock basis. Your statement in your first post is incorrect. Morgan core Durons do possess the MOESI instruction set, as does the XP. MOESI is not exclusive to the MP. Durons 1Ghz and up are no different than the MP/XP Palamino cores other than in respect to cache size and rated FSB.

Oh, by "not supported" I meant not supported by many of the popular mainboards. That's mainly because AMD doesn't want you to. MOESI is absent in the original Duron and Thunderbird Athlon, unconfirmed in "Morgan" Duron and officially there in Athlon MP and the identical silicon XP.

Smaller L2 caches do not have as much of an impact on SMP performance as they do on Intel platforms, since the CPUs do not share a single bus. So you just get the normal performance delta you'd expect from not having as much cache, not worsened by any congestion side effects like on Intel platforms.