*OFFICIAL* Alienware M18xR1/R2/18 Benchmark Thread - Part 3

This the right program that informs temperatures among others which would it be?

 

I am not sure I understand your question, my friend. Obvious, the one on the upper left is GPU-Z. What you see in the upper right is a Sensor Panel for AIDA64. What I primarily use for temperature monitoring is HWiNFO64. I output the information to on-screen display using RTSS v6.0.0.6228.

 

Thats on a single adapter?

 

THANKS (aida64 sensor)

 

Yes, it is. But, with manual control of the CPU using ThrottleStop profiles. The nice thing about the dual setup is not having the manage anything... just let 'er rip (within reason, of course).

This is also with a single AC adapter, and no need to toggle CPU speeds... just running wide open on CPU and GPU. This is why for the average gamer and once-in-a-while bencher there is no real need for a dual AC adapter setup. It's only needed for number-chasing.



This is number chasing with a dual AC adapter...

 

`

Yeah, thats what I been thinking to, dual setup really only needed if you go hardcore OC and then testing the system with program (typical benchmark programs) that really can load everything 100% at once, a state (which we will probably NEVER see in real life gaming (but I sure which we could).

 

Got my 3720XM today! Installed it and my triple pipe heat sink. I did a round of benches to see how it stacks up against the stock 3720 and it came out worse.
but then I fired up xtu and trixx and managed this with no extra voltage
http://imageshack.com/a/img34/4844/opae.png
http://imageshack.com/a/img203/7370/cve9.png
http://imageshack.com/a/img513/3923/3r4g.png

Not sure how much I can push it even with extra voltage, I guess I'll find out.... tomorrow

 

4930MX wPrime 32M and 1024M @ 4.5GHz

 

Compare that wPrime 32M and 1024M scores to my 2920XM's best... (albeit at higher clock speeds)



...and, the 32M and 1024M runs I just did at a matching 4.5GHz with the 3920XM... 4930MX got pretty darned close, don't ya think?

 

The fact that we settle for pretty darned close is a testament to just how terrible Haswell is. Clock for clock more efficient my butt. (just noticed you're running 1600 with 4930MX but 2133 with 3920XM, so OK this might not be a completely fair comparison)

 

That's kind of my point about Haswell not being more efficient when I made that facetious comment... "Pretty close" should apply to 3920XM being "almost as good" as 4930MX... this is totally backwards. While the wPrime scores are still totally respectable, it should be better than last generation. The memory speed doesn't really have an appreciable effect on wPrime scores as far as I can tell. Notice that the 2920XM is also 1600. I can swap the memory out between them if there needs to be that level of comparison. It think the best we might hope to see from that exercise is a reversal of the "barely better" scenario if we're lucky.

I'm thinking about trying to reprogram the stock 1600 RAM to 1866 with Thaiphoon Burner. I can flash it using the M18xR2 and drop it into the 18 one stick at a time to verify it is bootable on the 18 without having to worry about if it's not... just flash it back to the stock firmware using the R2 if it's not.

 

Nah, I think you're spot on with the "barely better" scenario at best.

So far I've seen no indication that the 4930MX is performing anywhere near what's expected of it. Although given the power-handling limit that Clevos run into as well, I'm starting to think more and more that the mobo must have something to do with it. Probably something along the lines of nobody ever expected the 4930MX would be capable of drawing 140+W, so the mobo simply wasn't designed to handle that much power. (cheap MOSFETs?)

Not entirely unreasonable given your 3920XM barely draws 100W at 4.5 GHz, and all the BS "Haswell = efficiency" propaganda that Intel was spreading prior to Haswell's actual release.

 

Ok I have lots of questions so I will number them

1. What is the conversion for flex to mV or is it more complicated then that?
2. Out of curiosity what is the CPU throttling gauge in XTU and how is the throttling remedied, (increasing the turbo boost power?) its always at 0% for me so its not a problem.
3. What is reasonable flex/voltage to increase the multiplier? (At the moment I have added 133mV to achieve 4.3Ghz from 3.9Ghz)
Edit: make that 160mV @ 4.3 to pass vantage cpu tests
4. Throttlestop is acting strange, the turn on/off button doesn't really seem to work and it wont hold maximum frequency when running the 32M and 1024M benchmarks, it starts at 4.3 and but ends at 3.5 (should I post some screen shots of my settings?)
5. XTU is much better at holding the max frequency but still has goes up and down by what seems to be 1 or 2 multipliers is that normal?
6. The reference clock in XTU, is that worth playing with? and what the best way to find the sweet spot for it? Do I find the highest reference clock that is stable and them increase multipliers or increase multipliers first and play with the reference clock to fine tune the total frequency?

 

hope that broadwell next year will do good, doubt it though

 

the fastest of fastest SSDs, nope not samsung 840 pro

SanDisk Extreme II 240GB 6-Drive SSD RAID Report - Secondary Volume Benchmarks - PCMark 8 Extended

imo they should do neutron GTX cause i recall corsair had a fairly powerful SSD, on par with OCZ's vector in writes.

 

'
I like my ocz vector, with or without raid 0. They are better than samsung 840 (pro) :laugh:

 

Hopefully this will give you something to sink your teeth into as your are tuning your new toy.

 

1 Flex in the BIOS is the same as 3.90625mV, or 1 click in XTU "Additional Turbo Voltage" - watch the numbers next to the slider in XTU when you click one click at a time.
1. Ok sweet that's easy
2. Don't burn too many calories on that. XTU benchmark doesn't easily translate into something meaningful in real-world scenarios. In XTU that is a thermal throttling measurement. You don't need XTU to measure that because it's obvious by looking at the core clocks when the CPU gets hot. That's kind of like checking air pressure on a flat tire, LOL. Unless induced by high temperatures (thermal throttling), "Power Throttling" (erratic clocks) is often because of the MSR lock being enabled in the BIOS or Pri Plane (XTU "Core Current Limit") needing to be increased, or the Turbo Power Limits being set too low. The "throttling gauge" in XTU does not reflect power throttling, but again that's obvious on face value and does not need to be measured.
2. Kinda makes you wonder why they add stuff like that, it just clutters the place up
3. What is reasonable is whatever it takes. There is no one size fits all setting. It varies by CPU. Not enough you will get freezes and/or 0x124 STOP errors (BSOD) and too much will cause more heat and potentially throttling from that. Use as much as needed, but no more than necessary. Trial and error is required to determine what your CPU needs.
3. I was just trying to put the voltage I have added in perspective of what is considered alot/little.... Cpu usually uses 1.XX volts and we can add about another volt with XTU or Throttlestop. I guess to much much voltage is never a problem if you only add what is needed as temps will always be the limiting factor, like you said... unless you start using dry ice
4. See information above about settings in general... will probably solve your problem with ThrottleStop, too. Jut be aware that you need to set limits for TRL and TPL values in ThrottleStop equal to or great than your BIOS/XTU settings or enabling ThrottleStop will undo all of them. ThrottleStop is phenomenal if used properly, but you have to use the correct settings to realize a benefit.
4. I will go over the throttlestop thread again and make sure im not doing anything silly
5. It's normal with XTU benchmarking or stress tests because the loads do odd things with the CPU. It's also expected with the wrong power settings, or misusing ThrottleStop, but again, forget about XTU benchmark or LinPack benchmarks because they are more or less irrelevant and have little practical application to real-world results in gaming or benching.
5. I had used XTU benchmark to try and find an initial stable-ish setting to start running the 3dmark benchs mainly because I am still trying to get my head around Throttlestop, but I had to add way more voltage to pass the vantage cpu tests then I had to, to pass the XTU bench
6. Depends on what you are doing, and what kind of CPU and chipset design. Sandy and Ivy don't like it so much, but AMD and older Intel CPUs handle it a lot better than using a higher core multiplier. It can be useful, but it can also reduce stability. It also changes clock speeds of memory, PCI/PCIe lanes, audio bus, and other components besides the CPU that often balk at being overclocked. Save this tweaking for last. Once you get everything else optimized you can bump that up to find out what's stable. This varies by system and CPU as well, but around 104.00MHz and beyond is usually where system stability starts to noticeably degrade and begins to cause issues for Sandy and Ivy Bridge. Where this is most useful is achieving higher clock speeds with a lower multiplier, or eeking out a few extra MHz from a locked CPU that really isn't well suited for overclocking. It often causes instability in things other than CPU benchmarks. Games and GPU benchmarks can get super flaky and unstable with too much BCLK, and so can multimedia apps and media players. If you're running more than 101.00MHz BCLK and start having new BSOD, freezes or app crashes, even odd buzzing or humming noises or audio distortion from your speakers with things like YouTube, Windows Media Player and iTunes, and 3D games, that you never had before, chances are you got a little too carried away with your BCLK.
6. I have a very basic understanding of the BCLK but that makes alot of sense, its usefulness is basically limited because you cant increase it to much on the Sandys/Ivys and because its affects alot of other things which we dont have cooling/voltage control over in a laptop

Thanks again Mr. Fox, I will keep playing

 

The nice thing about ThrottleStop is that it is so simple to use once you know what it does. These might be helpful links. One is from my YouTube channel and the other one of my Imgur account.

https://www.youtube.com/watch?v=xb2HPbk_O1I

http://i.imgur.com/ZPMW6Wn.jpg

Substitute your owns setting for CPU power (voltage/amps/watss) since the screen shot is show 4930MX, but the ThrottleStop program settings (Options button) are the same. Should give you an idea of how to use the profiles, too. Just remember to use the same power settings as what you determine works best using BIOS or XTU adjustments and plug those values into ThrottleStop.

 

SuperPI 1M Score @ 4.7GHz

 

Turns out all my throttlestop setting were fine, but both the MSR lock and the power setting in the bios needed to be diabled/raised
Just passed vantage cpu tests @ 4.3Ghz for a score of 29.7K, I just need to find some cooler air now... 45 flex is getting me up to 98C

 

Can you manage less flex without Bsod? 45 is alot for 4.3

 

Awesome that you found the BIOS settings that needed to be tweak.

Whoa, yeah, I agree with Brother TBoneSan... 45 is way too high on the flex. It you truly need it that high it might be due to the prototype CPU, but 3920XM runs 4.3GHz at 10 to 18 flex all day long with very cool temps. I only use 38 flex for 4.5GHz. If you can lower it to 15 or 20 at 4.3GHz and still run stable doing so make fix your high temperatures.

 

Yeh your right 45 was a bit much much 38 and 39 will pass vantage cpu on the first run but not the second? No bsod just vantage crashes, temp are better tho hi 80's low 90's.
Edit: A full vantage run wont pass at 40 flex... having said that I don't know if it passes at 45 flex either.
At least I have a bit of scope to push the BLCK as far as temps and voltage go if i cant get 44x multiplier

 

The crashing is likely to be caused by something else. Usually low flex will induce a BSOD, a crash or program failure is often something else. Can you try increasing the Pri Plane to something like 1300 with those lower Flex setting and see if that stops your crashing...?

 

Will try that tonight.

Edit: Pri plane set to 1300 in the bios vantage still freezing @4.3 and 38flex
might try resetting the bios and going though the settings again to make sure, I am going to follow the video Mr. Fox is there anything in there that I should set differently other then the ram (I only got stock 1600Mhz, do I need to up the ram voltage at all?) cpu multipliers and power limits?

 

Put the power limits and time window values to the max in the BIOS and see if that helps.

 

I think the crashing is simply a voltage issue as that is what seems to fix it

Anyway last night I set up a fan under my M18x outside in the cool air and this is what happened

3Dmark06
Vantage
3DMark11
Firestrike

I'm just not sure if my 7970m's will take 1.1v (running 1.075v at the moment)

 

Assuming tess is not disabled in CCC, your 3DMark11 and Vantage scores would be the highest I have seen for 7970M CF. (If tess is turned off in CCC, they're going to be artifically high and will not count when submitted online.)

Edit: Legit with tess! Nice job! http://www.3dmark.com/3dm11/8197214

Other than one 3DMark11 run a little higher ( this one) every higher submission on record for 7970M CF is a bogus cheat with tess disabled, so you've got some better than average AMD GPUs there Brother Kenny. Congrats on the score.

 

For sure! Kenny you've got a stompin' set off 7970m's. Very nice score indeed. Far higher than I could ever get mine. Well done!

 

This is for the folks that say 32GB of 2133 is a waste...



Moral to the story is... that's only true if you don't know how to put it to good use.

 

Thanks guys, it was a lot of fun.
While they are good scores the over-volt and newer drivers have definitely helped a lot, as for a good set of 7970's, SlickDude80 has a 7970m in his M17x R3 that will go to 1Ghz on the core with stock voltage! :O
I think I'm just one of the few people willing to push the 7970m's in a M18x

 

For some reason this reminds me of Willy Wonka. Definitely a rare high score for the most powerful mobile GPU that Big Red has to offer.

You've got a golden ticket... la, la, la.

<iframe width='560' height="315" src="//www.youtube.com/embed/TW_xGlL1XOc" frameborder='0' allowfullscreen=""></iframe>

I was able to run my 7970M CF setup at 1GHz on core. The drivers were absolutely horrible and my 3DMark11 score was exemplary at that point in time, but substantially lower than yours. [ LINK]

 

Guys! You need to remember...It's not always the guy with the highest clock that wins. It's usually the guy with the best efficiency that wins.

 

That's right. Likewise, a powerful piece of hardware with screwed up firmware and/or drivers doesn't do anyone any good.

 

Yeah I hear a lot of bragging in other threads about getting the highest clocks/volts etc without any speak of performance.

 

Here's a trip down memory lane...

Anyone remember overclocking a GPU with RaBiT? If so, that's good... I don't, LOL.

 

Wow that is a nostalgia trip. Can setfsb overclock the cpu?

Acer Aspire 1690 Series Review (pics, specs)

my 1694wlmi review has a link at the bottom to my overclocking experiences

 

Lowest card i have is the 7150M and a gma 950M.

 

LOL... we have a lot to be thankful for now, don't we? I think the X1300 may have been the weakest discrete mobile GPU ATI ever made, or at least feels that way at the moment.

I see my driver reset itself on the Aquamark3 run. Here it held all the way through without crashing.

I might be able to with a utility... maybe SetFSB. I have not looked into it. Nothing is adjustable in the BIOS. This thing has been sitting for a really long time. Decided to tinker with it. Found a Core 2 T7400, 4GB of Samsung PC2-5300S 667 and a brand new unwrapped Dell X1400 256GB GPU all for about $50 on eBay for everything, LOL.

I'll have to take a look at your link. Thanks.

 

Not even close, weakest notebook I upgraded had ati rage 4mb chip. Pentium 3 700 mhz goodness. Did not even come with wireless so i removed the 56 modem and put a mini pci 56 wireless card in which made it the fastest way to get data onto the machine.

 

oh damn, got me beat by a long shot!

 

With an audigy 2 pcmcia card (hardware sound chip) and upgrading the 128mb of ram to a whopping 384mb meant xp could run and it could play back dvd movies and output 5.1 audio without dropping frames.

 

Dammit kinda makes me wish I never tossed out the old Acer PII 333 MHz desktop I had. IIRC it had 64 megs of ram, a 2GB HDD, and an integrated Mach64 4MB display chip. Came with Win95, upgraded to Win98SE and ran for 7 years (97-04) without a single hitch. Finally retired it after 04 because a Pentium II in 2004 was just ridiculous. But other than it being incredibly slow, it was still chugging along like a champ after 7 years. Good times indeed.

 

If it was a mobile 333 mhz chip it had to be from 1999 onwards.

 

It was a desktop PII, a Deschutes I think. But you're right, Deschutes was only released in 98 so it had to have been 98 when I bought the desktop.

 

rofl my dad has a working 286 stored in our garage i think i played pacman on it when i was 3 years old 5.25 diskete

 

Anyone remember playing QBASIC Gorillas, GWBASIC Bananas and Snakes or am I really dating myself now, LOL? It was so long ago that I don't remember if it was on the Tandy 8088 "beast" or the 286 that replaced it. I do remember when I got really crazy and went all out for a 486SX. In almost no time it wasn't enough to satisfy the need for speed and converted it to a DX using an aftermarket Evergreen Technologies math coprocessor and upgraded to about 1 MB of system RAM. My family thought I was insane and would never need that much power for anything. At that point Gorillas, Bananas and Snakes ran way too fast to control them and I had to disable the math coprocessor to keep those games from crashing, but it was just right for Commander Keen and Wolfenstein 3D, then Quake. I don't remember clearly what display adapter it had, but it was definitely an ISA card. I think is was a Diamond something, with like 256 KB of vRAM. They all played great on that bad mamajama. Most of my friends at the time were wussing out like sissies and going with Windows, but DOS was the only option for the man's man. I finally gave in to the peer pressure around 3.11 for Workgroups. Plus, I had to use it for work whether I wanted to or not. Then came Windows 95, but I still liked MS-DOS and IBM-DOS better until Windows 98. That lasted me a while until I finally broke down and bought a Quantex PII-350 with a first gen ATI Rage card and a 19" CRT that was gigantic and weighed a ton. From that point forward I began building them because nothing prebuilt was powerful enough to get the job done. And, it was more fun and cheaper to build than buy.

 

The first computer I used was a bbc micro and the first game was the land the space ship one which my mum programmed in.

 

I had a 80186 with 300 modem and i thought i was the man!!!!!!!!!!. My dad brought it home and then left on a business trip. Needless to say I tore it apart to see what it was.
The first game i think i played was zork. It was text based.