AlienwareTech @ E3 - Alpha SteamBox Console Replacement

That's a great price. I can't say anything kind about the Haswell turd CPU, but it is what it is. Could be worse... at least it's not a crappy AMD CPU. If it's one of the rare samples of a Haswell that doesn't have throttling or heat issues, it should do a fairly decent job with Liquid Ultra thermal paste. Hopefully, it will handle at least 4.3GHz stable without catching anything on fire.

But, I'm right there with you. If I had $4K in discretionary cash I would go buy a new P570WM with a 4930K and 780M SLI and put my M18xR2 in the closet as a cherished memento of a better place in time. I still have the original box, cloth cover, manual, cap, alien head domed plastic stickers, etc. All that I am missing is the moth balls.

 

That is an insanely good deal, if you're ready to buy I would pounce on it before someone else grabs it.

As Mr. Fox you will need Liquid Ultra for the 4930MX. Heck I'm using Liquid Ultra even just for the 4900MQ and I'm not even a hardcore bencher, but Haswell just runs asinine hot, and doesn't help that there's no AC in my unit so my room is basically burning during the summer.

 

I know. Fortunately this system has 780m's instead of 880m's so I have one less issue to worry about. I guess Haswell will be a learning curve. I will share my findings.

 

Every chip will be different, but just to give you an idea, even with a -80mV undervolt my 4900MQ can easily OC to 3.9GHz stable (stock is 3.6GHz). The default voltage is pretty whack for most Haswell chips.

 

Only one can hope that Broadwell is much cooler and sufficient. Oh, how nice it is to finally dream again.

 

Yeah my 4940MX with a -80 undervolt will still run up into the 80s when something pushes it and that's only at 3.79GHz

It will rarely hit 3.9GHz with all 4 cores but when it decides to, screeching fan and 85C minimum.

Sent from my HTC One_M8 using Tapatalk

 

Sorry to hear that, Brother Ethrem. The Alienware 18's main problem is the power handling limit. Even with the crippled BIOS my 4930MX overclocks pretty well and I can easily bench with the CPU at 4.5GHz and sustain manageable temps with Liquid Ultra and manually running the CPU fan full blast. (The fans in the 18 are not very loud. Even when all three of them are running full blast it is not particularly loud.)

The real problem is when you try to push the video cards along with the CPU overclock and it runs out of juice. Not being able to use the dual AC adapter mod that works on the M18xR2 is a real setback to benching.

In the grand scheme of things, I guess Alienware 18 owners have less to complain about, but that still does not excuse the fact that it cannot keep up with the M18xR2 in terms of performance and has fewer features. I don't like having to use a sliding scale or grading on a curve by comparing it to less capable current generation machines. I think all models manufactured by each brand should be, at minimum, benchmarked and compared in light of what that model's predecessor was capable of, with an expectation that it match or exceed the features and performance capabilities of the product it is replacing. Alienware missed the mark with the 18 in that respect just as NVIDIA missed the mark with GTX 880M. This is not a favorable trend in technology.

Here is an example of the CPU overclocking capability of the Alienware 18. This is not maxed out, but it is a recent example from last week were I was trying to regain some performance that was lost with the release of BIOS A08.

Spoiler

 

Yeah... my 4940MX won't do 4.5GHz. It struggles to do 4.2GHz on two cores as a matter of fact.

I've tweaked with the amps and all of that stuff (how do you think I overloaded my adapter the other day lol) but it won't stabilize past 3.9GHz. Maybe if I got some Liquid Ultra and a converter box... But even then, if its anything like my 4770K, it will take a lot more voltage to get to 4.5GHz. My 4770K won't POST @ 4.5GHz with 1.35v

 

Sounds like you may have gotten a dud of a 4940MX in that case, although if you could share your XTU settings that might help.

 

To me it looks like the i7-4930mx overclock better than i7 4930mx.

 

This is the only way I can ensure 3.9GHz, normally I just run it with a -80mV and let it do its thing

Spoiler

 

Cpu is far too hot for this clock frequency. Over 10 degrees celsius too high for that clock speed. I used Gelid extreme earlier and a good application with good results in 4x39-4x40 with a higher voltage than you and better temperatures than what you have. I think the application of your cooling pasta is too bad. Just a poor application of thermal grease leads to 10 degrees difference between a good application. Even 4x41 with Gelid extreme was no higher temperature than what you have. It is up to 20 degrees Celsius difference between a bad application of a good thermal grease and a good application of Liquid Ultra. This I know because I have tested.

 

Well I repasted it with IC Diamond twice (because the first time I didn't put a line all the way across the die and the first and last cores were 2-3C hotter that the others), I think that the CPU cooling in the Clevo is not adequate. It's just a simple copper block that loops two heat pipes up to a fan sitting on top plus the area that the fan vents out of is much smaller than what the GPUs get.

And that was the 8 thread test... 4 threads maxes at 78C

Sent from my HTC One_M8 using Tapatalk

 

Remember My Alienware 17 also has only two heatpipes and not 3 pieces that Alienware18 have. Although my machine has only 2 pieces are not my cpu hotter an Alienware 18 with 3 pieces heatpipes. There is always 3-7 degrees difference in temperature of such core 0 or 1 and core 4 Do not make the thin line of thermal grease all over the processor core, only 3/4

 

I may get around to repasting again... for everyday usage, the temps are fine. I don't usually see above 80C with the workloads I give the machine.

 

The application of thermal grease is more important than the quality of cooling paste.
It is worse with too much thermal grease than too little.:thumbsup:

 

Yeah but IC Diamond has a larger tolerance for too much vs too little compared to some other ones.

Its such a pain in the to apply it though.

I wonder if I can use MX-4 on it... Its much easier to work with.

 

Remember that IC Diamond is thicker than normal thermal grease. With too much thermal grease, it is worse than other thinner consistency. Remember to warm up a thicker thermal grease (put the tube in a bag and put the bag in a cup of warm water about 55 degrees Celsius) before applying it on the cpu. This is IMPORTANT. Maybe Gelid Extreme (very good paste))

 

One of the benefits of ICD and gelid extreme is that they last much longer than other pasta.
I helped a friend with and apply new thermal grease on an Acer. Termperature of this machine was very high. This machine has only one fan and cooling pad. During the test of 3dmark vantage cpu and gpu was 93-96 degrees Celsius. Used Noctua NT-H1 thermal gras. Tested again and the temperature of the CPU was 65 and GPU was 72 degrees Celsius during the same test. CPU and GPU use the same heatpipes. The machine was colder than it was purchased new.

 

I have seen that recommendation before and I don't really understand it because I don't have problems getting it out of the tube and it seems to spread just fine (I've repasted my GPUs probably 10 times with it because I was going nuts over temps).

I have a ton of MX-4 though and I'm almost out of that tiny tube of ICD... maybe I should do some experimenting with MX-4 and see what happens. It was too thin for my GPUs but hopefully there is adequate pressure for the CPU die...

On the point about longevity, MX-4 is rated for something like 7 years.

 

If that run was with auto fan then it's mighty impressive. Here's what my 4900MQ with Liquid Ultra looks like:

Spoiler

Runs just as hot if not hotter than your 4940MX, not to mention 1 second slower. And this is with Liquid Ultra no less. Granted room temp is 25C (77F) right now, but that's no excuse. I'm pretty confident I applied Liquid Ultra properly too.

 

I helped a friend with and apply new thermal grease on an Acer. Termperature of this machine was very high. This machine has only one fan and cooling pad. During the test of 3dmark vantage cpu and gpu was 93-96 degrees Celsius. Used Noctua NT-H1 thermal gras. Tested again and the temperature of the CPU was 65 and GPU was 72 degrees Celsius during the same test. CPU and GPU use the same heatpipes. The machine was colder than it was purchased new.

I have tested Gelid extreme and without heating the thermal grease. The result was worse

 

Well your chip is being fed a LOT more power than mine... 110A? You shouldn't need more than 80 to get 3.9! You also bumped your TDP up to 95 as opposed to my 71W.

You would probably get a good 5-6C lower than that if you tweaked your values. I tweaked mine until I found the lowest stable point for 3.9GHz - if I plugged 95W and 110A, I'd have throttle in seconds I bet.

It makes no sense that you have to heat the thermal paste first, it doesn't change the conductivity of any thermal paste that I'm aware of. The only reason I've seen to heat up thermal paste is to make it easier to spread which isn't a problem if you apply it in the correct pattern (X on a square like an IHS on a desktop GPU, line across the middle for a rectangular surface) or you use a lower viscosity paste (MX-4 is significantly easier to spread than ICD)

 

So this is what it looks like with max fans:

Spoiler

I did another run with 75W TDP and 80A as the current limit, made no difference whatsoever unfortunately.

Spoiler

I honestly ddn't think it would make a difference though. I've long suspected that since the it's called turbo power max and current limit, it only sets a cap but doesn't force the CPU to completely use up all that power. For example during the 1024M run my CPU was only drawing about 60W, nowhere near the 95W cap I set. Seems like it's indeed the case here.

 

View attachment 113417 This is with 4x42

 

But Firestrike's physics test is such a short bench it won't truly reflect how hot your chip will run under sustained load. If you want an apples to apples comparison please run ThrottleStop 1024M bench and report back your temps.

 

Hmmm your CPU functions differently from mine. Anything I set it to, it will gobble it up...

Keep in mind, I'm using static voltage in my benches here. Dynamic voltage almost always pushes more than the chip actually needs.

Fans on max

Spoiler

]

 

What I meant was that by and warm up the thermal grease, it is easier to apply. With the thermal grease is heated, it is easier for the thermal grease squeezed out of the heatsink. There will be a thinner application of grease from it. Google translate ;-(

 

Pressure from the heatsink *should* be enough to do that without any heating but yes it does help it get out of the tube easier. Honestly I have no idea why my IC Diamond gives me no problems coming out of the tube. I push the plunger and hold it right above the for then when the drop starts to form I push it onto the die and slowly push til it comes out then carefully lift up and let it slowly string together so it doesn't bubble.

Sent from my HTC One_M8 using Tapatalk

 

Makes sense, you have an unlocked MX chip, so it will probably take whatever you throw at it. My locked MQ chip will only take as much as it needs to accomplish the task.

Yeah dynamic voltage usually adds about 30mV or so under load I think, that would maybe account for 2-3C max.

Dammit now you're making me wonder if I actually did do a shoddy job of repasting Liquid Ultra. Either that, or under sustained heavy load, the heatsink becomes saturated and heat dissipation from the fins rather than heat transfer becomes the bottleneck. (comparing the CPU temps in the Valley benchmark in your Clevo thread makes me believe it might be the latter, at least for my own sake )

And also we can probably conclude I have a pretty shoddy 4900MQ as well. (look at the TS bench times)

 

This is what I got with XTU when I used dynamic with -80mV, note I had to increase amperage and watts to get 3.9GHz (it was running 3.8GHz on the same amperage and watts of the static voltage)

Spoiler

You can see temps went up as did power requirements. I actually hadn't played with this before, this is good information to find out just how much of a difference it is.

Your TS bench time is a bit off from mine... is there something else the MX has the MQ doesn't? Faster cache or something? I think its pretty much negligible the difference you have versus mine. What's your XTU bench though? My XTU bench is always bad.

Its 854 @ those settings.

As for the repaste... I don't know... because I am pretty sure I did a bad job on my paste job, I just didn't feel like doing it again so I settled on it. Its lower than it was out of the box so that's good enough for now.

 

Old test View attachment 113418 Unfortunately, I have a useless internet speed

 

OK, so 3-4C temp difference. That might be the difference between ICD and Liquid Ultra then, assuming I did not royally screw it up (I know I definitely did not the first time, and this is my second time repasting with CLU).

I just realized, I kept cache @ 3.8GHz instead of 3.9GHz LOL. Which may account for that 1 second -- sounds like a lot, but it's only a 0.4% difference, so it's probably what it is.

 

Envious of 4.2GHz... not envious of temps.

 

Fire Strike physics testing is kind of a joke. You can have a great GPU and a pathetic CPU and still walk away with a fairly impressive overall score. It's good if all you want is a graphics performance test, but the results are misleading because it makes some systems appear to be better overall performers than what they really are.

 

I was always under the impression that Liquid Ultra was like 7-10C difference from ICD. Hmmm.

Well maybe my chip isn't as bad as I thought? But the fact remains I can't get the clock speed stable past 3.9GHz... I believe 4GHz needs 1.2v, I don't remember exactly what it was, but it was over 1.175 for all four cores.

 

Liquid Ultra was 7-10C from MX-4, but ICD is superior to MX-4, so maybe 4C is about in the ballpark.

I'm honestly not sure what it is, maybe I really did screw up the second paste job with CLU, and that would be very sad. Or maybe I should've kept acid off my CPU die and heatsink rofl

Also, I actually like dynamic voltage better precisely because of that +30mV or so overvolt. Cuts down on the number of 0x0124 BSODs (Vcore too low) because there's a small buffer.

 

Yeah, I hate BSODs... I WHEA'd earlier when I forgot to set my multiplier to 39 instead of 40 when I applied my 1.111v lol. Its crazy with these Haswell chips... they are really finicky... Usually voltage scales fairly well across clocks but that doesn't seem to be the case with Haswell at all. My desktop 4770K is a perfect example... was a nightmare getting it to 4.4GHz and it still randomly will go unstable on me.

I don't remember these problems with Penryn, Kentsfield, or my Athlon 64.

 

If you run at a higher voltage and a higher overclock. Then the temperature go much higher. Then you will see that you should use liquid ultra. There is a big difference between 4x38 - 4x39 and 4x42 and 4x43 in terms of temperature during tests. It will be a much higher wattage during the tests.

 

Yeah I rather not mess with it because I don't think my Clevo cooling can match an Alienware's cooling.

 

So I learned more by spending 40 minutes in this thread today than I did the entire time I was benching in the blind. Just goes to show how important a community is in terms of tweaking.

Also, I'll take that 4C as some consolation that I did not completely botch my Liquid Ultra paste job. Though I did find it very hard to squeeze any out when I tried to repaste last week, maybe the stuff really has degraded while sitting in the syringe...

If you're gunning for CPU-only benchmarks, you can try the triple pipe heatsink mod I posted earlier (I really shouldn't call it a mod since all I did was a component swap). It's basically like using Liquid Ultra minus all the risk. You won't want to game with this setup, but GPU benching might also be OK since the GPUs never quite reach their max temps in these 3DMark benches I find.

 

View attachment 113420 View attachment 113419 See this test report. These temperatures are very high even this the processor uses little watts and a low CPU clock speed. This is the original thermal grease that Alienware use. Therefore, you see that a good application and a good thermal grease needed.

 

Nah, its no point to overclock the CPU if I can't use it for practical purposes like gaming. And besides, I finally got temps of that slave card under control, I don't want them running wild again!

Well that's not saying much... I've seen Alienware's stock paste jobs... They could use the best paste in the world and you would still have astronomical temperatures. They might as well put a thermal pad in there because that's how thick their paste job ends up being anyway.

 

See the second photo I included. This is the new Alienware those tested. This is even worse. A poor quality or a bad application.

 

Only difference is that one machine uses slightly higher volts and consumes a higher wattage.:cry:

 

Wow this thread expanded since I went to bed and then to work... I have pulled the trigger on a P370SM for that great price. I also have a host of other parts to plan and buy such as an additional adapter plus converter box, a P370EM heatsink set (will be fun to play with even if GPU2 temps may be a bit higher) and some other goodies.

I will also be looking to change the 60Hz LCD to something else. I can easily buy a eDP cable from Eurocom and I have sourced what looks to be a reliable retailer that offers many screens to choose from and even lists the interface whether it be 40pin LVDS or 50pin eDP such as the 120hz LG display and even the IPS screen that is in the Dell and HP workstations.

I'm hoping to remove the branding on my new laptop and look at customising it somehow. I want there to be a small tribute to Alienware on it.

 

These are new numbers on Cinebench 11.5 and Wprime 1024. Tests were run on my Alienware 17 with an ambient temperature of 22 degrees Celsius. Using only my modified CoolerMaster Notepal U3 (2x140mm and 1x80mm fans) Fans run on 12 volt pc power supply. Cpu running on my standard overclock 4x42. What do you think about the temperatures and results of the test of Wprime 1024? View attachment 113436 View attachment 113437

 

You could have just picked up an m18x R2 and stuck with alienware. sucks alienware has become so stale that they are losing customers that truly love the brand. that's pretty bad.

Sent from my SM-N900T using Tapatalk

 

Maybe it will force them to wake up and change their policies and engineering.

Sent from my HTC One_M8 using Tapatalk

 

I love my R2, its still under warranty and does all that is asked. However, All I can say is that unless AW pulls their finger out by next release I'm done, I'll be forced to go else where... no choice. You'd think AW would want to hold onto their enthsiasts that need a DTR that's capable of powering its internals. Fact is there are no real DTR powerhouses offered anywhere that can be bought new except Clevo/Sager. For someone that's after a portable Extreme beast machine there's slim pickings. What's sad about it is that AW doesn't even seem like they're going to put up a fight.. just roll up the white flag like a bunch of wusses instead.