• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Socket 939: A chasing after the wind?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

sappo

Member
Joined
Aug 12, 2002
Location
guam
So Ed keeps telling me to wait for 939. He seems to think it will be fast and worth the wait. But then again that's what he said about the Prescott and DDR2 and PCI-E. Prescott was DOA. From what I'm gathering DDR2 isn't going to be worth it for a long time. Now many review sites are beginning to question the need for PCI-E.

This seems to me like a classic case of the tail wagging the dog.


Questions:

1) 939 ONLY offers dual channel over 754's single channel, right? If the nforce2 boards are any indication dual only has a performance advantage of like 5-10% over single in SYNTHETIC tests. In real world tests it's much smaller. Is this right?

2) People seem to say 939 will be expensive. Is this true? Even if they are relatively inexpensive, 754 boards are DIRT cheap. I wouldnt expect 939 to be as cheap as 754 for a while.

3) That said, what the heck is the big 939 deal anyway?!?
 
1. you can't directly compare nf2 and nf3 250. the nf2 had a maximum of 400mhz memory speed through the processor. doubling that with dual channel didn't help at all like trying to drive 2 cars down a 1 lane street. with the nf3 250 (939 chipset) you will have 1000mhz of ram speed through the processor. dual channel at stock 200 would equate to 800mhz of ram speed, well within the 1000 mhz. the scores for dual channel FX-51 and 53 are sky high even with registered ECC ram, just think what regular ram would score.

2. at first i thought 939 would be expensive too, but ed has posted a couple times about amd having decent low end options around 200 bucks which is within my price range. the mobo's aren't going to be much more. perhaps 25 more than their 754 counterparts. the thing i am looking forward to most is not the 939 and dual channel, but a GOOD board, not this hack/slash job nvidia did with the nf3 150.

3. there is always buildup for the new product. you just have to learn to see through it and judge for yourself if there is really anything to be excited about.
 
1. The nforce-2 Socket A, and the Nforce-3 Socket 939 design are not comparable in terms of performance increases. Socket A, has one 64bit wide memory channel running at a maximum of 200mhz (400DDR). The memory controller on the nforce-2 was a dual channel memory controller, which was capible of accessing the RAM in 64bitx2 chunks at 200mhz speed. Unfortunatly the bottle neck in this design was the FSB of the CPU which only allowed 64x1 bandwidth so the extra bandwidth provided by the nforce-2 memory controller was mostly wasted on Socket A systems.

Where as the socket 939 systems have 2x64bit memory controllers, so it will be able to fully utilize the memory bandwidth from two channels of memory. Realistically depending on the application I would expect a real world performance gain of 3-5%, which isn't noticable. In other words you won't be able to tell the difference between a 64x1 or 64x2 system all else being equal. I wouldn't pay a premium for this design for performance as there is no major performance benifit, for most applications.

2. 939 will be more expensive then 754 as this is their premium line and will be marketed accordingly. But at this point no body knows how much it really will cost, its just a guess until real product arrives. PCI-E is another case in point the current generation of cards won't benifit much from the design, until they are redisgend to take advantage. Same thing happend when AGP came out, there was no performance benifit for the switch, but as time passed and cards were optomized for the new interface performance increase. Same thing will happen with PCI-E, though its really not majorly faster than AGP 8x.

3. Personally as is generally the case the 1st generation of any new design isn't substaintially better than the previous design. Its the new design allows for faster future ramp ups that the older design can't possibly keep pace with. Besides isn't it time we got away from all these legacy standards, and had only one BUS on a computer. No more USB/Firewire/PCI/AGP/SATA/IDE/Ps2/Parallel/Serial/Ethernet/Floppy etc. etc.

That was the thought of the IEEE1394 standard and the eventuall direction of PCI-Express is to have on bus that everything plugs into. Makes computers cheaper, simpler, and more reliable.
 
If the difference proves to be 3-5% as you say, I could care less about waiting for a socket 939! Personally, I think it's just plain stupid. I think I'd speak for most if I said I wouldnt spend more than $30 dollars on a motherboard, ram, cpu combo to go with a 939 comptuer. And I seriously doubt we'll see a 939 rig that is that competitive by the end of the year --especially since it's a lot more expensive to have two dims that one!

Now I'm not anti-new technology, but it's gotta have at least one of two thigns:

1) A very noticable performance difference, or
2) and a very un-noticable difference in price.

Cases in point:

A 52X CD-RW. They're basically the same price as any other writer AND they're really fast.

Athlon 64: Very noticable performance difference (especially considering a A64 3000 is significantly faster than an XP3200. I guess the only thing holding me back on getting one is there's no motherboard with a PCI lock...... YET!

Or an AGP 8X video card. They arent too much faster than AGP 4X, but all the mid to high performance cards 8X. So in that instance i'd instinctively go for an 8X card.
 
rottenspam said:
Besides isn't it time we got away from all these legacy standards, and had only one BUS on a computer. No more USB/Firewire/PCI/AGP/SATA/IDE/Ps2/Parallel/Serial/Ethernet/Floppy etc. etc.

That was the thought of the IEEE1394 standard and the eventuall direction of PCI-Express is to have on bus that everything plugs into. Makes computers cheaper, simpler, and more reliable.

Er, I'm afraid I dont quite get you. Are we supposed to plug our ps2 mice directly into a pci-e slot? A common intervace sounds really cool, but with things like printers, keyboards, and mice I dont see how that would work.
 
939 performance is going to be the same as Socket 754 as the extra bandwidth will be cancelled out by the halved L2 cache. Also, the lowest clocked 939 chip will be the 3500+ so no, it will not be "reasonably" priced at the outset. Also, expect the usual issues with new chipsets if they implement PCI-E x1 and/or PCI-E x16.

939 is going to offer no performance improvement over 754 and it will be expensive as there are no plans for cheaper A64s on the platform. By the time you'll upgrade again you'll be looking for a new mobo as there will be more features that you want as is always the case. Besides, don't expect to see volume 939 production at least until late Q3 which should keep the chips at a premium price even compared to the same speed 754 chips.

In the end there's really no reason to wait; it's much ado about nothing. Buy now and enjoy.
 
I agree with sappo, 939, thats bull, I mean when 939 come there will be plans for hammer 2 whatever, it never ends
 
sappo said:


Er, I'm afraid I dont quite get you. Are we supposed to plug our ps2 mice directly into a pci-e slot? A common intervace sounds really cool, but with things like printers, keyboards, and mice I dont see how that would work.


You misread, he's doing an analogy.

He's saying how stupid(and he's right) it is to deal with 12,000 different ports and all of their compatability issues. He's saying they're trying to convert EVERYTHING(Mouse, keyboard, digital camera, printer) to firewire.

Compare this to PCI and AGP, why should our video cards have different slots than everything else, logically, it doesn't make sense. With PCI-E, it won't be, just one standard, the way Firewire will be for peripherals.
 
OK this has been discussed on an earlier thread, first up cpu prices for 939's, excluding the FX range, will be the same as those for the 754's. AMD cannot merely up the price, irrespective of quantity produced, unless they up the rating of the cpu - also they will not do that in isolation as they are not market leaders and need to see what Intel are doing with their prices so expect the top of the line non FX 64 bit 939 pin to be around the 420 dollar mark (same as present). Aside from technical innovations the 939 mobo may offer it is also future proofed whereas that cannot be guaranteed for the 754.
 
Bar81 said:
939 performance is going to be the same as Socket 754 as the extra bandwidth will be cancelled out by the halved L2 cache. Also, the lowest clocked 939 chip will be the 3500+ so no, it will not be "reasonably" priced at the outset. Also, expect the usual issues with new chipsets if they implement PCI-E x1 and/or PCI-E x16.

are you reading the same reviews i am? the 3000+ is right behind the 3200+ in every test. less than 2% difference in almost every case. exactly the same performance in some cases. the 3000 has 512k cache whereas the 3200 has 1mb. this means the cache has little to no effect on performance. having nearly double the memory bandwidth is going to improve performance alot more than 0-2%.

939 is going to offer no performance improvement over 754 and it will be expensive as there are no plans for cheaper A64s on the platform. By the time you'll upgrade again you'll be looking for a new mobo as there will be more features that you want as is always the case. Besides, don't expect to see volume 939 production at least until late Q3 which should keep the chips at a premium price even compared to the same speed 754 chips.
you have a 939? no? so how do you know 939 is not going to offer any benefits?
In the end there's really no reason to wait; it's much ado about nothing. Buy now and enjoy.
there's reason to wait if the buyer still isn't sure. you shouldn't ever buy something because someone says so. if you aren't 100% positive i would advise not to buy.
 
The the 939 will be able to take advantage of DC in much the same manner as the Intel 800FSB CPU's so this should be a sig performance boost unlike socket A DC.
 
Lets not forget the important fact that s939 will be AMD's socket of choice for quite some time whilst 754 will be dying off...
 
I would wait before making my decision and read the reviews, there alot of speculation about 939 and not enought facts, but if you need to make a choice it up to you and no one else.
 
I have a fairly high powered rig at the moment and have no real need to upgrade for the next year or two, so I get the benefit of watching other people's mistakes in the mean time.

The way I see it just now, Socket 754 is going to be the next Slot A. Sure, it is the biggest and baddest around at the moment, but it won't last that long. Socket 939 has a future, probably not quite as prestigious as Socket A did, but certainly more so than Slot A ever did.

I'll wait around happily until then. Roll on Socket 939.
 
Biggles 266 said:
I have a fairly high powered rig at the moment and have no real need to upgrade for the next year or two, so I get the benefit of watching other people's mistakes in the mean time.

The way I see it just now, Socket 754 is going to be the next Slot A. Sure, it is the biggest and baddest around at the moment, but it won't last that long. Socket 939 has a future, probably not quite as prestigious as Socket A did, but certainly more so than Slot A ever did.

I'll wait around happily until then. Roll on Socket 939.
754 is going to be supported at least through the end of the year. speculation is that 754 will become the value line with duron labeled chips, probably devoid of 64bit capability and lower cache.
 
Maxvla said:
1. you can't directly compare nf2 and nf3 250. the nf2 had a maximum of 400mhz memory speed through the processor. doubling that with dual channel didn't help at all like trying to drive 2 cars down a 1 lane street. with the nf3 250 (939 chipset) you will have 1000mhz of ram speed through the processor. dual channel at stock 200 would equate to 800mhz of ram speed, well within the 1000 mhz. the scores for dual channel FX-51 and 53 are sky high even with registered ECC ram, just think what regular ram would score.

Uh I think you forgot one major thing. All A64s have a onboard memory controller, northbridge/southbridge has nothing whatsoever to do with memory performance anymore. Unless you doing inter processor (SMP) communication nobody really cares about the HT speed, also its not a 1000mhz FSB, it 1000mhz serialized, sure Rambus ran at 800mhz but but it didn't give 8x the bandwidth over pc100.
 
Aren't you guys forgetting the main attraction of socket 939? The 90nm process should overclock far better than the 130 nm socket 754 chips. They should also run a bit cooler too. That's what I'm looking forward to at least...
 
saaya said:
you should fe wait for 939, not worth upgrading yet, and i dont think 939 boards will cost a lot.

What's a lot, is $175ish a lot?
This one was never in stock because it wasn't released yet and it's a 754:
http://www.zipzoomfly.com/jsp/ProductDetail.jsp?ProductCode=241577

tio said:
Aren't you guys forgetting the main attraction of socket 939? The 90nm process should overclock far better than the 130 nm socket 754 chips. They should also run a bit cooler too. That's what I'm looking forward to at least...

It's April 3, 2004 and you will not see anything like that until 2005, it may be as long as a full year from now before you see 90nm Winchester CPUs and maybe as long as the summer after next before they're affordable.


It'll be nForce3 250 Socket 754 mobo + a socket 754 CPU for most people in 2004.
 
Back