Latitude E4300 Owners' Corner

MySaturdaySelf · 07-11-2011, 02:31 PM

Quote:

Originally Posted by Dillio187

I see you are using mis matched memory chips. I have had display issues identical to yours in the past while doing this, you may want to revert to 2x2gb or get another 4gb stick to make an even 8gb.

thanks for the tip. it worked. now I got good reason to upgrade ram. lol

nando4 · 12-03-2011, 11:13 PM

Welcome to the E4300 Owners' Corner! A place to liase with other owners and share useful tips/tricks/mods to get the most satisfying user experience from Dell's last great ultraportable with an optical drive.

Release date: 12th August 2008 (pre-GFC build quality)

OS, drivers and disassembly

Manuals, Win7 OEM sources : the additional Dell installation contents added to a Win7 CD.

Storage

Second hotswappable HDD/SSD via optical drive bay
Add e-sata: Flush-type expresscard for fast external storage
USB 3.0: add USB 3.0 to a E4300 via it's expresscard slot.
e-sata/usb cable or enclosure, useful for external e-sata storage. Eg: convert optical drive to be external unit

Expansion

DIY eGPU to attach a desktop videocard via expressport. Provides HDMI/DVI and accelerated graphics.

Tweaks + other mods

Manual temperature control via the FN+15324 trick
Overclocking a E4300 with P9400 cpu to 3.16Ghz (333x9.5 @1.25V)
DC optimize optimize the software configuration to maximize battery life, idling at 5.5W with a HDD!!

Schematic available at notebookschematic.com. Can use the free E6500 schematic as a reference. Many details being very similar.

Reviews: NBR &, laptopmag, PC World,

So then how does one go about extending the functionality of the E4300 in a cost effective manner? Read on..

Acknowledgements: a cast of many E4300 and other NBR users contributing to this thread. Many thank yous.

Dell Latitude E4300 versus HP Elitebook 2530P

Spoiler :

nando4 · 12-12-2011, 07:34 AM

INFO: Overclocking a E4300 with P9400 cpu to 3.16Ghz (333x9.5 @1.25V)

WARNING: Overclocking means running components faster with an increase in operating temperature. Observe voltage and cooling precautions. I take no responsibility for any damage. User beware.

CPU (AC/DC voltage)	GHZ/BCLK/multiplier	Win7 WEI	cpumark	wprime 32M	TSbench 32M
P9400-2.40 1.25/0.925	3.16/333/x9.5^12	6.6/6.8/4.2/3.5/x.x	2200	26.25	32.82
P9300-2.26 ?????/?????	3.00/333/x9^12	6.5/6.7/4.2/3.5/x.x	2091	27.79	34.83
P9400-2.40 1.063/0.925	2.53/266/x9.5^1	6.2/6.2/3.9/3.4/x.x	1762	32.99	40.25
P9300-2.26 ?????/?????	2.40/266/x9.5^1	6.0/6.0/3.9/3.4/x.x	1670	34.36	42.38

^1 Requires a stage1 bios setting change and configuration of Throttlestop to run a dual-IDA overclock using an extra 0.5 multiplier.
^2 Requires a Stage2 PLL pinmod as noted below.

Introduction

I am typing on Dell E4300 with a P9400-2.40 CPU. After almost giving up I managed to overclock it to noticably quicker 3.16Ghz and perform a stack of DIY eGPU benchmarks with full stability - no freezes or BSODs. The brief on how to do it is below.

Stage1: overclocking with an extra 0.5 multiplier (dual-IDA)

I was able to dual-IDA overclock to a 9400@2.53Ghz after this stage. Disable Speedstep in the bios as shown and start from point (3) of the instructions here to gain an extra CPU 0.5 multiplier via Throttlestop. So a P9400-2.40@x9.0 will now run at 2.53Ghz@x9.5.

Stage2: FSLx pinmod the E4300's Silego SLG8LP554VTR PLL from 266 to 333Mhz

1: P9400 SuperLFM, LFM, HFM and IDA voltages. Also shows the 3.16Ghz OCed P9400.
2: The E4300 PLL and voltage regulator schematic snippet
3: The voltage regulator near the CPU. VID0-VID6 are arranged around top left corner starting third in from the top.
4: Turn the motherboard over and the VID0..6 resistors mirror the location of the voltage regulator and make CPU voltmods very easy.
5: The 333Mhz PLL pinmod. Add a 20k resistor between the PLL's pin23 and Vcc.

Increase the PLL frequency from 266Mhz -> 333Mhz (FS_C/FSC_B/FS_A=0/0/0 -> FS_C/FSC_B/FS_A=1/0/0). The secret is to use a 20k resistor from the FS_C PLL pin 23 to Vcc (I used the resistor attached to PLL pin 1 for Vcc). It must be 20k as the FS_C pin must set in a range b/w 0.7 to 1.5V. If the PLL samples > 2V then it engages TEST mode and system will not boot at all. With the 20K resistor a voltage divider is created with the existing 10k resistor attached to pin 23 so the PLL will sample 1.1V on bootup. That sets a logic 1 FS_C for 333Mhz operation.

The PLL can be accessed without disassembling the system. Remove the RAM coverplate, remove plastic cover below the Latitude On socket (similar to infinity symbol). That plastic is pinned down by the chassis standoffs so would need to be cut around the edges to remove it. There you'll see the PLL and there's sufficient room to be able to solder in the 20k resistor at the PLL pin23 position and Vcc (3.3V).

Other issues when increasing the PLL frequency from 266Mhz to 333Mhz

1. Need good RAM - Hynix 1GB/4GB modules failed to complete boot. It would BSOD. A 4GB G.Skill and 2x2GB Samsung RAM boots fine I'll be looking at how to slow down that Hynix RAM so it can work. Likely need to use Thaiphoon Burner as SPDTool only works with DDR2.

2. The CPU needs more volts - the IDA voltage of 1.2375 was not enough to run the CPU with complete stability at x9.5 under full load. Everest's CPU cache test would fail after about 10 seconds. If I didn't run IDA mode, then the CPU could only use 1.15V which was stable only up to a x9 multiplier (333x9=3.0Ghz).

The fix I applied was to remove the VID0 voltage regulator resistor, effectively setting VID0=0, so that the 1.2375V setting was now 1.25V. That was enough to get a stable operation. Earlier on I lifted the VID2 resistor and routed VID1 to that pad which gave me 1.2875V. That was also completely stable but the CPU temps were too hot for comfort. I was seeing more than 90 degrees at full load. The VID0 mod was better as temps remain under 85 degrees under full load.

I believe a P9300 CPU wouldn't need any CPU voltmods. It's lower 3.0Ghz (333x9) dual-IDA mode would probably be OK with the IDA voltage of 1.2375V. Certainly, I could run a stable 333x9@1.15V on my P9400, but IDA was disabled. This would be a IDA mode on a P9300 and they tend to require more voltage.

Dell E6400/E6500 could also be overclocked

I don't have a E4300 schematic but I found a free E6500 schematic to help me along. Looks like the PLL and voltage regulator details are the same for the E4300/E6400/E6500.

So the same overclock could be applied to a E6400/E6500 with those systems having the advantage of socketted CPUs. A E4300 has a soldered CPU. Means the voltmod could be done on the E6400/E6500's CPU socket like shown in BSEL Mod on a socket P explained with photos so no soldering is necessary there.

Conclusion

My P9400 is overclocked to give near full-power i3 performance yet can still be undervolted to 333x6 (2.0Ghz @0.925V) so it gives better battery life than a 1st gen i-core. Eg: My DC Optimized settings see idle power consumption of 5.5W with a 7200rpm HDD . I quite like the flexibility of the Penryn C2Ds. The OC completes the excellent E4300 ultraportable package

References used

Latitude E6500 schematic for the PLL and voltage regulator circuits
Intel Core2Duo CPU VID table to see how VID pins will alter the voltage delivered to the CPU.
SLG505YC264 and SLG8LP564V Silego PLL datasheets. Close approximations to a SLG8LP554VTR whose datasheet I wasn't able to find.

nando4 · 12-12-2011, 07:43 PM

INFO: DC optimize a E4300 for maximum battery life

Above: P9400's SuperLFM. LFM, HFM and IDA voltages

Applying the following tweaks gives improved battery life over the factory-delivered state:

0. Remove the LCD bezel and mask out the Automatic Light Sensor window (bottom left) with black tape. Then can leave the bios in ALS mode and Windows will retain the brightness settings on reboot or resume.

1. Run SLFM 266x7@0.925V (stock clocked) or 333x6@0.925 (overclocked) using Throttlestop. Use "DCExitTime=5" and "ExitTime=5" in throttlestop.ini so TS exits once it's set this up to reduce monitoring load.

i. Set the Profile 1 to be AC, set profile 4 to be Battery. Set 'start dual-IDA' on Profile 1.

ii. On profile 4 check SuperLFM, EIST, C-state. Set it to 0.925V at x6. Increment up to x6.5, x7, x7.5 testing at each one to see which is stable. Move back one multiplier if unstable. You want the highest stable multiplier at lowest voltage (0.925V) for most battery savings.

iii. On profile 1 set x9.5 at 1.15V. Right-click and choose 'start dual-IDA'. Pull back the voltage one level at a time, check EIST to set it, then right click 'start dual-IDA' to test for stability. Identify the lowest stable dual-IDA voltage for lowest running temps on AC.

2. brightness set to 3 bars

3. wifi adapter configuration: 'power transmit' set to 5. Highest. Ad-Hoc power management set to "maximum savings", disable 802.n if your network doesn't support it. Scanning for the N mode requires more power.

4. Set 4500MHD graphics to "maximum power savings". Ensure have a single display setup. Using the "clone display" adds 1.5-2W extra overhead.

5. BIOS has LAN disabled

6. Disable the audio device in Device Manager. This saves approx 0.4W. Muting the audio device doesn't have the same affect.

7. Disable system paging file to reduce SSD/HDD access.

8. Install the Intel RST driver then apply Add_AHCI_Link_Power_Management_.reg. This will reveal AHCI Link Power Management in Power-> [Plan]->Advanced->Harddisks. I set these options to HIPM+HIDM and 100ms respectively as shown below:

9. Install Flashblock Firefox plugin to prevent flash pages from kicking up power consumption.

10. The wifi adapter raises power consumption after some time of use. The workaround is to disable then re-enable it.

So with this DC profile my system idles at 5.5W (!!) wifi on two bars at idle with a 7200rpm harddisk. This is very impressive. My 52Whr 6-cell is lasting over 5-6hrs as confirmed by batterybar. A power-efficient SSD could extend that to ~7hrs.

Undervolting the CPU using a VID_x pinmod

Quote:

Originally Posted by avlan

Is it possible to undervolt CPU lower than 0.900V (for example by resoldering voltage resistors)? If yes, is it any way profitable?

Yes it is. I showed the VID resistors in the above posts which is what you'd be working with.

If you applied a logic 1 to VID4 then could go b/w 0.7875-0.9V . The thing is you'd want to figure out which voltages are bootup/default/stock and find a way to keep those, but modify some other intermediate values. Eg: remove the VID4 resistors and connect VID4 to VID3. Then look up the VID table to see what the new actual voltage would be when the VIDs b/w 0.925-1.2375 are selected knowing that you are now looking at the VID3 column and picking out the voltages where VID4 = VID3.

My P9400 can run 2.0Ghz@0.925V so has plenty of headroom for undervolting below the SLFM floor of 0.925V if wanting to run it at 1.6Ghz (6x266).

Before you go to that trouble I suggest do the steps above to DC_optimize the E4300. I reduced the idle power consumption from ~7W to 5.5W with a 7200rpm HDD. A SSD should bring that under 5W making it onpar to a LV 2530P.

nando4 · 12-12-2011, 07:50 PM

DIY eGPU implementation: P9400@3.16Ghz 13.3" Dell Latitude E4300 + NVidia GTX560Ti @x1.Opt

Notebook

US$300 13" Dell Latitude E4300 P9400@3.16Ghz
4500MHD 4/8GB DDR3 160GB HDD + DVD
ICH9M expresscard 1.0 slot (2.5GT/s)
Windows 7/64 and Ubuntu 11.10/64

DIY eGPU parts

US$96(US)/US$113 delivered PE4H-EC2C 2.4 OR
US$81(US)/US$93 delivered PE4L-EC2C 1.5+SWEX
US$200 NVidia GTX560Ti 1GB 384 CUDA cores OR
US$130AR NVidia GTX460_ 1GB 336 CUDA cores
US$5 salvaged 12V/17A "550W" ATX PSU

Benchmarks (highest OCed eGPU) 3dmark: 06/vant.gpu/11.gpu=15288/14305/3804, RE5.dx9=97.5, dmcv4.dx10_s4=117.8]

Config

CPU
RAM

GPU

DX9

DX10

DX11

Ports

OS

3dmk6^

RE5
var|fixed

FFXIV

Mafia2^!
1080p

3dmk^
vant.g

dmcv4
scene4

3dmk11
720p

Unigine
Heaven

Dirt2#
1080p

1080p

720p

External LCD

P9400
@3.16
2x2GB

GTX560Ti
x1.Opt

95.9 &|47.8 &

44.1 &48.2

24.5/39.7 &27.3/41.5

Win7
64b

External LCD

P9400
@2.53
2x4GB

81.0 &|38.2 &

43.1 &44.3

24.6/34.2 &26.4/35.0

Internal LCD

! = two back-to-back runs using result from the faster second run
& = 880/2100->1020/2300(GTX560Ti) eGU overclock result capture or text.
# = min/average, London multi-car track with all HIGH except post-process=MED. cmd used "DiRT2.exe benchmark example_benchmark.xml", output saved to Documents/My Games/Dirt 2/Demo/benchmarks
^ = 1280x1024 'internal LCD mode' provided via notebook VGA/DP port to external LCD


US$298 GTX560Ti+ PE4H-EC2C+PSU	Setup 1.x showing system details + GTX560Ti eGPU	Optimus driver setup

Idea reference: DIY eGPU experiences

Software Setup

1. If you have 4GB or more RAM installed then you will always see an error 12 against your eGPU in device manager. This is because the C2D Dell Latitudes' bios set TOLUD=3.5GB where there is no 32-bit pci-e space available to accomodate the 256MB window required by the eGPU. The fix is
a DST override (use the E43100 package) that extends the root bridge into 36-bit space so the eGPU can then occupy 36-bit pci-e space. Shutdown the system.

2. Poweron your eGPU, poweron your system, boot Win7. The GTX560Ti is detected as Standard VGA.

3. Install the 285.62 modded Verde driver. Disable NVidia High Definition sound to maximize video bandwidth.

4. OPTIONAL for eGPU overclocking: Install MSI Afterburner.

How to startup the eGPU

1. Poweron your eGPU, poweron your system, boot Win7. The GTX560Ti will be available with no errors against it in Device Manager. Only issue is the wifi will have an error and not work. The fix is to either:

disable then re-enable the wifi. This can be done manually or by writing a batch file with a devcon command to do it for you.
Use DIY eGPU Setup 1.x and perform a 36-bit PCI compaction on the eGPU. When wanting to use the eGPU do so by booting via Setup 1.x's automated startup via startup.bat bootitem. Your startup.bat will look like shown here.

OPTIONAL: use sleep-resume method. Boot Win7, sleep Win7, attach eGPU via expresscard slot and power it on, resume Win7. If a LCD is attached to the eGPU then make it the Main Display in Display Properties to output accelerated graphics to it. NOTE: this method does not allow Optimus to output accelerated graphics to the internal LCD - the eGPU must be detected on bootup for that. You must set the PCI Reset Delay slider on the PE4L/PE4H to the 0 second (disabled) position for this to work BSOD-free.

2. Use the Safely Remove Hardware And Eject Media systemtray icon when finished using eGPU.

Comments

3dmark06/RE5 both see good performance improvements from a 25% CPU overclock.
I was surprised to find that with stock 880/2100 clocks this GTX560TI can be powered by a basic 12V/17A ATX PSU. Though if trying to overclocking's at 1020/2300 it did require a second 12V/15A PSU to supply power via one GTX560Ti pci-e lead. I don't intend to overclock the GPU during regular use since there aren't any significant real-world gains to be had doing so.
At this point in time I would however recommend a overclocked GTX460 over a GTX560Ti. The GTX560Ti delivers at best 15% more performance but at over 55% price premium. Better to save the money and put it towards a 28nm GTX6xx (Keplar) scheduled for release in the next 3 months.

avlan · 12-14-2011, 11:25 AM

Is it possible to undervolt CPU lower than 0.900V (for example by resoldering voltage resistors)? If yes, is it any way profitable?

nando4 · 12-14-2011, 12:17 PM

Undervolting the CPU

Quote:

Originally Posted by avlan

Is it possible to undervolt CPU lower than 0.900V (for example by resoldering voltage resistors)? If yes, is it any way profitable?

Answered at the bottom of this post.

avlan · 12-18-2011, 07:28 AM

Quote:

Originally Posted by nando4

Spoiler :

I applied most of these to my Arch Linux (which I use mostly), along with laptop-mode-tools. Now I have 6W with WiFi, instead of 12W, idling \o/

tridc · 12-20-2011, 05:57 AM

Could you show your undervolting voltage result in RMClock? Seem my settings is not good, I dont know why I cant decrease more voltage at each stage.
Thanks

nando4 · 12-20-2011, 06:08 AM

Quote:

Originally Posted by tridc

Could you show your undervolting voltage result in RMClock? Seem my settings is not good, I dont know why I cant decrease more voltage at each stage.
Thanks

Throttlestop is imho far better than rmclock. To get 0.925V need to engage the SuperLFM mode. See screenshot below. Means the SuperLFM floor voltage of a P9400 matches a L9xxx CPU as found in a HP 2530P or Lenovo X200s.

You can see I'm running overclocked 6x333 (2.0Ghz) at 0.925V. With your stock-clocked system I'd suggest you try 7x266 (1.86Ghz) or 7.5x266 (2.0Ghz) at 0.925V to see which is the highest CPU frequency you can sustain at the lowest voltage.