Envy 15 2nd Gen Benchmarks

Just to confirm, all my posted 3D benchmark scores were run at the standard default resolutions. While the posted 1366 x 768 res scores are good they cannot even be posted as results on the Futuremark website do to the lower than standardized resolution, essentially with 1280 x 1024 you are processing ~30% more image data than at 1366 x 768.

@Moviemarketing

Of course your CPU scores will be higher in some cases since you have an i7 820QM which your comparing to a i5 540M...... The trial version does not make any difference as it is designed to run at the standardized resolutions and settings for the benchmark, there is really no reason to run at any other settings unless you want to see how your specific machine performs at different resolutions / settings, these runs would not be comparable 'benchmarks' for performance though...

It appears from your scores that you are not overclocking your GPU at all, with moderate OC'ing and the 820 you should easily be able to break the 10,000 barrier in 3Dmark06...

 

Don't be too sure, I lost the 10000 advantage when running on standard resolution.

 

I figured you would, you'll need to be at about 620 / 1120 OC's to hit 10K, my max runs were at ~633/1133. FYI, upping the memory speed will have a much greater impact on the scores than upping the core speeds...

I think you misread the bottom half of my earlier post. That was directed at 'moviemarketing' who has an i7 820QM CPU which should bump up the default scores at the same OC's by ~200 - 300 pts due to the CPU difference, I've seen a couple of 10,500+ scores from i7 Quad CPU's Envy's...

 

I will try that later tonight, by lowering my Core and upping my Memory. I just re-ran the test and got 9885 after closing some useless processes.

 

That should help, I always do a restart, close / turn off all uneeded apps & anti virus and then turn off wifi, B/T and mute the audio before a Max OC attempt. My thinking on the last 3 items is that they do require some power that may, however slightly, lower the bus voltage available to the GPU/memory...

Good luck...

 

Hahahaha. Mute the audio to ensure available power for overclocking? While far-fetched, it is logical... but a little extreme! I have no interest in an OC that's not stable for long term gaming/other use...

Ah, so that's how you get the 10K so easily. The Quad.

Sigh. I am temped to try an i7-620M. I already bought an QS/ES i5-540M.

Anyone interested in this i5-540M if I sell? (QS/ES) I will be doing full compatibility (Envy 15), stability, feature, temperature and power testing to ensure this QS/ES model is not inferior to OEM in any way. . .

 

I have the 540M. The i7 Quads can add ~200-300 pts on top of my scores in 3D benchmarks due to the higher CPU scores.

From what I've read the 620M is very inefficient power wise, especially the early ES versions. The production models use over 10W more at idle than any of the i5's, so if you want decent battery life I would recommend against the 620M. Here is a link showing the differences in performance and power usage, scroll down to the last graph and click on it to enlarge it; Notebookcheck: Review Intel Core i3/i5/i7 Processors ?Arrandale?

For gaming I doubt you would see any difference in performance with the 620M compared to the 540M since the GPU / GDDR3 is going to be the limiting factor...

 

Yes, I've read that, the only single comparsion available on the internet. There are alot of people who dispute these results (well some in the Sony forum at least) and I'm particularly interested in verifying them myself. I have a kill-a-watt, the skills, the platform and the 540M. If true, then I'd certaintly keep the 540M over the 620M. Anyone want to help out by offering to take my 540M / or 620M or other financial support?

My bad on your chip. Great score with the 540M.

 

You may not need all that to test the power draw. The newest 'Everest ultimate edition' (by lavalys) app seems to have a very accurate reading of the CPU Wattage which can be set to update as often as every 1 sec. For reference my 540 seems to idle at 5.43W and at ~4.5W on battery.

I was leary of actually paying for the newest version but now that I have it I think it was worth every penny, very good app with all sorts of cool features and info on my system that I didn't know was available, plus a great burn in / full load testing function.

EDIT: here is the link to download it, IIRC there was a free 30 day trial 'coupon code' (but I can't find the link) http://www.lavalys.com/support/downloads

 

Well, I will give your advice a shot, but here is what I got by luck! Let me cool this down a bit and give it a super go in a few with your tidbits.

 

I'm gonna try a run in the near future with the laptop in front of my air conditioner, since I'm in hawaii and my ambient temps are higher than mostly everyone else.

Air intake vents are in the front or bottom?

 

Both the 2 large ones in front and the 2 on the bottom which are right below the fans. When doing a 100% burn in test I noticed a slight drop in GPU temps (~2-3C) when I propped up the back edge with a book....

 

So is it unstable at higher speeds, because 70C is not THAT hot..? (I'm just asking things, I know it's insane to go higher )

Is there already a run with entry, high and extreme presets? (vantage)

Someone already run the PCMark05 battery test or MobileMark 2007 ? (min)

 

check the first few pages for benchmark listings....

 

When I get my replacement envy I can take some cpu power draw measurements for you. It's an ES 620m.

 

let me know what you think of the 620. Almost bought that instead of my 820

 

Is there a particular "moderate" OC guide you would recommend?

 

just use the amd clock tool and increase speeds very slowly. i ran 3dmark06 and vantage in between clock changes. if you start seeing any tearing or artifacting, dial down your clock speeds. As a safety precaution, don't overclock beyond that point (in fact I wouldn't recommend overclocking at that setting again). Keep in mind most cards have to be seriously abused somehow to be damaged by overclocking (or defective). The PC has a safety mechanism built in to shut off in the case of too high temps.

Also, you get more bang for your buck with overclocking the memory. For most users I believe, their 5830 easily overclocks a couple hundred points without any real stress (from 800-1000)

 

I installed this one
AMD GPU Clock Tool v0.9.8
techPowerUp! :: Download AMD GPU Clock Tool v0.9.8

It says error - no valid devices found.

 

This seems to be the only one that works: AMD GPU Clock Tool v0.9.26.0 For HD 5870 download from Guru3D.com

IMHO, you really don't need to worry about hurting the GPU by OC'ing to much, it has several 'safety' features to prevvent damage, the worst that will happen is a 'blackscreen' shutdown mode that you will have to push and hold the power button down and then restart the computer.

I would start at 550/900 and then work up in 10 or 25 Mhz step until you see artifacts or the temps get above 80C... My max stable clocks are 633/1133 which seem to be above average.

Good luck

 

That is extremely scary to think about. That happened to me 10-15 times per day with my first Envy 15, no overclocking whatsoever. I returned the unit for refund and ordered a replacement which [knock on wood...erm, I mean, magnesium alloy] has not had that issue so far.

 

i get blackscreens when i overclock to about 560/1100

even with an air conditioning unit on high fan at like 68F pointed at the intake grills

 

Do you use a laptop cooler? I use the Zalman ZM-NC1000

So to make sure I understand correctly,

(1) I go to the box labeled "Engine Clock" which reads "499.97" and I should increase it to 550; and

(2) The box labeled "Memory Clock" which reads 799.87 I should increase to 900

(3) then hit "OK" or "Set Clocks"?

Then run the PCMark06 again and see if there is significant improvement?

How about if I'm working without AC and I want to increase battery life, can I adjust the settings in a way to "undervolt" or "underclock" that will make the battery last longer? Is this safe?

 

This 'blackscreen' is not the same thing you were experiencing on the otherr machine. It is what happens if you 'crash' the GPU form OCing, it fully recovers and does not cause any continuing balckscreen problems once you lower the clocks back down.

Sorry to hear that, sounds like you got an 'average' GPU but above average memory speeds, have you tried bumping up the memory past 1100 and leaving the core at ~550? The memory speed makes the most difference in performance as that is the bottleneck of the graphics capability...

Any temp issues when it crashes? Unless I do a max OC furmark run, my GPU core temps never get above 69C during the benchmark runs @633/1133

 

@moviemarketing

You set the clocks as you described then need to use 'set clocks', then leave the app open (or minimized) while using it. You can also use the temp tab and activate the GPU temp sensor readings - play around with the controls until you get a sensor to list in the top 'box', there are 3 that work but I forget how to activate them the 1st time (my envy's not here ). When your done make sure to use the 'restore default clocks' on the main page.

I Have been able to gain ~20+ minutes battery life by using lower clock settings when on battery, IIRC 60/120 was stable while 50/100 works but had issues with playing flash or SD wmp video. With an 820QM I don't expect you'll get near as much improvement since your CPU uses so much more power than the 540M

EDIT: FYI, The 5830 is a 'dynamic' GPU that adjusts clock speed as needed based on graphics demand (as long as you have 'power play' enabled in CCC). The standard clocks (on AC) vary from 199/299 to 499/799, when on battery the range is 99/199 to 299/299. When you use the clock tool it overrides this feature and stays at your settings until changed or you restore defaults or close the program.

 

Looks like I got defeated at 621/1133. Anything higher and the Envy will black screen and crash on the third graphics test. Here is what I got on my last run last night. The temperatures got hot as I didn't run it with a cooling pad and the room was very stuffy hot (considering I have been running the benchmark a few times prior to this one).

Most likely, I will stick gaming to 600/1100 maximum. I don't feel comfortable running any higher than that in the long run.

 

Well, like always though if you OC it could lead to your vid card frying Even when it's not a permanent OC.

Speaking from personal exp.

 

It is not a video 'card', it is a single die (silicon chip) GPU just like a CPU, It has 4 temp sensors, fan speed sensor, overvoltage limits and firmware and software controlled shutdown features if anything gets out of spec. In other words the GPU will not let you actually 'fry it' without protecting itself first. So IMO OCing will never allow you to damage it, shutdown will occur first.

 

JJB is right, the card will failsafe before it fries, but Sneis is also right that repeatedly failsafe will cause the solders to disintegrate in the long run. At least most of my gaming is done at 550/1100 (Haven't touched a game that needed more juice), and at 550/1100, the temps are around the mid 60's, so it is all good. I only juice it to the max for occasional benchmarking and stability tests or if I really need to run Crysis (which I don't). It is nice to have an Envy that can surpass 600/1100 on OC-ing, but as a safety thing and general practicality, I don't need to OC that high to play what I need to play in the long run.

 

Sorry but the 'solder' will not desintegrate with repeated shutdowns, don't know were you ever heard that one. Typical solder melting points for electrical use is 220C - 280C and the solder used on the actual terminations to the die are high temp variaties with 300+C melting points. The 5830's max allowable temp is 100C so I'm not really worried about melting solder joints....

 

In that case, I should run my games on a higher OC then, hmmmm...

 

Just want to touch on this. The solder joints may fail and are under risk for failing exactly for high temps, repeated shutdowns, etc. It is not about the melting point, but because temps may range from 60°F to 200°F, at a significant % of solder's melting point, the temperature extremes result in creep and stress variations which can lead to fatigue and eventual cracking of the soldier joints. This has happened with, in particular, Nvidia graphics in laptops recently, but should be avoided with proper material selection and soldering techniques. It is probably a non issue with this laptop.

 

Alright, well the reason why I brought this up:

My Envy 15 was artifacting 100% of the time; part of the reason why I ended up RMA'ing it.

Bear in mind, I overclocked only to run 3dmark2006 and ran stock otherwise. I didn't even get the chance to legitimately game with the laptop.

My temps were always as expected, nothing out of the ordinary. I also confirmed that the GPU cooler was making proper contact and tried swapping the CPU's.

Ram is usually the culprit in artifacting in my experience, bad chips I guess.

 

Sorry but with todays solder compounds, process and material technology the problems you describe with creep and stress (actually called the thermo coefficient of expansion) have not been issues since the early surface mount technology was developed 20+ years ago. If Nvidia had this type of problem recently it had to be do to a bad lot of solder or a quality control problem with the process. I am not saying it can't happen, just that if it does it's not due to the actual solder, it's the quality control of the materials or process.

 

Ok I finally figured out Everest.

For my 620m at idle I seem to be between 4.65W to 5.25W on both battery and plugged in. Gotta figure out the load draw.

 

DirectComputer/OpenCl with Catalyst 10.4 Drivers and Stream SDK 2.1


Default ATI 5830 500/800
==========================================================
ATI Mobility Radeon HD 5800 Series @ 500 MHz (1002 / 68A1 / 1522103C)
Intel(R) Core(TM) i7 CPU Q 720 @ 1.60GHz (8 logical CPUs)

Benchmark version: v0.45
DirectCompute: D2269.4
OpenCL: C1193.9
CPU: M31.6

AMD multi-vendor Miniport Driver
amdkmdap 8.14.01.6105
Windows 7 x64 Enterprise Edition (build 7600)

========================

A slight increase in computational improvement speed

 

Ok,here we go:
This was from a little over 20 minutes on the stability test in Everest, CPU only.

 

Alrighty, last post for the night I swear. This screenshot shows a single run in 3dmark06.

Everest measures the cpu temps and cpu voltage.

Set gpu clock tool to overclock to 600/1100 and set to measure gpu temps. Should have more headroom but since this is Envy#2 now I didn't feel like pushing my luck. You can also make out in 3dmark that the cpu is indeed the 620m.

For those that care this is BIOS version f.26.

 

@Mrsneis

Hmmm, your GPU seems to be running quite warm under load. Mine maxed out at 61.5C at 625/1125 in 3D06 (maybe the f .26 BIOS? I was on f.24) and my fans never really ramped up to near max speed. Maybe try running the Furmark burn in test with the same OC (will run warmer) and it will tell you if might have heatsink or some other issue. For reference IIRC my GPU after a full hour of Furmark @600/1100 stayed at 77C, this was while running the everest CPU stability test with the 1st 3 boxes checked at the same time.

In the everst stability test you need to check the stress FPU, Cache and memory to get the full power draw as these all are powered from the CPU. You should also see a drop in temps doing this since it take some load off the cores.

Good to see the low idle wattage, that is a far cry from the 10+W extra they got in the notebookcheck test comparison.

 

I saw that another poster claimed he was getting significantly higher temps with f.26 and f.18 compared to f.24 ~ problem being however that with f.24 it's taken down due to your call ><

 

Here's take 2 using f.24. The first one was everest stability test with CPU, FPU, Memory, and Cache ~ more intense than my last run.

The second one is the same run in 3dmark06 @ 600/1100.

Looks like my temps are pretty consistent and that voltage draw isn't as extreme as notebookcheck made them out to be. For the GPU temp though, I'm not incredibly worried about the temps but the question lingers, what if I re-apply the TIM? In my experience it would at best lead to a few degrees shaven, unless something is totally wrong; but in that case I would be artifacting and crashing all over the place anyways.

Also, I'm using f.24 is this different than f.24a?

 

Got 10761 @ 790/1120 video overclock. I am running an ES/QS i5-540M from ebay that runs great.

I will be comparing it shortly to a retail i5-540M and the i5-430M. Preliminary results actually put this ES i5-540M at LOWER power consumption then the stock i5-430M.

Looks like 11K 3DMARK06 should be possible here with a quad, or even a 620M

interesting, my system will do 840 core in furmark. only will do approx 800 core in 3dmark06.

mine is begging for a mild VMEM volt mod and ram coolers...

 

@MsSneis

The numbers look reasonable, your well under anything close to maximum allowable temp ranges. Now that I think about it the numbers I posted with everest and furmark were run some time ago and i can't recall which BIOS I was on then (maybe f .18?). I want to rerun the stability tests with f .24 and compare to your numbers but it will have to wait until I am back home next week, I only have the 65W adapter with me which won't give full CPU performance.

Not sure about the 'a' in the F .24 BIOS, everest shows mine as "F.24" in the 'summary' list in the 'DMI' section at the bottom of the list, is this what yours shows? or does it have the 'a'?

 

Wow you have a very 'special' 5830, that is by far the highest running core speed posted here. Please confirm what resolution you ran the 3Dmark06 at, in theory there is no way you should have gotten 10,700+ with only 1120 memory speed at the default 1280 x 1024 resolution. The memory speed is the bottleneck for performance on the 5830 not the core speeds, in other words. at 790/1120 and 1280 x 1024 res it should have crashed because the memory would not be able to keep up with the core data.... Do you have the 1366 x 768 res display?

The 540M and 520M have been documented by several reviews to use less power, at idle and full load, than the 430M, have not seen any explaination as to why this is.

 

I have the 1920x1080 glossy of course. the benchmark was run at default 1280x1024

The core overclock added some performance definitely. Picked up a few hundred 3dmarks, not much compared to what you get from the mem overclock, but there's some there.

 

Actually you picked up more than 600 points compared to the last 625/1125 run I saved the results from, great score but it still baffles me that the memory is able to keep up with that high of core speed.

 

kicked the clocks up a notch, but I did spot a few small artifacts.

I also set the ATI 3D settings back to default. TIP: set the MIPMAP settings to high performance for a few 3D marks. I was running this for the 10700 run...

10823 3DMARK06 this time. Would be ~10900 with the MIPMAP setting back where I had at at high performance.

I actually run 4x anisotropic filtering, adaptive AA and high quality MIPMAP for gaming. These settings take off more "3DMARKS"

my gaming clocks it looks like will be a stable 750/1100.

So who wants to let me borrow their 920XM for I can get that 11500 run in? Mabye 12000 if I voltmod the memory and add cooling.

Give me a reference to run mine at to see what the result will be. I'll do a run at 625/1125 and post. We'll make sure theres not something else going on raising my scores.

 

625/1125 to try and match your run:

10179

so going from

625/1125 to 790/1125
10179 to 10823

core clock change 26.4%
3dmark score change 6.32%

My results are in line with yours. 650 point gain from GPU speed 625-->795.

That actually only qualifies as a partially "bottlenecked" GPU, 6.3% is not insignificant... I'll take it. Consider a large portion of the benchmark is also CPU and platform bound as well.

500 stock --> 790 stable 58% overclock! lol.
800->1125 not bad either, 40.6%

Anyways, these things are not so black and white. The structure in which and the amount of data needed to process a frame is fantastically complex. Even when bottlenecked, there will always be gains from overclocking the bottlenecked element. In this case, the core can make more efficient use of the of the available memory bandwidth thru faster instruction times (thus it will request new data from the memory "sooner") and reduced effective latencies to memory access (faster clocked memory controller--> less wait for a "sync" with the memory.)

I think i'll run at 750/1115 for gaming unless I run into a problem. Seems stable...

 

Well you got very lucky with the GPU, bet it may be a 5850 graded part labeled as a 5830 due to demand.

My 625/1125 run scored 10,144 pts. so yours seems to match at those clocks.

I just compared my GPUZ readings at my max (almost stable) clocks of 650/1133 to what you posted. One very odd discrepancy jumped out at me, look at the Bus Interface, I show x16 @ x4 and yours shows x16 @ x8 , The only other difference seems to be that I am running the 10.5 driver which should not effect Bus Interface AFAIK. Any ideas??



View attachment 49836

 

can we have a run with vantage?