Max RAM Speed for Alienware 18

I always used to think that the max RAM speed you could install is 1866MHz since that's what most custom PC retailers like XOTIC PC and Powernotebooks offer as an optional update.

But I saw in the BIOS under the performance section that you can go up to 2600 MHz if I remember....

is this true? will my Alienware 18 work with the faster RAM? and is it worth it to upgrade or not really?

 

I have not seen any RAM past 2133 MHz, so there aren't any XMP or SPD profiles for 2600 I assume. What you COULD do, is buy 2133 MHz RAM and attempt to overclock the frequency and loosen the timings.

 

so are you saying that 2133 MHz. RAM will work??? I have never seen anyone with such a setup hence why I asked

 

I know people that have made it work in the M18x R2, I can't see why the Alienware 18 would be any different.

 

Strange thing, the Crucial Website shows that my current RAM 1866MHz. is the max:

http://www.crucial.com/usa/en/compatible-upgrade-for/Alienware/alienware-18

 

I have 2133Mhz Corsair Vengeance in my AW18 it will not run above 1866Mhz

 

I think you have right on that. I think Mr Fox test this earlier with his AW18 Viking. Only 1866 mHz.

 

thanks, strange why in the BIOS under performance, where it says Memory at the bottom, you can change it from Auto to 2166 and above, can you try manually changing the speed?

 

so why are there higher speed options in the BIOS?

 

Maybe a Garbage Dell bios..

 

not surprised! Thank god this is the last Dell computer I am buying!

 

It will run 2133, but you need to use Thaiphoon Burner and tweak the tCKmin value to bump the RAM clocks up to 1068.38MHz and then it will run at 2133. User mw86 at Tech|Inferno was able to manage 2400MHz from his Kingston Impact memory using this technique. By default, 1866 is the max.

This is the fastest SO-DIMM available now: http://www.kingston.com/us/hyperx/memory/impact LOL x100 at the "ECO-FRIENDLY low-voltage" nonsense part of their ad campaign. Give me a break... how extremely stupid.

See these results with the Alienware 18 that I posted about a week ago.

 

Right, thanks for the Mr. Fox, but this doesn't answer, why does the BIOS have manual settings for more than 1866 if it wouldn't support it normally without such tweaks?

 

The BIOS does support those settings, otherwise I would not be running at 2133 and mw86 would not have hit 2400. The tweak is required to circumvent a compatibility issue. It's not the fault of the memory, it's the BIOS. All of my 2133 RAM that runs at 1866 in the 18 boots and runs at 2133 automatically in the M18xR2. I think the obvious underlying answer is that Alienware did not set up the BIOS correctly, and they did not thoroughly test all of the settings. Yeah, I know my surprised look is on my face about that, LOL. If that one thing were the only thing about the BIOS that they borked, life would be a lot easier.

 

It is most likely an even more useless bios in latest 2015 model series of Alienware .. New Alienware can not use raid and overclocking of the processor has one step and that is when you hook up a G / A amplifier. And of course the amazing hybrid functionality. Dell will probably work even more for that hybrid functionality to improves the perfomance. Nice with several charge cycles of the battery even if you only use your laptop connected electricity grid. And you have no chance to recover a fully crashed bios on the alienware 17 R2 if you try to overclock with xtu and get crash.

 

Wow. Thaiphoon burner should never be a requirement to run high clocked RAM.

 

It's only a matter of raising the memory clock by about about 1MHz to get the BIOS to recognize that it needs to run at 2133 instead of 1866.

 

There is actually a way to do this without the burner - check my old post here: http://forum.notebookreview.com/thr...higher-speeds-on-aw18-r1.768341/#post-9878147.

It's a little annoying to do since you have to experiment with the memory timings (not all timings are published), but you don't have to deal with the RAM firmware and can do it in the BIOS. The memory speed setting actually chooses the memory ratio.
I run mine (which are rated at 2133MHz) at 2400MHz.

 

And these are the modules that I'm using for the 2400MHz - I've managed to get it to run at up to 2566MHz, but it starts corrupting after 24 or so hours of burn testing so I've stuck to 2400MHz.

 

Care to share what settings you used?

 

The quality of the memory controller in your CPU can have an impact.

 

They're in the link (in the photo of my BIOS: http://forum.notebookreview.com/attachments/1-jpg.119284/) - the setting that I have in the screenshot is somewhat stable for 2400MHz (it does have silent memory corruption that only presents after an extended time - three days continuously under load), so I recommend that you do your own stability testing (which you should anyway since silicon lottery). It's a shame that we cannot "overvolt" these modules to 1.5V though they're rated for that as well as 1.35V.

If you change the ratio to 7, they'll run at 2133MHz (with slight headroom on the clocks from the SPD). The BIOS is derpy with the timing names (two tRCs for some reason, and I've never seen tRPab before anywhere - the link also has guidelines for these.

The SPD timings for Corsair Vengeance 2133MHz to get them to run at 2133MHz would be (default SPDs except for the odd AW BIOS ones):

2133
Custom setting
7
11
11
11
31
8 (tCWL - there was a mistake in the post earlier which said that you should use tCL-1, but that only applies to very low latencies - I've fixed it now)
33
42
0 (2nd tRC)
278
16 (tRPAB)
7
9
16
9
2N Mode

Anyway, use and tweak at your own risk, but I'd recommend not touching the NMode since that will likely necessitate a BIOS reset.

Edit: The multiplier should have been 7 - fixed now. Oops.

 

Also, if anyone knows more about tRPab maybe they could explain that more. I'm a mathematician and my knowledge of semiconductor design is cursory at best.

I couldn't find anything except in technical manuals (JEDEC standard) - it stands for tRowPrechargeAllBanks, while tRP can mean tRowPrechargeSingleBank OR tRowPrechargeAllBanks (abuse of notation). Clearly tRPab > tRP, but I don't know how much would be optimal, so the rule of thumb I've included is from experimenting and going through various code that simulates microarchitectures (e.g. https://stuff.mit.edu/afs/sipb/contrib/linux/lib/jedec_ddr_data.c) + a couple other Google hits.

 

Your tRFC is very high at 278, try and get it below 200.

 

Yeah - those are published SPD settings for the particular module at 1066MHz, and they most certainly can be tightened

 

Yeah I read that as your own, whoops

I get my own down to 105 but that's not normal for chips to be doing up at 2133mhz.

 

Does anybody know what would happen if I mixed CL10 and CL11 RAM together? The 2 pairs are both 1866MHz, 2 x 8GB.

 

Probably nothing other than the CL10 would run at CL11. When you mix memory modules, the (s)lowest common denominator becomes the Achilles' Heel that sets the bar for the rest of the memory modules. It should still function normally and it is very unlikely you would notice any difference. You might measure it with a memory benchmark, but other than that it should be mostly irrelevant. The only thing you don't want to do is try to use DDR3 1.5-1.65V modules in a system that is designed to use DDR3L 1.35V modules. Results can vary from works OK to won't boot, so don't mix DDR3 and DDR3L.

 

Sometimes it can pick some really odd sub timings too, but it will always err on the side of caution and set looser timings.

 

Sweet. Thanks, guys. What happened was that a pair of my Kingston HyperX 1866 DDR3L went bad ( computer kept bluescreening, I checked with memtest and other diagnostic tools to narrow it down to that pair). Sent the RAMs to Kingston and they sent back the newer HyperX Impact with same speed but they're CL10 (vs the old ones were CL11). I still have a pair of CL11 in my machine but they're under the keyboard so I'll probably mix the 2 pairs for a couple months before I switch out the remaining CL11 pair for the newer CL10 ones.
It was bizarre. I'd never encountered bad RAMs before. Well, not Kingston at least.

 

Yeah, that's not very common for Kingston. Their products are usually top notch, although nothing is perfect. Good that you isolated the problem to the offending modules. Shame they did not give you the option to RMA all 4 sticks and get 4 matching sticks in return. Since the new 15 and 17 only have two RAM slots, maybe you can unload the good RAM sticks on one of the new 15 or 17 owners and use that money to help offset the cost of 2 more matching sticks of Impact.

 

The new sticks should be able to do the old timings. Does it save the current timings when you go to do custom ones? If so select default timings with the old sticks with the new ones out, reboot and then select custom timings and use the standard timings, then put the new sticks in. They should be able to match the old ones and give you the same performance as before.

 

So which program do you guys use to check active timing settings? HWInfo and RAMMon reports max/min supported timings, and CPU-Z's Memory tab reports CL11 timings but I'm wondering it's the actual active timing or it's just pulling the lowest available one.

 

I believe that CPU-Z memory tab is correct, though it only shows a handful of timings. Also, it might not show all SPDs on the SPD tab (my SPD tab only goes up to JEDEC #6 at 1000MHz).

 

What is show in the screen shot for the "Memory" tab is actual (active). Multiply x2 to get the effective rate, which is 2133 in this example. If I am understanding your question, there is no "lowest available one" so to speak. All the modules will run at the same speeds and timings, and that is determined by the slowest module if they are mismatched. The information on the SPD tab is more or less useless... ignore it. It is not accurate most of the time and not complete. Use a program like AIDA64 (one option, there are others) to see the actual SPD information.