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

1.6ghz 1MB L2, mobile Athlon 64 @ newegg...

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

amdking

Member
Joined
May 21, 2002
is anyone gonna try one of these? Im anxious to see how well they overclock, they are all CG revision and really inexpensive and default voltage of 1.4v
These may be gem overclockers....but noone has tried yet...
anyone goin too?
I wish I could LOL
 
amdking said:
is anyone gonna try one of these? Im anxious to see how well they overclock, they are all CG revision and really inexpensive and default voltage of 1.4v
These may be gem overclockers....but noone has tried yet...
anyone goin too?
I wish I could LOL

Apart from how well it can overclock (which at this point probably no one knows among us), I see there is some potential problem with this one for overclocking, ..., will post what I think later, ....

Let other give some comment first.
 
Re: Re: 1.6ghz 1MB L2, mobile Athlon 64 @ newegg...

hitechjb1 said:


Apart from how well it can overclock (which at this point probably no one knows among us), I see there is some potential problem with this one for overclocking, ..., will post what I think later, ....

Let other give some comment first.
???
Would love to hear your input. What's wrong with it??

1.4v spec looks great, as does 1mb L2 cache. It's a CG Clawhammer like mine; 2.4-2.5ghz should be a safe bet with it.
 
Re: Re: Re: 1.6ghz 1MB L2, mobile Athlon 64 @ newegg...

Gautam said:

???
Would love to hear your input. What's wrong with it??

1.4v spec looks great, as does 1mb L2 cache. It's a CG Clawhammer like mine; 2.4-2.5ghz should be a safe bet with it.

That is what I would think logically at first, but ....

if it is totally unlocked, it would be perfect to try.

The problem is it is only unlocked to 8x.

Given the 2.5 GHz overclock potential of the CG revsion, one would need a motherboard with FSB setting of 300+ to achieve, unless running memory and fsb not at 1:1, penalty unclear for this.

Even if there are memory to run at such speed, which does not, at most 270 MHz or so, it would cost a fortune for those memory modules.

Corollary, A64 with multiplier x9 is borderline for overclocking.

A64 with multiplier x10 is best choice for now, as motherboard and memory most likely would be able to run at FSB setting of 240-250+ MHz so memory bus, HT bus and CPU can be maxed out about the same time.
 
Async is your friend. :)

Don't forget that the A64's memory controller is much more efficient than the off-die ones found in AXP chipsets. I don't run at 250x10 sync myself; I run at only 210 with a CPU/12 ratio. Running with this divider is rather analogous to running at 210x12. Efficiency per Sandra is consistently 95%, regardless of divider, a very different story in comparison to the AXP. Running HTT speeds of up to about 350 are no problem at all, so up to about 2.8GHz, there shouldn't be any issues.

Edit: Some of d]g[ts' results:

d]g[ts said:
I was bored so I did some testing allbeit limited.
A64 3200+ run at 2407 in all test. Ram is single stick of KHX3000 with 2-2-2-5 timmings @ 200 (CPU/12 divider as listed by CPU-Z)and 2.5-4-4-8 timmings @240 ( I know I could do better timmings here but these are average crap timmings for PC4000 ram)

Hexus Pifast
1:1 52.05 seconds
CPU/12 52.14 seconds

SuperPI 8meg
1:1 7 min 50 sec.
CPU/12 7 min 38 sec.

SuperPI 1meg
1:1 37 sec
CPU/12 37 sec

Sandra CPU arithmetic
1:1 10495/3817/4912
CPU/12 10495/3817/4911

Sandra CPU multimedia
1:1 23053/24787
CPU/12 23052/24787

Sandra MEM bandwidth
1:1 3728/3728
CPU/12 3135/3136


All but 2 test are equal Superpi 8meg seems to like tight timmings no matter what, And sandra memory bench I have always taken with a grain of salt.
I would have done some 3d test for you but if ya look in the sig you will understand why I didn't
If ya have another bench you want me to try LMK and dig me up a link.

On my 25K run ( Had 3 so far all different setups )
1. A64 @ 2430 ram @ 220 2-2-2-6 vid card @ 517/406 25056
2. A64 @ 2475 ram @ 247 2.5-2-2-6 vidcard @ 500/400 25114
3. A64 @ 2475 ram @ 247 2.5-2-2-6 vidcard @ 500/400 25213
 
Last edited:
Gautam said:
Async is your friend. :)

Don't forget that the A64's memory controller is much more efficient than the off-die ones found in AXP chipsets. I don't run at 250x10 sync myself; I run at only 210 with a CPU/12 ratio. Running with this divider is rather analogous to running at 210x12. Efficiency per Sandra is consistently 95%, regardless of divider, a very different story in comparison to the AXP. Running HTT speeds of up to about 350 are no problem at all, so up to about 2.8GHz, there shouldn't be any issues.

If you are correct about the ASYNC memory efficiency to 95%, then

can we get this scenario to run with 250 GB or K8T800 PRO

FSB setting = 320 MHz to max out CPU around 2560 MHz with X8 multiplier.

HT bus at 320 x 4 = 960 MHz

Memory with async 4:5 to run memory bus at 320 x (4/5) = 256 MHz (say using DDR 500 modules).

If so, problem solved.


PS: For most A64 bios setting, memory bus is CPU_memory_divider instead of the memory_FSB_ratio (though either way can set ASYNC).
I.e.
memory_bus_frequency = CPU_frequency / CPU_memory_divider
e.g. ASYNC, memory_bus_frequency = 2400 / 12 = 200 MHz


PS:
I looked at a pre-released 250 GB motherboard MSI Neo K8N, the highest FSB setting is 300 MHz, so highest CPU can get is 2400 MHz, with a X8 multiplier, which is less than what one would expect for the CG revision.

So a 1.4V version mobile 754 3000+ with 1M L2, x9 multiplier would be a better choice for CPU overclocking.
 
Last edited:
hitechjb1 said:


If you are correct about the ASYNC memory efficiency to 95%, then

can we get this scenario to run with 250 GB or K8T800 PRO

FSB setting = 320 MHz to max out CPU around 2560 MHz with X8 multiplier.

HT bus at 320 x 4 = 960 MHz

Memory with async 4:5 to run memory bus at 320 x (4/5) = 256 MHz (say using DDR 500 modules).

If so, problem solved.

Shouldn't be an issue at all. One could use, for example, 2.5x320 and remain in spec exactly. I run at 2.5x250 for 625; no problems here. I have run at 310x2 in the past and things seemed fine, but many have run into the 350's on both the K8T800 and nF3 150 alone.

PS to your last, most do overclocking inside of Windows using Clockgen. The BIOS maxes aren't usually relevant. I don't even remember what the maximum selectable in my BIOS is; I think it's only 250.
 
Last edited:
So if what you said holds, I am looking forward to order a 754 1.4 V mobile 2800+/3000+ 1 MB L2 when a reasonably good 250 GB motherboard is available, plus DDR 500 memory to try.

Probably I would prefer a 3000+ than a 2800+.

Price tag for these three?

CPU $180
Motherboard $120 (?)
2x512 MB DDR 500 ?
 
Go to a local Circuit City when they have a promotion on Kingston ValueRAM. You can look through the packages until you find one with Hynix BT-D43 chips, which are capable of 270-280MHz using 4-4-3 timings and 2.6-2.8v quite consistently. They are the same IC's found in many of the good PC4000 modules. Extremely inexpensive as well; around $70 for a 512mb module IIRC.

But, I thought that for the nature of your work, 939 was better suited?
 
Thanks for the hint of the memory.

Back to the memory bandwidth efficiency:

From d]g[ts's result:

Sandra MEM bandwidth
1:1 3728/3728
CPU/12 3135/3136

If I read and understand the number, the CPU is at 2407 MHz, CPU:memory is 12:1. Memory is running at 2407/12 = 200 MHz.

Max BW at 200 MHz = 3200 MB/s
The Sandra BW = 3135 MB/s

So memory BW efficiency is 3135 / 3200 = 98% which is good.


But from your run
On my 25K run ( Had 3 so far all different setups )
1. A64 @ 2430 ram @ 220 2-2-2-6 vid card @ 517/406 25056
2. A64 @ 2475 ram @ 247 2.5-2-2-6 vidcard @ 500/400 25114
3. A64 @ 2475 ram @ 247 2.5-2-2-6 vidcard @ 500/400 25213
Increasing memory from 220 to 247 only improve the 25K number very little. Was it due to drop in video card overclocking?
 
The 3DMark results are d]g[ts as well, sorry for not clarifying. 3DMark at the point he was at was quite dependant on the GPU, and the result just illustrates that running the memory even at lower speeds may not hurt real-world performance significantly, and the dividers certainly don't hurt very much. Sorry, I took this quote in haste, it wasn't showing what I thought it did.

One other thing- for people with much, much, much more electrical experience than me, a voltmod should let an A64 fly. They do scale very well with voltage, and 1.7v/2.5GHz is very impressive. The potential of what a vcore mod would deliver is tremendous. With 2.1v and sufficient cooling, I'd say 3 GHz is in the crosshairs.
 
Gautam said:

...
One other thing- for people with much, much, much more electrical experience than me, a voltmod should let an A64 fly. They do scale very well with voltage, and 1.7v/2.5GHz is very impressive. The potential of what a vcore mod would deliver is tremendous. With 2.1v and sufficient cooling, I'd say 3 GHz is in the crosshairs.

Do you mean the motherboard cannot set higher CPU voltage.

I looked the AMD tech doc for 754, there are five pins VID[0:4] which define the default voltage. I think if one spend time sometime to look at it, some wire trick mod can be done as in the case for XP.


Just glancing through the AMD tech doc briefly, the max absolute ratings on Vcore is only 1.65 V, which is much smaller than that of the various Tbred B/Barton of 2.05 to 2.20V.

The max voltage to the HT I/O circuit is 1.5 V (even lower).

The max voltage to the memory controller SDRAM I/O circuit is 2.9 V.

For details, refer to the AMD tech doc.

The abs max voltage table is also in
A64 CPUs, chipsets, motherboards

So be careful about applying too high voltage to the A64 CPU.
 
Last edited:
So the A64's are locked also ? Hmm that sux, i need to buy new memory then cuss my pc3200 Stick can only does 215fsb or 220fsb. And if i need 250 to run max overclock then ....
 
hitechjb1 said:


Do you mean the motherboard cannot set higher CPU voltage.

I looked the AMD tech doc for 754, there are five pins VID[0:4] which define the default voltage. I think if one spend time sometime to look at it, some wire trick mod can be done as in the case for XP.


Just glancing through the AMD tech doc briefly, the max absolute ratings on Vcore is only 1.65 V, which is much smaller than that of the various Tbred B/Barton of 2.05 to 2.20V.

The max voltage to the HT I/O circuit is 1.5 V (even lower).

The max voltage to the memory controller SDRAM I/O circuit is 2.9 V.

For details, refer to the AMD tech doc.

The abs max voltage table is also in
A64 CPUs, chipsets, motherboards

So be careful about applying too high voltage to the A64 CPU.

Yes, the memory controller is known to die after using vdimms above 3.2v or so. But plenty have run 2v for months without any problems(albeit on phase change cooling for the most part). They certainly put out a lot more heat than Bartons did, but that's to be expected. This explains why all motherboards seem to max out at 1.7v for voltage. :-/
 
they are locked from their default multiplier and up...you can use any multiplier from default and down.
Its not much of a problem as long as you find a motherboard capable of high fsb...in this case 300+ fsb w/ adjustable LDT and PCI/AGP locks

you dont have to worry about memory speeds much either because A64's first system where even in ASYNC they do nearly as good as SYNC for memory speeds.

I think that if I go for this mobile chip...which I just might give it a shot, and will pair it up with a Gigabyte K8nPro or even their new 250GB board, but I dont see any major differences between them in overclockability so...with time I guess I will decide =P

I love the approach to the A64's system overclockability. The discoveries made over time since they have been released have been to prove very flexible and decent gains.
 
Gautam said:
Go to a local Circuit City when they have a promotion on Kingston ValueRAM. You can look through the packages until you find one with Hynix BT-D43 chips, which are capable of 270-280MHz using 4-4-3 timings and 2.6-2.8v quite consistently. They are the same IC's found in many of the good PC4000 modules. Extremely inexpensive as well; around $70 for a 512mb module IIRC.

But, I thought that for the nature of your work, 939 was better suited?

HRM.....dont A64's love extreemly low latency ... I mean even over high speeds..or does the A64 like both?
Traditionally I know P4's love high latency and dont care about low as much as Athlon XP's do.
 
The A64's pack a much greater punch at higher latencies than AXP's do. As far as what sorts of memory timings and dividers you use, it's very similar to the P4, in that they don't get hurt much by loose timings and dividers don't hurt them at all. But, yes, low latency is preferable. At 210MHz 2-2-2 I do about as well as others at around 270 with 4-4-3 timings, so if you do go A64, it will love your 3500EL. The K8NS Pro with the nF3 250, non-GB is the best choice ATM.
 
Originally posted by hitechjb1

If you are correct about the ASYNC memory efficiency to 95%, then

can we get this scenario to run with 250 GB or K8T800 PRO

FSB setting = 320 MHz to max out CPU around 2560 MHz with X8 multiplier.

HT bus at 320 x 4 = 960 MHz

Memory with async 4:5 to run memory bus at 320 x (4/5) = 256 MHz (say using DDR 500 modules).

If so, problem solved.

Originally posted by hitechjb1
Shouldn't be an issue at all. One could use, for example, 2.5x320 and remain in spec exactly. I run at 2.5x250 for 625; no problems here. I have run at 310x2 in the past and things seemed fine, but many have run into the 350's on both the K8T800 and nF3 150 alone.

PS to your last, most do overclocking inside of Windows using Clockgen. The BIOS maxes aren't usually relevant. I don't even remember what the maximum selectable in my BIOS is; I think it's only 250.

Under the scenario that the BIOS FSB setting is not high enough (say 300+ MHz) to max out the CPU with a max multiplier of x8, ....

If Clockgen is used to set higher FSB setting, does that mean that after every reboot one would have to get into Clockgen to readjust the CPU frequency, .....

If so, motherboard with higher BIOS FSB setting (300+) or a CPU with 9x multiplier is preferred, not using Clockgen.

9x CPU is not much higher in price than 8x ones, potentially higher chance of overclocking, as price and yield mature, higher multiplier CPU will be more affectable.
 
Back