Okay, i had to join the X2 fun!

[n7]

There they are...
Teh X2 3800+
Installed
Dual P95s

I was lucky & got the Toledo core, like i'd sorta hoped.

So far it doesn't appear i'll be getting a fantastic OC though, sadly.
I got excited & initially just kept bumping her up till i was at 2.65 GHz, but P95, SP2004, & SuperPI all would fail instantly, so i've backed down to 2.5 GHz to do some testing before trying for anything more.
As usual on the damn Neo2 (yet i still love teh mobo), vcore fluctuates wildly from 1.52V to 1.55V under load...
I will try to back down on vcore later if it turns out it doesn't help.

The really great thing though so far isn't the CPU though, it's my new Thermaltake Big Typhoon.
It was a challenge to get on, but so far it seems to have made a huge difference.

With my Venice & former Zalman 7000B-AlCu, i reseated the HS about five times, & i'd still get 65C under load (yet perfectly stable), so i assumed my temps sensor was borked.
Now with the X2 & TT Big Typhoon, i'm getting 55C under load, which i am still guessing is somewhat off.
Regardless though, it's a 10 degree drop with a CPU that should theoretically run even hotter :Q!
So i am very happy, especially since it's just as silent as the useless Zalman was

Pardon the long babble, i just felt like sharing



Edit: It seems like my X2 doesn't like stress-testing programs so far.

P95 is stable at 2.4 GHz, but not higher...

However, everything i've been doing has been just great @ 2.6 GHz, so i am not overly concerned that Memtest & P95 fail.
I use teh PC for video encoding & gaming, & so far, all has been well @ the "unstable" 2.6 GHz

I may have to try a new bios or two to see what happens.

 

[AnImuS]

Originally posted by: n7
There they are...
Teh X2 3800+
Installed
Dual P95s

I was lucky & got the Toledo core, like i'd sorta hoped.

So far it doesn't appear i'll be getting a fantastic OC though, sadly.
I got excited & initially just kept bumping her up till i was at 2.65 GHz, but P95, SP2004, & SuperPI all would fail instantly, so i've backed down to 2.5 GHz to do some testing before trying for anything more.
As usual on the damn Neo2 (yet i still love teh mobo), vcore fluctuates wildly from 1.52V to 1.55V under load...
I will try to back down on vcore later if it turns out it doesn't help.

The really great thing though so far isn't the CPU though, it's my new Thermaltake Big Typhoon.
It was a challenge to get on, but so far it seems to have made a huge difference.

With my Venice & former Zalman 7000B-AlCu, i reseated the HS about five times, & i'd still get 65C under load (yet perfectly stable), so i assumed my temps sensor was borked.
Now with the X2 & TT Big Typhoon, i'm getting 55C under load, which i am still guessing is somewhat off.
Regardless though, it's a 10 degree drop with a CPU that should theoretically run even hotter :Q!
So i am very happy, especially since it's just as silent as the useless Zalman was

Pardon the long babble, i just felt like sharing

:thumbsup:

 

[Zim]

1.53v?? Sizzle!

 

[Pabster]

Originally posted by: n7
There they are...
Teh X2 3800+
Installed
Dual P95s

I was lucky & got the Toledo core, like i'd sorta hoped.

So far it doesn't appear i'll be getting a fantastic OC though, sadly.
I got excited & initially just kept bumping her up till i was at 2.65 GHz, but P95, SP2004, & SuperPI all would fail instantly, so i've backed down to 2.5 GHz to do some testing before trying for anything more.
As usual on the damn Neo2 (yet i still love teh mobo), vcore fluctuates wildly from 1.52V to 1.55V under load...
I will try to back down on vcore later if it turns out it doesn't help.

The really great thing though so far isn't the CPU though, it's my new Thermaltake Big Typhoon.
It was a challenge to get on, but so far it seems to have made a huge difference.

With my Venice & former Zalman 7000B-AlCu, i reseated the HS about five times, & i'd still get 65C under load (yet perfectly stable), so i assumed my temps sensor was borked.
Now with the X2 & TT Big Typhoon, i'm getting 55C under load, which i am still guessing is somewhat off.
Regardless though, it's a 10 degree drop with a CPU that should theoretically run even hotter :Q!
So i am very happy, especially since it's just as silent as the useless Zalman was

Pardon the long babble, i just felt like sharing

:thumbsup:

Looks like the 3800+ is the chip of choice these days. Not that I regret my 4400+ purchase

 

[n7]

Bah, i am just getting annoyed now...

One of the P95s just failed, & one SuperPI failed @ 2450 MHz

I've got my RAM @ ultra loose settings too, so it can't be that...

Is there anything else i need to do for these programs aside from setting their affinity to CPU 0 & 1?

 

[Markfw]

Well, what is your divider ? What voltage ? And I don;t trust Prime95, F@H is my indicator. However, superpi is quick, and if it works, I hardly ever have to go down from whatever worked there.

 

[doc2345]

Originally posted by: PabsterLooks like the 3800+ is the chip of choice these days. Not that I regret my 4400+ purchase

Of course there's no reason to regret your purchase. You knew what you were getting! You can buy what you want or you can "roll the dice" and hope for a great OC chip.... If your first 3800+ purchase doesn't work out then you can buy another one. Now you've spent $150.00 more than you needed to, you have 1/2 the cache, less the performance, out $750.00, and probably still don't have what you were looking for. If the second one doesn't work out, the hole just gets deeper and the game continues......

Sorry, I just don't understand the concept....... For all that won the bet on the first roll of the dice, three cheers for you, for the rest of us, make an inform decision and purchase what you are looking for....... It may not ultimately give you "extreme bragging rights" that some people here enjoy, but it does save you money, time, and grief??..

 

[n7]

RAM is at 100 (1:2) divider with extremely loose timings (3-4-3-8).

I know what the the RAM can do anyway, since it's stable at the setting in my sig with my Venice, but i figured i should take it outta the equation anyway.

I will be rather pissed if i cannot get 2.4 GHz stable, but then again, i may just say screw P95 & SuperPI.
If i can re-encode video 24/7 & play UT2k4 all day, i don't really care too much about what P95 says...

 

[Neptune3000]

I want one!

 

[sandorski]

nice mountains

I keep trying to talk myself into an X2 3800.....

 

[RichUK]

:thumbsup: you got a toledo core aswell

 

[Furen]

have you tried any other divider (besides 100) yet?

 

[DDDavey]

Quick question...what's the Toledo core, and how can you tell?

 

[RichUK]

Originally posted by: DDDavey
Quick question...what's the Toledo core, and how can you tell?

look at this

if you look at the etching on the IHS

the CD at the end of the first string "ADA3800DAA5CD", indicates it is a Toledo core, where as the same string with "BV" at the end indicates a Manchester core

The etching on the IHS is known as the CPU steppings

IHS (Integrated Heat Spreader), the metal plate on the top of the core, as see in this picture.

 

[govtcheez75]

I can't wait until mine comes!

 

[DDDavey]

Thanks for the explanation Rick...I figured it was some stepping "code" but I wasn't sure. I remember looking for a certain stepping 3 yrs ago when I built my last rig.

What's the main diff between the Manchester & Toledo core? I assume the Toledo is a better O/C'er, is that all?

Thanks again.
Dave

 

[RichUK]

Originally posted by: DDDavey
Thanks for the explanation Rick...I figured it was some stepping "code" but I wasn't sure. I remember looking for a certain stepping 3 yrs ago when I built my last rig.

What's the main diff between the Manchester & Toledo core? I assume the Toledo is a better O/C'er, is that all?

Thanks again.
Dave

The Toledo core is originally manufactured as having 2x 1Mb L2 cache per core, which is what the 4400+, and 4800+ fully take advantage of. The X2 3800+ using the Toledo core, has half of the L2 cache on both the cores disabled, therefore its ability to only utilise 512Kb of L2 cache per core.

Im not sure whether the Toledo X2 3800+ cores are either, failed L2 cache cores that are re-badged, or their yields were not sufficient for the higher clocked 4400+/4800+ cores, or whether it was just simply released to combat the cheap Intel 820 dual core, and also whether the Manchester core was not being produced in high enough numbers . I think it is the latter two as the yields seem to be pretty good when people are usually hitting faster speeds then the top level 4800+.

The new Manchester core is to be the actual AMD X2 3800+ I presume once the fabs are shipping sufficient amounts. The Manchester core also has a smaller die, and reduced transistor count, as it has 512kb of L2 cache per core by default no more. I think from what i have heard and seen the Toledo core is the better Overclocker at the moment, and once the Manchester core is tweaked you might see higher Overclocks, maybe.

The Toledo E6 core has a transistor count of 233.2 million, and a die size of 199mm^2

The Manchester E4 core has a transistor count of 154 million, and a die size of 147mm^2

However, the disabled cache on the Toledo core only amounts to 154 million working transistors the same as the Manchester core, even though the core is still larger at 199mm^2.

EDIT: Anandtech info on the above here

People see the Manchester core as 2 x Venice cores, and the Toledo as 2 x San Diego cores. This is logical as the AMD K8 as been mentioned before, was designed with dual core in mind. They accomplished this by using their Cross Bar design, so all you would simply need to do i would have thought, is tape out an extra San Diego/Venice core on the current design so to communicate both cores on the Cross Bar. Of course tweaking the manufacturing process in the fabs will mean that AMD will be able to produce a higher volume of working cores, as on the current Venice ad San Diego.

 

[McGeyser]

If you want more from your X2, take a look at the DFI NF4 Ultra D mainboard, NOT the infinity series. The DFI NF4 Ultra-D board has smaller increment memory dividers, such as the 180 or DDR 360 setting, that works wonders. Also the PWMIC or mosfet power regulators are very high performance and produce solid power. I would bet you will stabilize your Vcore, and with your existing Ram, hit 2.6GHz prime stable, especially with the Typhoon.

 

[DDDavey]

Originally posted by: RichUK
The Toledo core has a transistor count of 233.2 million, and a die size of 199mm^2

The Manchester core has a transistor count of 154 million, and a die size of 147mm^2

However, the disabled cache on the Toledo core only amounts to 154 million working transistors the same as the Manchester core, even though the core is still larger at 199mm^2.

Awesome, thanks for the explanation. I was just reading this here too as you replied:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2484

So with the smaller die size of the Manchester, does that mean that it's going to run cooler? Or does the larger die size of the Toledo have the advantage of more surface area, allowing more heat to escape to the HSF?

 

[Skiracer]

What is your LDT multiplier? If still at 5, that is your problem. You need to keep the LDT bus below 1000 MHz. I have a Toledo core 3800+ currently running at 2.5 GHz, LDT multipler at 4 (1000 MHz), RAM multipler at 1.8 (PC360) for a memory clock of 450 MHz (900 DDR) all on a DFI Lanparty nf4 Ultra D motherboard at STOCK voltages. It runs Prime 95 stable overnight and never gets over 52C with the big typhoon cooler I'm using too. These toledo core X2 3800 CPUs are amazing!

 

[RichUK]

Originally posted by: DDDavey

Originally posted by: RichUK
The Toledo core has a transistor count of 233.2 million, and a die size of 199mm^2

The Manchester core has a transistor count of 154 million, and a die size of 147mm^2

However, the disabled cache on the Toledo core only amounts to 154 million working transistors the same as the Manchester core, even though the core is still larger at 199mm^2.

Awesome, thanks for the explanation. I was just reading this here too as you replied:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2484

So with the smaller die size of the Manchester, does that mean that it's going to run cooler? Or does the larger die size of the Toledo have the advantage of more surface area, allowing more heat to escape to the HSF?

I could not say, i would presume it would be about the same, with regards to say a 4200+ on a E4 core and a 4400+ on an E6 core (comparison of same clock and volts but different die size), as with the reduced die size you have less transistors producing heat but a smaller surface area to transfer heat, and the opposite for the larger Toledo core except for a larger surface area in which to transfer heat, im not sure what trade offs would benefit the most on a thermal design. Where as with the X2 3800+ Toledo core with the disabled L2 cache still on the same sized die, you might think the extra surface area would result in better heat transfer, but i honestly would not know I wouldn?t like to say, although in theory I would agree. What i do know is AMD need to sort out the thermal pad they stick on the cores to connect the core and the IHS of the dual cores, as i have heard that they are poorly attached, and thus poor heat transfer.

 

[Aenslead]

Originally posted by: n7
There they are...
Teh X2 3800+
Installed
Dual P95s

I was lucky & got the Toledo core, like i'd sorta hoped.

So far it doesn't appear i'll be getting a fantastic OC though, sadly.
I got excited & initially just kept bumping her up till i was at 2.65 GHz, but P95, SP2004, & SuperPI all would fail instantly, so i've backed down to 2.5 GHz to do some testing before trying for anything more.
As usual on the damn Neo2 (yet i still love teh mobo), vcore fluctuates wildly from 1.52V to 1.55V under load...
I will try to back down on vcore later if it turns out it doesn't help.

The really great thing though so far isn't the CPU though, it's my new Thermaltake Big Typhoon.
It was a challenge to get on, but so far it seems to have made a huge difference.

With my Venice & former Zalman 7000B-AlCu, i reseated the HS about five times, & i'd still get 65C under load (yet perfectly stable), so i assumed my temps sensor was borked.
Now with the X2 & TT Big Typhoon, i'm getting 55C under load, which i am still guessing is somewhat off.
Regardless though, it's a 10 degree drop with a CPU that should theoretically run even hotter :Q!
So i am very happy, especially since it's just as silent as the useless Zalman was

Pardon the long babble, i just felt like sharing

Crispy boy. You got a really nice rig, there.

I like the power supply. Really nice cables.

 

[Heinrich]

Originally posted by: Skiracer
What is your LDT multiplier? If still at 5, that is your problem. You need to keep the LDT bus below 1000 MHz. I have a Toledo core 3800+ currently running at 2.5 GHz, LDT multipler at 4 (1000 MHz), RAM multipler at 1.8 (PC360) for a memory clock of 450 MHz (900 DDR) all on a DFI Lanparty nf4 Ultra D motherboard at STOCK voltages. It runs Prime 95 stable overnight and never gets over 52C with the big typhoon cooler I'm using too. These toledo core X2 3800 CPUs are amazing!

This story and n7's sound quuite similar to my experience. I have my LDT at 4, my chip at 270x9, RAM at 200 (1:1) and it runs prime '95 and games constantly. It also does fine at 270x10 in the "real world" but fails prime within 5 minutes. I have to run all the way down to 250x10 for prime to be stable overnight. I'm not sure that I trust it when Prime fails so very quickly, so I'm sticking with 250x10 for now. Perhaps after some "burn in" I will fiddle some more. Still, a 25% o/c from 2ghz to 2.5 is nothing to be sad about. Newegg has the 4800+ Toledo at $$$$$$$$884. There is no 2.5

Oh, and I have a thermalright XP-90.

 

[rise]

Originally posted by: Heinrich
Oh, and I have a thermalright XP-90.

that made a big difference for me. got rid of my 7000cu and popped in an xp-90 with the stock fan from svc. got my temps down so that i'm 14+ hours dual priming now, 270x10 1.44v @ 48-49c. pwmic runs ~55c but that should be fine.

thats about all i want to push it although i may try 275x10 if i don't need more volts.

 

[Gronich]

n7

I thought I would list my current setting for you for the Cell Menu to see if there is anything glaringly different between your settings and mine...

Ignoring the Memory set-up to what you think is correct for you I use the following:

High Performance Mode [Manual]
Aggressive timing [Disabled]
Dynamic Overclocking [Disabled]
Clock Spread Spectrum [Disabled]
HT Frequency [3x]
Cool 'n' Quiet [Disable] - I use RMClock to do the same task.
Adjust CPU Ration [StartUp] i.e. 10x
Adjust CPU FSB Frequency [280] - I started at 220 and used Clockgen to push it up.
Adjust AGP Frequency [66] - PCI/AGP is locked when you start overclocking by default.
Adjust CPU VID [1.325v] - I use the Max. of 1.35v + 10% for 2850Mhz+
CPU Voltage [Over VID 8.3%]
Memory Voltage [2.85 V]
AGP Voltage [1.60 V]

I would say you are running a degree or two higher than what I get at the same voltage/speed. I have seen between 58 and 60oC when I was priming at 2.85Ghz with 1.55 volts which I thought was unacceptable as it runs at 52oC at 2.8Ghz with 1.4 volts.

Hope this helps.