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

How would you upgrade to an A64 system?

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

hitechjb1

Senior Member
Joined
Feb 2, 2003
Many members would like to see the various options for building an A64 system polled. With the help of Maxvla (a senior member), this is the poll thread:

How would you upgrade to Athlon 64? (Poll Thread)
Max selection = 2 (have to be done in a single polling session)
Poll closing data = Aug 31, 2004.

As A64 evolves and conditions change, may create new polls again down the road.
Have fun. :)

==================================================================
It looks like the first release of 939 CPU would be at least $500 (as of June 04), ....
Ref: http://www.ocforums.com/showthread.php?t=301383
End of July 2004, price drop for many 754, 939 CPU's, by about 50-150 USD.

This thread A64 CPUs, chipsets, motherboards
contains informations about the various A64 platforms (940, 754, 939, ...), A64 technologies,
informations about the various CPU's, chipsets, motherboards,
as well as performance difference, price performance tradeoff of the various platforms, ....

Please express what you may do or plan to do if going for an A64 system to upgrade existing or build a new one.

Consider just CPU + motherboard, possible scenarios:

1. Build and stay with 754 system only
754 2800+/3000+/3200+ NewCastle/ClawHammer
250 GB or K8T800 Pro motherboard (w/o PCI-e)
around $400, have price performance A64 all the time, but not top end A64
Price drop end of July 2004 for 754, 939 CPU, cost becomes $300 - 350.

2. Build 754 system now, rebuild a "better" 939 system later
754 2800+/3000+
250 GB or K8T800 Pro motherboard (w/o PCI-e) around $400
Price drop end of July 2004 for 754, 939 CPU, cost becomes $300 - 350.

"Better" is up to your interpretation, e.g. it may mean 939 with PCI-e.
Year end/early next year, ebay above get $200 back
939 3xxx+ (low PR 939)
motherboard with PCI-e at that time
around $500
net $400 - $200 + 500 = $700, have A64 all the time, last into 2005 year end

3. Use existing system, build a "better" 939 system later
Existing system = nforce2 + xp/barton or whatever currently using
"Better" is up to your interpretation, e.g. it may mean 939 with PCI-e.
Towards year end/early next year
939 3xxx+ (low PR 939)
motherboard with PCI-e at that time
around $500, but missing A64 for half a year, last into 2005 year end
(possibly ebay existing system for some money back)

4. Build "any" 939 system as soon as parts available
939 3500+
motherboard w/o PCI-e around $650
"Any" is up to your interpretation, e.g. it may mean 939 without PCI-e to get 939 up and running ASAP.
** if 939 motherboard has PCI-e, this choice would become more favorable for lasting into 2005 year end
Price drop end of July 2004 for 754, 939 CPU, cost becomes $500.

5. Wait for 939 90nm
939 90 nm
motherboard with PCI-e

6. Wait for dual-core A64 system

7. Wait for DDR2
DDR2 memory modules
This would most likely be towords end of 2005 (if it happens)

8. reserved

9. reserved

10. Some other ways
If so, how.
 
Last edited:
I have a palomino currently - so for me its option 4 and an entire upgrade of cpu, mobo, ram, dvd burner and graphics card.
 
OC Detective said:
I have a palomino currently - so for me its option 4 and an entire upgrade of cpu, mobo, ram, dvd burner and graphics card.

I thought you would choose 3, upgrade towards year end.

Option 4 means now with expensive 939.

Maybe I did not make it clear in the first post.
 
About a month after the nForce 3 250/K8T800 Pro are released, I intend to sell my Mobile Barton, and my Socket A motherboard, and upgrade to a S754 system, motherboard and processor to be determined.

I will wait untill the nVidia nForce 3 250 and VIA K8T800 Pro is released, because I'm pretty sure that a newer motherboard with a newer chipset will offer superior overclocking options, and will have a higher resale-value for when (if) it is time to switch to a 939 system.

I will wait a month after the first wave of nForce 3 250 and K8T800 Pro boards hit us so that I will have more information available, in order to determine which board is best for my purposes, mainly overclockability, overvolt features, and availability of memory voltage modifications.

The price of this probably wouldn't be too bad, considering that my older hardware will be sold for a little bit of money to help fund the upgrade.

Does a S754 Athlon64 exist, or is one forecasted, that has an 11X stock multiplier? If so, this is the processor I would get, even if it's more expensive because of a higher PR.

And that's my upgrade path plan :)
 
I'm going with 5. Stay with Socket A until way into 2005 until a 90nm Socket 939 Winchester becomes affordable... unless something comes up that would make me realize that there is a valid reaon for actually needing the extra speed + a good deal comes along.

The only people going with 939 now are the ones for whom $ is no object. I share the view that 939s will not be affordable until late Q4 2004 or Q1 2005 so Socket 939 is a 2005 thing, not 2004 imho.

I look at this whole thing as just hype...

Mind you, we're talking upgrades here. If I were buying a new system, it would be Socket 754.
 
felinusz said:

....
Does a S754 Athlon64 exist, or is one forecasted, that has an 11X stock multiplier? If so, this is the processor I would get, even if it's more expensive because of a higher PR.

And that's my upgrade path plan :)


Current for 754, the 3400+ is rated 2200 MHz with x11 multiplier, based on ClawHammer w/ 1 MB L2. It has the desktop 1.5V and the mobile DTR 1.5V versions.

But I think the best price-performance CPU now are the
mobile 1.4V 3000+ (x9), mobile 3200+ 1.4V (x10), mobile DTR 3200+ 1.5V (x10)
all with 1 MB L2, based on ClamHammer core,
until the PR moves down in price.

May be by the time that the 250 GB and K8T800 Pro motherboards are readily available in a month or two, the 3400+ would become better price.
 
3400+ will come down in price same time as Intel cuts their prices probably - that is pencilled in for 22nd August - down to $278.
 
I'm sticking with my rig (soon to have a shiny new 2600+M in it ;) i hope i don't crack this one!) until i go to college. Then i'll build a new rig probably based on the Athlon 64 unless intel has a better offering. Not sure what number that is.. probably 5 because college is next year.
 
But I think the best price-performance CPU now are the
mobile 1.4V 3000+ (x9), 3200+ 1.4V (x10), 3200+ 1.5V DTR
754 CPU, all with 1 MB L2
until the PR moves down in price.

But your highest available multiplier plays a much larger role with Athlon64 overclocking, because the chips and chipsets are capable of such high FSB speeds, much like Intel machines. As such, access to the 11X mutliplier (and below for the DTR versions) is worth the extra money involved, in my opinion. It's either that, or more money spent on superior RAM that can handle 250+ MHz. With an 11X multiplier, RAM that can do 240 MHz is not going to hold you back, and it's likely that with good cooling, both your processor and memory will be maxed out around the same time as you scale upwards :).

May be by the time that the 250 GB and K8T800 Pro motherboards are readily available in a month or two, the 3400+ would become better price.

I certainly hope so! I'm hoping that as more hardware is released, prices will really drop fast, as popularity and purchasing goes up.

The only people going with 939 now are the ones for whom $ is no object. I share the view that 939s will not be affordable until late Q4 2004 or Q1 2005 so Socket 939 is a 2005 thing, not 2004 imho.

If someone can afford an S939 platform, and is thinking of buying one, they still have to ask themself whether a S939 setup will really be worth the amount of money it's going to cost over a S754 setup. I'm pretty sure that 939 prices aren't going to be moving south anytime during the next year.
 
felinusz said:


But your highest available multiplier plays a much larger role with Athlon64 overclocking, because the chips and chipsets are capable of such high FSB speeds, much like Intel machines. As such, access to the 11X mutliplier (and below for the DTR versions) is worth the extra money involved, in my opinion. It's either that, or more money spent on superior RAM that can handle 250+ MHz. With an 11X multiplier, RAM that can do 240 MHz is not going to hold you back, and it's likely that with good cooling, both your processor and memory will be maxed out around the same time as you scale upwards :).
...

This issue about high multiplier for A64 in order to max out the CPU, HT bus, memory bus can be resolved by running memory frequency, CPU frequency and HT bus frequency with ASYNC locking, without resorting to expensive memory modules running at high speed. Memory speed is clocked as high as the memory module is allowed (until spending to get better ones).

Gautam, among others, who have 754 systems running pointed out with test results that the 754 single channel platform can consitently deliver 95%+ memory bandwidth efficiency in ASYNC mode of operation over a range of CPU:memory divider, ....

It should be pointed out that for 939 platform whose main advantage over 754 would deliever an additional 80% effective memory bandwidth (based on some preliminary test data, see link below).

I estimated for air/water cooling, a CPU with x9 multiplier suffices to max out the current CG revision CPU overclock range. And for water/extreme cooling, a CPU with x10 multiplier would better fit.

As price, yield and CPU overclockability mature, the higher PR and multiplier will move accordingly to our favor, i.e. higher PR and multiplier for lower price.

Setup scenario for a 754 system would be:

hitechjb1 said:
Typical Overclocking System

Price performance system
- A64 754 1 MB L2 ClawHammer, CPU revsion CG
(as price moves down, move to higher PR with small price differential)
- 3000+ with x9 multiplier (for air/water cooling)
- 3200+ with x10 multiplier (for water/extreme cooling)
- 1.5V desktop, 1.5V mobile DTR, 1.4V mobile CPU
- 250 GB or K8T800 pro motherboard
- 2x512 MB DDR500 or overclock equivalent
- SLK-948 or heat pipe
- As of May 04, cost around USD 600.

Target setup:
BIOS FSB setting: 250 - 300 MHz (FSB is a setting, not physically per se)
CPU 1 MB L2 ClawHammer CG revision
- x9, target to 2.25 - 2.7 GHz
- x10, target to 2.5 - 3 GHz
Memory at 250 - 300 MHz, ASYNC, effective BW = 3800 - 4560 MB/s
HT at x3, at least 750 - 900 MHz w/ DDR, or x4 from 1000 and up

Ref: A64 CPUs, chipsets, motherboards
 
Last edited:
I guess the question here is whether the extra 300+ points (supposedly) for dual channel is worth paying the extra price. Hence the comparison is for me a 3500+ v a 3200+ (former on 939, latter on 754 but exactly the same frequency and cache). If the 3200+ is around the $280 and the 3500+ around $500 then to me it is - even before considering the differential likely to decrease by September and despite only the say 5% performance improvement. I have to buy a number of other items for my upgrade (really only keeping case and monitor - rest of parts will be incorporated in a system I will give to my daughter). Hence the premium becomes a much smaller factor in my calculation as my outlay is greater plus I am future proofed.
 
hitechjb1 - I see your point, and it is very well made with the examples you present. However, I'm of the school that really, really wants my memory to be run 1:1, regardless of how efficient and effective the memory dividers are (Although naturally, my Socket A experience with memory dividers has lent an amount of bias to this view).

Perhaps I'm applying this incorrectly to the Athlon64 case. The way I see it is as follows:

Memory run 1:1 @ 240 MHz

>

Memory run out of sync @ a speed<240, with Processor FSB run at a speed>240

In the latter case, the potential of your memory is not being realized, unless is maxes out quite low. Overall system performance will be decreased. Maybe I'm incorrect in my choice, but I would rather take advantage of a higher multiplier to make the most of RAM known to max out at around 240 MHz, then run the same RAM, or lower quality overclocking RAM, at a crippled speed so that my processor can enjoy a slightly greater bandwidth (at the cost of my memory bandwidth, which is being held back).

I guess the question here is whether the extra 300+ points (supposedly) for dual channel is worth paying the extra price. Hence the comparison is for me a 3500+ v a 3200+ (former on 939, latter on 754 but exactly the same frequency and cache). If the 3200+ is around the $280 and the 3500+ around $500 then to me it is - even before considering the differential likely to decrease by September and despite only the say 5% performance improvement.

That's exactly what the question is :). The choice is up to you, and it sounds like you've got it very well planned out, especially when you take into account that you need to upgrade a lot of parts, and want a future-versatile setup.
 
3 would be gd, the 32bit performace increase is not worth so much $

64bits now are mainly for braggin and there is little 64bits software for normal users or games available

still need a long time before everything changes to 64bits

if u are really rich with $ to burn, of cos u can get them when 939 comes out
 
I've officially become an addict. My system's become a plaything for me, I'll admit it. I'm going to sell everything in a couple of weeks, right before I leave the country for vacation. Then, when I return in late August, I'll buy whatever the newest stuff is then. I don't need a car, so whatever I make has been going and will continue to go into my computer.

Felinusz, you have to view using dividers as using different multipliers. There's a CPU/mem divider. Mine's at 12, so it's like I'm effectively running at 12x209, with no hit in efficiency.
 
I know it's stupid, you don't have to tell me, but I'm for 2: price performance A64, mainly because my 2500+ barton is getting REALLY buggy...maybe just too much use at too fast speed? I don't know, but it's getting really bad off. I've had to step it down to 2090 Mhz (compared to 2135) due to random reboots and hotter temps.

Then again, there isn't any room on earth as dusty as mine:D . That's the problem I bet, soooo...anybody got a vaccum and hose to fit onto a window?:D :D :D

If I get wiser within the week (yeah...sure...like that will happen:p ), I might go for 5, keep existing system BUT replace bad barton with a 2500+ mobile barton and use it for the duration of the year.
 
murrayman said:
Just thought of something...

Why isn't this a poll?

Thanks for the suggestion.

I thought about it. I think let it run a little bit longer (few days ?) to make sure the different choices make sense, add and change some if needed. Then I may contact a mod to convert it into a poll.

In the mean time, please make a choice and post your opinion.


Suggestions and comments are welcome.
 
Added one more choice:

5. Wait untill 939 90nm CPU, DDR2


Now total 6 choices (increased from 5 choices).
 
Back