Pentium 820 D

[Rugger]

I really don't know where you guys are getting mobo or ram prices.

Newegg - the mobo price I listed was for an Asus P5WD2.

Yes, I buy a cheapo 820D, which for multitasking and multithreaded applications blows away AMD's best single core chips that costs much more.

That's like me comparing the 820D against an FX55 for gaming. Yes, the 820D beats a top of the line AMD chip in heavy multi-tasking - it's supposed to. The right comparison is vs anothe dual core processor like the X2 3800+.

AMD lemmings buys $1000 X2, next year the M2, oops it uses DDR2, no upgrade for the AMD lemming. New mobo, ram and chip required.

I started this stating that the $1000 chip was not worth the price/performance. M2 being available does not mean that s939 dual core processors wil NOT be available. More than likely, they will be available at higher speeds than right now at lower prices.

A year from now I buy another Intel chip that blows away the 820D for an upgrade.

Who says the new memory of choice won't be XDR or DDR3 and your DDR2 won't be obsolete? There are no guarantees. Given it is almost assured that DDR2 based systems will be available. If nothing else you could put it into an M2 system when your Intel mobo melts.

I can throw barbs just as well as you ,but it gets neither of us anywhere except mad. I've tried to have discussions with you, but it always ends up in rants. Bye.

__________________________________________________
<font color=red>You're a boil on the arse of progress - don't make me squeeze you!</font color=red>

 

[mozzartusm]

You are correct in your thinking about the P5WD2. Its the best by far for the Newer Intel CPUs.

ASUS P5WD2 Premium
Intel 3.73 EE @ 5.6Ghz
XMS2 DDR2 @ 1180Mhz

<A HREF="http://valid.x86-secret.com/records.php?PHPSESSID=792e8f49d5d9b8a4d1ad6f40ca029756" target="_new">#2 CPUZ</A>
SuperPI 25secs

 

[Zoron]

A year from now I buy another Intel chip that blows away the 820D for an upgrade. GEE WIZ, that was hard to figure out. AMD lemmings buys $1000 X2, next year the M2, oops it uses DDR2, oops new socket, no upgrade for the AMD lemming. New mobo, ram and chip required.

Oh really?!?!??

You actually expect the first dual-core boards from Intel to be viable beyond one year? I think you're seriously deluding yourself. You forget that AMD is likely overlooking DDR2 and going straight to DDR3... so what is Intel going to do? I'm sure that DDR2 you're so gung-ho about won't be compatible with DDR3 motherboards. Not to mention possibly a new socket and therefore a new CPU as well.

<font color=red> If you design software that is fool-proof, only a fool will want to use it. </font color=red>

 

[noko]

P4man, are you lost man? Put some Wheaties on brain stem. Let grow into a fully funtional part. Come back later .

 

[Rugger]

SUP MOZZ?! Thanks for the feedback on the board thoughts. What's your take on the 820D and price/performance? Have you gotten your hands on any of the DC's yet to take it for a spin?

__________________________________________________
<font color=red>You're a boil on the arse of progress - don't make me squeeze you!</font color=red>

 

[noko]

Bye. Now I didn't know we had any dicussions. The point is DDR2 memory will have a potential longer useful life for future processors. Will AMD go strainght to DDR3, I don't know, havn't heard that one. I do know that the expected sells for DDR2 next year should be about 68% of the market compaired to 32% for DDR.

So if I upgrade a X2 next year to a faster processor I will still be stuck with DDR ram. That may not be a concern for you or many others but something that does concern me.

 

[Rugger]

I just noticed the 5.6 OC on that EE - SWEET! Did you go sub-zero for that? I assume it's benchmarked and stable? You have screenies on your site?

__________________________________________________
<font color=red>You're a boil on the arse of progress - don't make me squeeze you!</font color=red>

 

[mozzartusm]

Ive played around with an AMD Dual core but not an Intel. At this point I dont see any reason to buy an Intel Dual core. Im sure that they will get it together at some point, buy as of now AMD runs far ahead IMO.

So your just now seeing the 5.6Ghz? I hit that at temps well above 0C. I think maybe it was 12 - 14C. I placed a lot of emphasis on the hot spots around the CPU. For some reason, ASUS see fit to put a heatsink on a couple of them and leaves the others bare. There are at least two places to the left of the socket that will easily go above 85C. I bought some copper tube and a pipe bender and with a little work, made some mini heat pipes. The only Screenie I had of 5.6Ghz was lost due to a corrupted HD. Its hard to keep the OS in one piece at that speed. ive got verified CPUZ's up to 5.3ghz and benchmarks at 5ghz. I also ran the folding prog for 24 hours at 4.9Ghz. I very seldom run that system below 4.5ghz. If they would fix the charts at CPUZ so that the same person couldnt be listed twice with the same rig I would probably still be in third place for RAM frequency.

Now for the important news. Wed afternoon things cooled down with that front moving through the area. Me and my grandmother went fishing and the Bass were feeding like crazy. Im wearing them out on that Flappin Shad I told you about, but im down to 1 new one and 2 that are in really bad shape. I cant find them anywhere.

ASUS P5WD2 Premium
Intel 3.73 EE @ 5.6Ghz
XMS2 DDR2 @ 1180Mhz

<A HREF="http://valid.x86-secret.com/records.php?PHPSESSID=792e8f49d5d9b8a4d1ad6f40ca029756" target="_new">#2 CPUZ</A>
SuperPI 25secs

 

[Rugger]

5.6 with above 0C - NICE! Mfr heat pipe, freezer for sub-zero temps - you're always outside the box when it comes to the OC.

I've been all over the place, literally <b>and</b> figuratively. I'm around Panama right now and won't be back home for another 3 weeks. I've been gone a week and I'm already itching to get my hands on a rod! I'm gonna have a huge itch by the time I get back. You should take a look at the Zoom Flukes or Super Flukes (same concept except the super fluke is larger. Rig that with a 3/0 - 5/0 hook (I use an extra wide gap worm hook). In order to reduce line twist you can put a good barrel swivel about 8-12 inches above the hook. I usually fish this with a spinning rod an 8-10lb test for longer casts because the rig is so light. The Super Fluke is heavy enough to use with a baitcaster, but the regular fluke really light. The flukes come in tons of colors - you know the colors that work best in your water so I won't bore you with that.

__________________________________________________
<font color=red>You're a boil on the arse of progress - don't make me squeeze you!</font color=red>

 

[Rugger]

I've also been throwing the Zoom Horny Toads with some success. Not a lot of huge fish, but getting a lot of hits. They seemed to like a steady, slow retrieve on the top over structure. I let it sink a couple of times and then would twitch it and got a couple of good fish like that too. I think the sit/twitch is similar to the profile of the swimmin shad/Super Fluke. I left before the temps started dropping any and the fishing was HARD. Hard work and hardly any fish! Still had a great time doing it though...

__________________________________________________
<font color=red>You're a boil on the arse of progress - don't make me squeeze you!</font color=red>

 

[mozzartusm]

Staying outside the box is my motivation. It didnt take me long to figure out that there are many half-truths floating around when it comes to computers. Sub zero and Phase is great, but how many people can afford to go that route.

I keep hearing about people tearing them up with Super Flukes, but I have never tried one. Ive got just about any bait you could ask for, but when it comes down to it I have 4 to 5 that I use. Spinner bait, Plastic(Worms, lizards shad CRAWFISH), Poppin Frog, and Rattle trap. Let me know when you are back or just before you get back and I will plan out a day up my way. The weather has been good lately, maybe it still will be. I consider rain to be good this time of the year.

ASUS P5WD2 Premium
Intel 3.73 EE @ 5.6Ghz
XMS2 DDR2 @ 1180Mhz

<A HREF="http://valid.x86-secret.com/records.php?PHPSESSID=792e8f49d5d9b8a4d1ad6f40ca029756" target="_new">#2 CPUZ</A>
SuperPI 25secs

 

[Rugger]

Flukes are <b>great</b> baits and easy to use. Standard profile is: Twitch, twitch...Pause...Twitch, Twich...Pause. Most of the bites will come on the pause while the bait is falling. I've got TONS of baits myself, but I think every angler has his "confidence bait/s." Baits that will be on the pole when the fishing is tough. I use a good bit of soft plastics, buzzbait and spinner bait. I'm working more with a jig&pig and twitch baits (straight Rapala or similar) and some more topwaters (Devil Horse and those Horny Toads).

There're half truths floating around everywhere...you just have to know how to sniff them out...

__________________________________________________
<font color=red>You're a boil on the arse of progress - don't make me squeeze you!</font color=red>

 

[Zoron]

Bye.

Finally... don't let the door hit you in the ass on the way out.

The point is DDR2 memory will have a potential longer useful life for future processors.

Perhaps... but only if you're using Intel processors.

So if I upgrade a X2 next year to a faster processor I will still be stuck with DDR ram. That may not be a concern for you or many others but something that does concern me.

Oh noze!!! I'll be teh stuck!! (Or should that be suck?)

Anyway, if it's such a damned concern for you, then by all means buy a [-peep-] 820D already. You're wasting everyone's time here by pretending to have a 'discussion' when really all you're doing is posting reasons you should buy the damn thing. You're not convincing anyone here but yourself... so quit wasting your keystrokes. DDR RAM is going to be around for a while yet... so don't get your panties in a bunch about it.

<font color=red> If you design software that is fool-proof, only a fool will want to use it. </font color=red>

 

[rettihSlluB]

Hey Noko:

Are you still fully committed that Intel EMT64 implementation is better than AMD's 64 bit offerings?

I bet this will change your mind and it comes from a person with great knowledge of microprocessor architecture (you can read the whole thing at AMDzone):

===================================================================================
As far as EMT64 goes, it is not fully implemented on the Prescott core. Prescott actually uses the PAE hardware to extend the addressing space. What this means is several things. For one, I/O hardware is incapable of accessing memory above 2GB. This isn't such a big deal with consumer hardware but can be a nightmare with professional hardware like DMA RAID controllers. This requires the program to use patches that copy the data down to the first 2GB of addressing space where it can be accessed by I/O and also to copy data from I/O back up to high memory which removes a lot of the benefits of the DMA hardware.

This hacked style of extended addressing should cause a distinct slowdown (compared to Opteron) when Xeon runs with more than 2GB of memory. This is why sites like Tom's have carefully avoided benching systems with more than 2GB of memory. Again, it won't bother consumer systems that much but this is a big hit on large systems where time is a factor.

Also, Prescott Xeon hits the ceiling at 36 bits of addressing rather than 40 bits like Opteron. If you need more than 64 GB's of memory with Xeon you are out of luck. These are two problems I know of right off the top of my head. Perhaps someone can mention others that they know about.
===================================================================================
Enjoy!

My Beloved Rig:

ATHLON 64 FX 55
2X1024 CORSAIR XMX XPERT MODULES
MSI K8N DIAMOND (SLI)
2 MSI 6800 ULTRA (SLI MODE)
OCZ POWERSTREAM 600W PSU

 

[noko]

BSer,

Thanks for the links, on LostCircuits article it has some inaccuracies such as this one:

The 955X Chipset

For the launch of the dual processor series, Intel has released a number of additional chipset spins that are pretty much based on the original 900 series design, albeit with minor variations. Flagship is the 955X chipset with support for DDR667 and HyperThreading, <b>followed by the 945 chipset that is limited to DDR533 and does not support HyperThreading</b>.

The 945 supports both 667 DDR2 and HyperThreading.

I've found this benchmark useful because I use WME9 to encode with:
<A HREF="http://www.lostcircuits.com/cpu/p4_840/10.shtml" target="_new">http://www.lostcircuits.com/cpu/p4_840/10.shtml</A>

Intel does better single processor over the AMD offerings but look at the dual core. Kinda interesting benchmark. Their I would have expected Intel to continue the tread but didn't. Still the 820D should do well there especially if OC to 840D speeds and beyond. I just don't see the benefit spending extra here for the performance improvement of the X2. In any case any dual core scenario beats any single core setup.

The Lightwave benchmark is just plain screwy and not sure if valid, if so then this particular area the X2 shines rather well. Now if I used Lightwave and this was done valid, X2 would be the only way to go!
<A HREF="http://www.lostcircuits.com/cpu/p4_840/14.shtml" target="_new">Caligari TS5.1 Vases</A>

This is more applicable to what I use but a much older version was tested in the benchmark, the newer versions multitask leaps and bounds better then TS5.1 did, even on my AMD system ( Ts will multithread as many cpu's you have available, even hundreds now):
<A HREF="http://www.lostcircuits.com/cpu/p4_840/18.shtml" target="_new">http://www.lostcircuits.com/cpu/p4_840/18.shtml</A>

In this case the 840D (if I OC the 820D to 3.2gh should be very similar) was 20% slower then the X2 4800+. I, myself don't consider the addition cost worth while for the extra 20%. Now if I use faster timings on the ram and OC the 820D higher then 3.2gh (I expect to) then the difference will probably be mute and really not worth the additional cost for me. How will the X2 3800+ performance compared, I would give it a -3% to -5% for the smaller cache, minus another 20% for the speed difference between the X2 4800* but once again I am sure that processor would OC as an option as well, now thinking about it, with the smaller cache of the X23800+ it may even OC better, hmmmm. The X2 3800+ is definitely an option here.

Now I have to eat some crow here, yes indeed the Intel 840D EE uses 60w more power then the X2 4800+. Hot mother!!! Figuring 75% efficiency of the power supply 60w/.75 = 80w more power from the Intel happy family at peak power of processor :redface: . I wonder how much Intel will consume oc to 3.6ghz???? How much does a 100w lightbulb cost running 120hours a week???

Now I found this very interesting from that article:

</b><font color=red>In the case of AMD?s Athlon 64 X2, the dual core has one distinct disadvantage, that is, each core has essentially only access to a single channel of memory. That is, both cores have to go through the system request interface and a crossbar switch to get access to the memory and, in a situation where both cores simultaneously request data, this can lead to contention. . .
<A HREF="http://www.lostcircuits.com/cpu/p4_840/" target="_new">http://www.lostcircuits.com/cpu/p4_840/</A>

</font color=red></b>

This showed throughout the testing with the dual Opteron benchmarks included with this review. Maybe this explains why at AnAndTech the multitasking scenarios Intel won out more then AMD but at single tasks AMD just blew away Intel.

Now if Tom's Hardware can do a real test of the dual core plateform with 2-4gb of ram (fast timings on both plateform) and do some rendering in the background and gameing in the foreground. That would be very interesting. Video encoding while rendering and playing a game . 64 bit testing as well.

 

[rettihSlluB]

Now I found this very interesting from that article:



In reply to:


In the case of AMD?s Athlon 64 X2, the dual core has one distinct disadvantage, that is, each core has essentially only access to a single channel of memory. That is, both cores have to go through the system request interface and a crossbar switch to get access to the memory and, in a situation where both cores simultaneously request data, this can lead to contention. . .
http://www.lostcircuits.com/cpu/p4_840/

Well, this is obvious. Only Opterons can benefit from non Uniform Memory Architecture because servers need faster acces and more bandwidth when handling memory-intensive applications like databases, etc.

It would be overkill to have a quad channel memory interface with dual core Athlon processors (it would be really nice, but motherboards would need 8 DIMMS to support it). Anyhow, the integrated memory controller is so efficient that there's no need to have a quad channel interface (well, not now), since both cores are fully fed with data; on the other side, Intel's dual core offerings have to fight for data because of the shared bus architecture. That's why AMD's offerings can scale a lot better than any processor Intel has released to date. That's why reviews favor the X2 over the 840EE (in a ratio of 90/10) and thats why the X2 can multitask a lot better than an 840EE even if you keep on stating that it doesn't . 😉

Right now, I'm planning to sell my current processor and buy an X2 3800+ and overclock the hell out of it.
I'll keep the rest of components you see on my sig.



My Beloved Rig:

ATHLON 64 FX 55
2X1024 CORSAIR XMX XPERT MODULES
MSI K8N DIAMOND (SLI)
2 MSI 6800 ULTRA (SLI MODE)
OCZ POWERSTREAM 600W PSU

 

[noko]

Very nice rig, hope you do it soon and let us known your results. I am sure you will be able to do a better job then virtually any reviewer has done so far. Plus OCing the 3800+ would be very interesting to say the least. Having 2gb of memory seems very prudent now as a side note. In your case it would be stupid to go with Intel, all you have to do is just replace the cpu with a X2. Power to ya.

 

[noko]

<font color=blue><b>You are correct in your thinking about the P5WD2. Its the best by far for the Newer Intel CPUs.

</font color=blue></b>

Holy smokers!!! You are right!

<A HREF="http://www.pcper.com/article.php?aid=143" target="_new">P5WD2 Review</A>

<b><font color=red>Me want so bad!</b></font color=red> 😱

 

[noko]

Now here is a very good review of both X2 3800+ and 820D using many 64bit benchmarks with multitreads.

<A HREF="http://www.techreport.com/reviews/2005q3/athlon64-x2-3800/index.x?pg=1" target="_new">http://www.techreport.com/reviews/2005q3/athlon64-x2-3800/index.x?pg=1</A>

Not too much multitasking but plenty of well thought out tests. Wide variety of CPU's as well from both camps. Best review I've seen done so far, plus they tested the Intel mahines with fast ram timings (about time) not that it matter to much in the end . I just wish it was at 2gb or even 4gb, in time it will happen. The 64 bit performance of Intel did not suffer and was consistent as far as I could tell through many tests. There are differences between Intel and AMD designed chips and sometimes it really depends more on what you use which one is bested suited for you.

For example, this 64bit multithreaded test using alot of vector math has the 820D 14% faster then the 3800+ for Interger portion of test and 19% faster in floating point.

<A HREF="http://www.techreport.com/reviews/2005q3/athlon64-x2-3800/index.x?pg=11" target="_new">http://www.techreport.com/reviews/2005q3/athlon64-x2-3800/index.x?pg=11</A>

That was a rare exception, otherwise the X2 3800+ virtually sweap the floor over the 820D. Also TechReport was able to get a decent overclock from 2ghz to 2.4ghz and it actually approached X4800+ performance, real nice.

Power wise, well I get to eat more crow here, take a gather at those numbers! 😱

I think I could win either way but still thinking going with Intel this time around for various reasons .

 

[mozzartusm]

Asus has ruled the Intels Mobos other than SLI since the 925 chipset. Ive been told that DFI is about to release one bad Motha for Crossfire.

ASUS P5WD2 Premium
Intel 3.73 EE @ 5.6Ghz
XMS2 DDR2 @ 1180Mhz

<A HREF="http://valid.x86-secret.com/records.php?PHPSESSID=792e8f49d5d9b8a4d1ad6f40ca029756" target="_new">#2 CPUZ</A>
SuperPI 25secs

 

[Guest]

Yah the new crossfire chipset from ATI looks like a real nice one! AMD/Intel alike! That saphire board anandtech reviewed was not bad at all. Intersting times in the chipset area with ULI/ATI putting out new and innovative chipset. Also I have to wonder what use we can have for that 3X PCIe16X board from eVga!

Asus P4P800DX, P4C 2.6ghz@3.25ghz, 2X512 OCZ PC4000 3-4-4-8, MSI 6800Ultra stock, 2X30gig Raid0

 

[Guest]

Wow noko, nice post! Like too see you being objective!

Asus P4P800DX, P4C 2.6ghz@3.25ghz, 2X512 OCZ PC4000 3-4-4-8, MSI 6800Ultra stock, 2X30gig Raid0

 

[rettihSlluB]

Really nice post noko 😉.

I have to say something about this:

For example, this 64bit multithreaded test using alot of vector math has the 820D 14% faster then the 3800+ for Interger portion of test and 19% faster in floating point.

http://www.techreport.com/reviews/2005q3/athlon64-x2-3800/index.x?pg=11

Well, there's a simple explanation about that and you can find the answers <A HREF="http://www.swallowtail.org/naughty-intel.html" target="_new">here</A>

SiSoft Sandra relies heavily on the Intel compiler, that's why we see such a performance drop in integer and floating point operations (we all know that the Athlon architecture has the most powerful floating point engine available for x86 processors) 😉




My Beloved Rig:

ATHLON 64 FX 55
2X1024 CORSAIR XMX XPERT MODULES
MSI K8N DIAMOND (SLI)
2 MSI 6800 ULTRA (SLI MODE)
OCZ POWERSTREAM 600W PSU

 

[noko]

Yes, that is a good article, the optimizations when applied to the Opteron netted 10% increase using SSE2. Makes you wonder if SiSoft was tainted on the multimedia benchmark, that is if they did use the Intel compiler and was unaware of this, how many other software packages? Has anyone else collaborated his findings?

Well the upgrade adventure has started, I ordered two SATA drives from NewEgg yesterday .

<P ID="edit"><FONT SIZE=-1><EM>Edited by noko on 08/10/05 08:04 PM.</EM></FONT></P>

 

[P4Man]

>SiSoft Sandra relies heavily on the Intel compiler, that's
>why we see such a performance drop in integer and floating
>point operations (we all know that the Athlon architecture
>has the most powerful floating point engine available for x86
> processors) 😉

I wouldnt be too quick to jump to conclusions. sisoft doesnt use x87 code afaik, which is where K8 reigns supreme, instead it uses either integer or SSE. There are quite a few workloads, especially SSE code, that even with proper optimization will fun faster on a high clocked P4 than on a K8. Now how representative mandlbrot creation is for anything else than well.. creating madelbrots is another issue all together.

= The views stated herein are my personal views, and not necessarily the views of my wife. =