'Laptops w. Intel Series 5 chipset can not take full advantage of fast SSDs'

This solution will kill battery life. Disabling the C-states will force the CPU into running at the maximum multiplier, which for most Core i-xxx processors, appears to consume between 25 and 30 watts by the CPU/South Bridge alone. For a 60Whr battery, you can expect around 1.25 to 1.75 hours of battery life, comparable to what you might achieve while gaming.

This assumes that you maintain the Disable Processor Idle state to 'Enabled' for the power policy setting while on battery. Of course, you have the option to only enable this setting while on the AC adapter.

As a side note, for those of you who have computers that share a common heat pipe between the CPU and the GPU, be aware that since the CPU will run hotter, the GPU will necessarily also experience elevated thermal load, regardless of clock settings and whether it remains idle. The heat pipe has finite limited thermal conductivity, and thus if one component yields an increased thermal load, the heat pipe will have less capacity for heat dissipation for the other component.

 

That's good news guys. I'm happy to see this (somewhat) solved.

I was considering getting a new notebook with PM55 chipset. Now at least when running on AC it will give full SSD performance.

 

@KolosoK

It will only affect the power plans that you actually change the 'disable idle state' setting in (after the registry tweak this option appears). So you have 6 power plan setting by default (3 plugged in and 3 on battery) and can select to disable the idle state for whichever ones you want....

@jakeworld

It won't kill battery life unless you set the battery power plan to disable idle state, which IMO would be silly unless you really need the SSD speed improvement while unplugged for some reason.

 

Just make sure you get one with a very good cooling system. As you read 'Jakeworld' is idling at 75 to 80C while I am idling at 65C with the same CPU....

 

Ouch. Many users have already said the Acer 1830T has quite noisy fans.

 

Amazing I'll try this out tonight! My battery currently lasts 2+ hours while playing Starcraft 2 on full screen brightness on Medium-High settings. I'm going to make the change for the High Performance plugged-in plan. Is this what's in the screenshots?

Also, I usually use a cooler under the laptop whenever I'm in high performance, so the heat shouldn't be much an issue for me

 

I kind of thought that would go without saying. I revised my earlier statement to clarify, but sometimes I forget to state the obvious
Also, a reason I idle at a higher temperature has a lot to do with the fact that my computer shares a common cooling solution with the GPU.

 

Yeah I have read that also. The Envy 15 has dual CPU fans and copper heat sinks and pipes designed originally to handle the i7 820QM, It's almost an overkill for the i5 family, I've never seen above 77C while doing a full 100% burn in test overnight and the fans never even came close to maxing out....

 

Thanks for the addtl tests. My temps are roughly the same as yours (though I haven't stressed my CPU much yet)

I now agree that SSDs fly!
+rep to you and ssassen

Here's before and after applying the magic tweak

 

So the conclusion is it can be fixed at the expense of heat and battery life.

What are the hopes for getting a better fix and who should it come from?

 

Is that the score for a Vertex 2 240GB CDM with normal settings?

I didn't know the Vertex 2 is that fast. (39MB/sec 4K random read and 88 write)

 

Yeah, those are awesome speeds! If I saw that I may have gotten it instead of the Corsair Force.

 

Left . = Processor Idle ENABLE
Right = Processor Idle DISABLE

Yes. Its my Vertex 2 240GB on my M15x

 

After the registry tweak all your 'power plan' advanced settings settings will have a new option under processor power managment for 'processor idle disable', just don't disable it in the high performance power plan while on battery and it should not affect battery life, I don't think improved SSD speeds should affect your game play while on battery.... see below:

 

You corsair should get the same or close result, IMO since they have the same ratings


EDIT:
Sorry guys, I think you misunderstood my post

I was referring to the Processor Idle State thing

Left CDM test was ran before setting the Proc idle > disable

 

He cheated I think your Force is just as fast.

 

I would assume that if we can get the same disabled idle state to just 1 core then the power draw and temps should be quite reasonable. I requested this from Ssassen earlier, if he knows how to adjust the registry for individual cores.... I see some options for 'core states' etc. but don't know what they will do when changed...

Also I noticed that 'LOUSYGREATWALLGM' ran his CDM at 100MB not the default settings Don't know how much of a difference this may make in 4K speeds....

 

We'll see after the fix

Also, can someone look into this:

 

err... default = 1000MB?

Thought 100MB because that what we often used to measure the small writes

 

The temperatures I get are very high with the i7 940XM
No matter how much I want those high 4k random reads/writes, I can't keep the CPU temps that high and the fans spinning like crazy

 

yep 1000MB, not sure how much of a difference it makes though...

 

Take a look here Link

 

Those temps aren't bad! My i5 goes up to 87C under load. It's meant to be able to be handle 105C, afaik.

 

Edit, NM....

 

If I get 87 C with 0%-1% load I will easily reach the TDP with 100% load (no reason to push the temps that high and risk my system)

This is what I get with 0%-1% load when I use the default settings


Scores much lower but the temperatures remains within normal limits

 

this is not a good solution.. i think intel should be informed somehow.. we need a chipset driver update..

 

Ouch, I'm sorry, I thought this was under full load! I take back what I said

 

Its a nice 'fix' to boost the performance of the SSD a bit but it cripples the CPU in my opinion. With this 'fix' the clocks are locked so if I run a program or a game that utilises only 1 core (or 2) this program or game will only run on a core that is clocked at a measely 1.6 GHz instead of running at 2.8 (on my i7-720 QM that is). So......... it cripples the programs that don’t utilise 4 cores or the quad core CPU's IMO. Dual cores fare a wee bit better.
Nevertheless Great find Ssassen (",)

 

iemie,

Hmmm... with my i3 (and no possibilty of Turbo), I had no way to see that coming - you're right that could limit overall performance instead of enhancing it.

There must be some regedit 'switch' that will give us the performance on the SSD side without taking the performance from the i5's and higher.

ssassen, hope you can find this one too!

 

just a note, the registry setting by itself doesn't actually do anything, it's only allowing us to see the option to disable idle state under power management.

 

Yes mr literal, we know!


But good point to the people reading/skimming these posts in the future.

 

HOLY CRAP! (Only as frank can say it)

I tried the fix and my temps are in the 70's. This isn't a fix or even a band aid, it's a license to fry my CPU.

This is not going to work for me.

It did raise my numbers from 16mbs to 25mbs on reads, and from 26mbs to 55mbs.

The more I think about it I wonder if this is a problem with how windows handles the ACPI. until we can get some idea of other OS results we don't know for sure.

 

!!!

Guys, stamatisx is right - the idle temps are ridiculous! I'm getting 80C with the idle thing disabled. Also, my performance didn't improve by much! I'm going to try a couple more times and revert back to normal. This is NOT a good solution.

EDIT: Got screenshots for y'all:







Yes, there was some minor improvement. Yes, random 4K writes did double in speed (see original speeds in the original post of this thread). However, it's not worth it, imo.

 

othonda,

lol...

You're surprised that expending more energy increases productivity?

All modern chipsets/cpu's are not inherently better at power management when they're doing work than previous generations - they are better at power management because they can ramp up to full speed and go to 'off' mode faster than yesteryear's parts.

There is no problem as far as I can see - if your temps climbed so high maybe you need to see how to adjust/enable higher fan speeds or even re apply the thermal paste on the cpu.

You do know that your cpu is rated for 100 degrees, right?

See:
Intel® Core? i7-920XM Processor Extreme Edition (8M Cache, 2.00 GHz) with SPEC Code(s) SLBLW

 

just wanted to update with my i7 2010 MacBook Pro's speeds.

Intel G2 160GB inside:



I assume my results are a little lower than normal because my drive is about 80% full, but I don't really know. Also, the Windows partition is relatively small and OSX doesn't yet support TRIM. not sure what the affect of any of that stuff is.

 

till,

I am not surprised, it's just the only place I see the increased productivity is in the consumption of power and the crystalmark benchmark. Burning up 45-50W while sitting idle is nuts on a laptop.

My cooling is actually working in the typical Sager fashion, very nicely. I do plan on upgrading the paste and doing some other mods to lower the temps, but for now it's staying stock. Temps in the 70's are not something I worry about as I have tested using occt and with 8 threads I stabilize at around 85c.

And to your last question, you don't know me very well to have to ask that question. I live, breath and engineer electronics for a living, so I have visited the link you gave me and downloaded the specs, read them and refer to them whenever I need to ever since I was researching my laptop.

 

Well, my U30Jc is night and day - not only in benchmarks, but in real world use.

I'm on battery power right now and switching that setting to disable idle states ate 20% of the power in 10 minutes! Yeah, this setting is not for 'portable performance' I agree. I also monitored my temps with throttlestop and before I switched, the temps were cpu: 39, gpu: 50. After switching to disabled idle states the cpu: 60, gpu: 64.

But, what a difference in performance:

LR3.2 flew, as did simply opening/launching QuickTime. My accounting software SA was on steriods and nothing seems to slow the computer down at all.

Will I use this on battery power? Not if I have a choice I won't. But plugged in and sitting on a Zalman notebook cooler, why wouldn't I?

I will do some comparisons in a few days, but I'm confident that real world productivity is above 20% with this 'tweak'.

What is most impressive to me is how fast the O/S responds. With this setting off (enable idle states selected), the computer feels like it is working through molasses to me (compared to a mechanical HD, let alone with the idle state disabled and the Inferno inside).

The best part about this is we all have a choice right now if we want to use it or not.

I'm surprised that it doesn't work for you (performance wise, real world), hope a universal fix is provided to everyone soon.

Even better would be more registry tweaks that allow us to control more precise options inside the power management tabs. Then maybe we will have the best of all worlds.

(I'm just thinking of the Tom's Hardware article posted by Phil a few pages back that showed that while disabling ALL power saving options gave the best SSD performance, enabling some of them only affected performance by a few %).

 

Thanks for doing that. It looks like the MBP is affected too. I don't think 80% full should have such a big impact.

Is there any way you can make your Windows partition less full?

The next question would be if OS X is impacted too. This could be measured by running X Bench without CPU load, and then comparing it with X Bench while at the same time a program is stressing the CPU for 100%.

 

I have a feeling we are going to be using another chipset by the time someone finds a solution.

 

Haha
this trick made my laptop the world fastest i7 620M machine in both multi and dual core coresmark results and went from 627th to 154th and 26th to 19th rank respectively
max benchmark temp 75C thanks to Ic diamond

 

Its nice to know what the problem is, but i wonder if well ever see a real solution to bring back full speeds and at the same time not make our CPUs perform the task of a central heating system (and i cant imagine what exactly would have to be updated - the chipset drivers, registry/power settings or bios?), since letting my i7-820qm idle at 64C is just ridiculous.

Ive written to Intel email support and i suggest that anyone who hasnt done the same yet to do so, since we have more chances of getting something done if more people report this. I mean both the chipset and the SSD im using are Intel products so id expect them to solve a problem between two of their own products... Ive also alerted Asus about this but i dont expect much from them considering they didnt reply any of my previous emails about GSODs and missing key strokes.

 

Good idea, everybody email Intel, also if you don't have an Intel SSD. After all it seems to be a chipset problem.

The more attention this problem will get, the sooner it might be solved.

I've posted it on OCZ forums too. I'm suprised none of the techsites like Techreport or Anand have picked this up yet. Does anyone have Anand's email? I can't find it on his site.

 

Yeah, I'm loving the speed up too!

Just a little update: two days ago (pre-tweak) a full MSE scan took just under 2:45 HH:MM to complete on 2.5 million items and that is leaving the computer alone to finish as fast as possible. (If I continued to use it, it would easily be over 3 hrs).

Today, I updated the virus defninitions as usual and switched idle states off - the scan completed in 1:37 HH:MM and this is while I used the notebook normally (email, this forum, banking, Excel spreadsheets, accounting software, 'playing' with images (no conversions, simply viewing RAW files in LR3.2, etc.).

This kind of performance difference is what I expected to see a year ago when I got my first experience with an SSD - I was told I was crazy to think mechanical HD's were superior - but, I know when something doesn't 'feel' right!



Anyway, this is amazing. I'm truly happy with the Inferno now and am heading over to Intel to join my voice to any and all others who will ask them to give us the performance and power saving features that we have been paying for a very long time.

I have a feeling that this is an SSD's sensitivity to timing issues and am expecting a fix not only from Intel, but also a joint effort on the part of MS too - at least for Windows users anyway.

 

gentlemen...first off i must ask.

You do know that CDM and AS-SSD write large amounts of data to your drives each time you run them and as such dramatically reduce the life of the drive?

General Discussion AS SSD Benchmark read from there,

main thread... General Discussion AS SSD Benchmark

most newer ssd now use nand with 5K P/Ec, running such benchmarks is just stripping life from your drives.

Why do you not this bench we are using on the OCZ forum.... Guide Random and Sequential Testing

It uses winsat and writes MB not GB worth of data to the drive and will still give you comparable results.

Last, you made a good catch here, Apple have been doing the same for years with their notebooks as anyone with a SF based drive will tell you...install it in a apple then test it in a PC and see a massive speed difference. All in the name of battery life and reducing heat.

 

ocztony,

Yeah, I have run CDM only a handful of times (and only in the last couple of weeks too).

Thanks for the link!

I was just thinking of searching the SSD Thread for the manual way of running the winsat benchmarks.

BTW, are you the 'Tony' in OCZ land? If you are... do you have a link or any idea when the SSD Toolbox will be released for Vertex 2's?

Thanking you in advance!

 

@ocztony

Sorry, but I am so tired of hearing people claim that your ruining your SSD's lifespan by running benchmarks etc. It is just not true, at least with intel drives. I don't have time to look up the exact info again right now, but with the rated R/W cycle life for intel x=X25/18M series it would take soemething like 13 years of daily 100% read / write cycles (fill and empty the whole drive) before you start to see any true wear. Get real! with wear leveling and TRIM any modern SSD is going to outlast any computer that you may currently be using. What is the oldest computer your currently using?? how about the oldest drives in any of your systems? By the time you even coming close to wearing out an SSD the thing will be obsolete anyway.....

Edit: I just checked my intel toolbox 'SMART' data on my primary SSD. It's 7 months old, has a total amount of writes of 4.93TB and the wear level is 0, or normalized level (life remaining) of 99 which is the same as when it was brand new.... Have probably run at least 200 CDM runs and other benchmarks and still have no wear indications. Again, no issue here

 

JJB,

while the theory supports your feelings, SSD's have not been in use for 13+ years for us to 'know' for a fact if indeed this is true or not.

I too read extensively about the Intel SSD's and was thoroughly impressed. However, I also remember reading that as each nand cell gets used it gets weaker and weaker and the 'theory' part gets thrown out the window.

And even if Intel has a handle on wear levelling doesn't mean that all other SSD's will too.

And you're right that any current SSD will be obsolete for enthusiasts a few years down the road, but that doesn't mean people won't still want to use them (if they still work and are reliable).

For example, I have more than a few HD's from before 2003 (Hitachi's) that are still serving a purpose - the computers may be gone and dead that they were first used in, but as long as the HD spins and let me access the data - I'll continue to use them.

Theory and MTBF numbers are all well and good - but it in no way replaces actual use over time. That, we have yet to see from any SSD.

 

the issue with SF drives is this. The controller relies upon it being able to compress data to keep its write amplification low...when you write to the drive with incompressible data you write without compression, this does eat into write cycles.

Intel/Micron nand has 5K P/Ec, we use it, so do Intel in the G2...this means an Intel drive will wear as fast as a SF drive.

Last, do not be foolish, any write to nand and then erase eats into write cycles, Intel use 7% OP, that is not a lot.

All i want is you guys to have fun and get at least 3 yrs from your drives, the more you benchmark needlessly the closer to not getting 3 yrs you will get.
It really is as simple as that...and remember nand burn is not covered under warranty, your habits are what is dictating the life of the drive, no manufacturer has any control over this.

 

we are testing a release candidate toolbox now...how soon it will come out I have no idea

 

@ocztony

Well as I said above, I am not worried in the least about wearing out my SSD's. 7 months and 4.93TB of writes with less than 1% wear according to the 'intel toolbox'.... Not going to be an issue as far as I am concerned and have no plans to limit my use of the SSD's in order to extend their 'life', at this rate I'll be to old to care before the performance degrades enough to cause problems.

If OCZ is truly concerned about thier drives starting to fail after 3 years then I can't see OCZ being around in 4 years.... Or any other drive manufacturers who have failures after 3 years....