*OFFICIAL* Alienware Area-51M R2 Owner's Lounge

Yes, it's likely even worse when you move to Xeon mobos, not sure if I saw any Xeon results.

 

My unit got shipped today !!!!

 

Latency is a problem sadly that won't be avoidable easily on the HEDT platform. HEDT focusses on sustained performance/reproduceable numbers while the consumer parts only need to crank out numbers.
It's because of that reason you pay so much for the licensing (and optimization of the platform) HEDT is better for long term support while consumer machines are better for machines that do not have a high impact availability (client machines for example which are fearly easy to replace in comparison to an server or a whole datacenter)

 

Nice. Looking forward to seeing your thoughts on it.

 

Are your thoughts still the same after a few weeks of use ?

 

Yeah i really like it. It runs cool enough for me and performance has been great so far, never throttles at all. Was playing the witcher 3 for 9 hours yesterday no problem at all.

I bought into it knowing i likely can't upgrade it which is why i bought the best CPU/GPU combo so the lack of upgrades doesn't bother me.

 

Glad to hear the good news on shipping and the thoughts on the system unfortunately my system has been delayed at least a day, Dell are trying to expedite. Also I've been given a £100 refund after I threatened to cancel so I recommend everyone does the same if facing any delay.

 

Port Royal doesn't care about cores, which is why you see people run 5-5.2Ghz in it.

 

He need to check out all dif benches from Hall of Fame to find the performance king.

 

Interestingly, quite a few Xeon W-3175X s in TimeSpy hall of fame.

Hmm, for $5000 one can easily build a desktop with a CPU almost 3 times as powerful as the 10900K in 51M...

 

I can't wait to get mine. I ordered it on July 10th! Hopefully it ships soon.

 

That's insane. The R2 gets 23196.

 

Yes, the dGPU is limited to x8 lanes in the R2

It is incorrect to generalize the x8/x16 issue in such a way. It is very situational. Some work loads will be affected and others will not. My Titan RTX in the AGA (x4 PCIE 3.0) exhibits between <5% and 10% performance drop for most games played between 1080p and 4k resolution. I can force the difference between the x4 and x16 link to be as much as 50% but that is with an extreme case of playing esports titles at 1024x768 and unrestricted framerates.

AGA uses x4 lanes and thunderbolt uses another x4, but otherwise your point stands. I knew what you meant, but other readers might not.

x12 lane configurations are not possible and the caldera doesn't support x8, so Alienware hung the TB3 controller off the leftover x4.

For most but not all. The 13r1 and 13r2 were based on Broadwell-U and Skylake-U respectively and used a PCIe multiplexer to split a single x4 PCIe 2.0/3.0 between the AGA and the dGPU.

What you don't find appealing about the AGA in the 51m is specifically why others really like the machine.

First, using the 51mR1 and an AGA with even a desktop 2080 is generally faster than the built-in 2080 dGPU, assuming you are outputting to an external display attached to the AGA. The x4 interface is not a significant bottleneck in most gaming scenarios for anything up to a 2080ti and even the 2080ti is about 10% slower at worst compared to x16. Meanwhile, the GPU in the AGA can easily exceed the 200w power limit the built-in 2080 is limited to and with an AGA, the CPU can run at higher frequencies since it no longer must compete with the dGPU for cooling on the unified heatsink. Faster GPUs and frame rates will exacerbate the AGA's x4 weakness for sure, but that doesn't mean a next generation card is definitively pointless.

Second, the x8 link for the built-in dGPU is absolutely not a bottleneck, even for the strongest desktop cards. A 2080ti on a PCIe 3.0 x8 interface generally performs within 2-3% of of the same card on an x16 interface. Any discussion about future cards is conjecture, but even if Ampere brings 2080ti levels of performance to a 51m compatible 200w power envelope, the x8 interface should not be a concern.

This is a nice thought, but the complexity to accommodate this is unfortunately impossible within the current motherboard design. While I would love an x8 AGA, the current caldera is also grossly inadequate for x8.

I'm hopeful we might get a decent AGA redesign when PCIe gen 4 hits Intel's mainstream platform. Since the x16 PCIe 3.0 link is just now starting to see measurable benefits over it's x8 counterpart with the 2080ti, I'm hopeful an x8 PCIe 4.0 bus with the same bandwidth will give more life to the AGA concept without costing any performance for the built in dGPU.

Wow, I know this is wccftech, but still...there is a lot of sillyness in that article.

 

Did you post some standard firestrike scores? 1920x1080...

Here is how my unit is performing after two months with non k 9700 and 2080 R1 after probably 300+ hours of BF5 multiplayer.. Would be nice to put a 9900 in it but i really dont want to deal with the temps as well as faster memory, wish I would have looking back on it but I spent as much as I wanted to spend with 700$ off. I do love my unit though, best gaming laptop I have ever owned by miles. If the R2 used the same keyboard which I love on this unit it would be a no brainer for me as my next in line but there are not that many great games out at the moment unfortently and BF6 will be out probably after the 3080 GPUs hit.

 

Here it is:



Interestingly your graphics score is higher? Is yours overclocked?

 

Interesting how the combined score takes a hit with the CL21 sticks, when compared against R1.

 

How many watts are being pulled through the GFX card on the R2? Does look lower than others but not too far off, could be the balance shifted completely to CPU due to the power hungry 10 core.

 

Just for comparison, here is my Fire Strike on my Area-51m R1 with a 9700K/RTX 2080

Fire Strike (Driver 442.23-BIOS 1.3.2) OC 20 MHz Core/150 MHz Mem

 

Wow,

If I'm reading that correctly your Physics score is 2/3 of what @ratchetnclank has and yet your combined score is roughly 10% higher.

I am shocked that the CPU's apparently vary so much, but also the Physics score when so much higher contributed so little to the combined test?

Its as suggested that when under combined load the CPU is drawing power away from the GPU?

What do you think of the scores and could there be any way to improve Rachets? Anyone else have any suggestions?

 

Since I am running a 9700K only, I have high hopes that I will be able to get much higher scores than what Rachet has. He has a good base score to start with but there is a lot of room for improvement like:

1) Disabling Telemetry
2) Disabling most background apps (except the ones related to the system/drivers (ie. Realtek Audio Console, nVIDIA Control Panel)
3) Undervolting
4) Not using Windows Defender (the 3rd heaviest Antivirus, see: AV-Comparatives Latest Performance Results)
5) Ensuring your nvIDIA Control Panel Power Management is NOT set to Optimal

Spoiler: nVIDIA Optimal Power Management Explanation

Right click on your desktop then choose nVIDIA Control Panel. Go to Manage 3D Settings from the left pane, then scroll down a bit until you see power management. If you see it's set to OPTIMAL which is the default setting when you install a driver. That's your issue.

Optimal Power means when there is no draw on the screen, the GPU clock speeds is set 0 MHz to save power and then ramps up once it needs to. Sounds great on paper, works like crap. This is the number one reason why anyone might experience crappy performance from their nVIDIA GPU. What's worse, is that's the default setting in the nVIDIA Control Panel after you install a new driver often leading people to blame the driver for bad performance when it's just the fault of nVIDIA's clowns.

Set the power management to Adaptive which puts the GPU at lower clock speeds when no GPU intensive apps are in use and it would ramp the clocks up when needed. That actually works. It's the best balance between getting lower heat from the GPU and good performance in games.

When benchmarking, for the optimal results, it's best to set the power management to High Performance.

Mind you, after you change the power management to whatever you set it to, a reboot is mandatory for the new clock speeds to take effect.

Classical nVIDIA Swiss Cheese

6) Not sure about the R2 but it's also worth checking if the nVIDIA Temp slider can be raised like on the R1:
Guide: How to unlock the GPU Thermal Limit Slider in Alienware Command Center

 

Wow, Dell must have upgraded your R1 to 2080 Super!

 

See: http://forum.notebookreview.com/thr...r2-owners-lounge.832848/page-92#post-11035038

Spartan =

 

Faster memory with tighter timings will help.

10900K // 28.7K Physics // 9.8K Combined // 2933MT/s CL21-21-21
9900K // 28.6K Physics // 12.4K Combined // 3333MT/s CL13-16-16

 

I asked in another section, but I do not think people understood what I was asking about so I will try it here.

I am not at all into gaming. I do not care about aesthetics. I only care about CPU power. I do scientific computation, and so I'm quite sure a lesser gpu would work for what I need to do, like look at 3D models and rotate them around, etc. It is pure computing power I am after.

That said, it seems from reading the posts here the i9-10900 will have voltage limiters placed on it? Dell seems to admit this in a chat I had with them, though they claim it will have minimal impact on performance. (See image below)

It seems like all the benchmarks people have been publishing for this machine are for the GPU. Does anyone know of benchmarks that test the Area 51m R2 for CPU, not GPU? I would most likely get the 10th Gen Intel® Core™ i9 10900K (10-Core, 20MB Cache, 3.7GHz to 5.3GHz w/Thermal Velocity Boost) and the 2070s since that is the cheapest GPU they offer. I want to know if it is that much better than getting one of these for what I want to do:


https://www.velocitymicro.com/wizard.php?iid=307

It comes with an i9-9900k as an option.

Thanks for any input! (See Update below)








UPDATE: It seems Dell now says there is no limit on the power. Can anyone think of a Gotcha that might accompany this conversation, as in something like, "We meant full power when running in serial. You can't expect all 10 cores to run at maximum speed!" Thanks!

 

You want to look at something like wPrime or Cinebench R20 - the latter is probably more AVX heavy. Just google say "10900 notebookcheck" to see all the benchmark results.

 

That was 192W draw on the GPU and 172W draw on the CPU.

 

Wouldn't the Clevo NH58AF1 with up to a Desktop Ryzen 3950X be a better choice at half the cost then?

Yes, its limited on power for the laptop, but the pure thread count and performance per Watt is much better for compute tasks over the Intel alternatives offered from Dell and Clevo.

 

I honestly hadn't thought about AMD, as I have been using Intel since my first Pentium III at 512 Mhz! But looking at the specs and benchmarks I may just change my mind. One question: "limited on power for the laptop" what does that mean? Does that mean it won't perform as well as it would in a desktop, say in the setup below? If so, how much degradation should I expect to see?

https://www.hidevolution.com/evoc-h...x-ryzen-9-3900-ryzen-9-3950x-gtx-1660-ti.html

 

The EOL Clevo P870Tm with the standard 2080 cards. And only 8 cores. All 2080 in notebooks is hard capped at 2100 (All of you can be winners ).



What clock speed are you able to reach on the Super card? I mean above 2100. Thanks
https://www.3dmark.com/fs/23165643

------------------------------------------------------------------------------



At 125W ? LOOL (Dell support ain't the best place to ask about such questions).

Ryzen desktop chips is capped at 65W (Eco Mode). Can read about it here...
https://www.tomshardware.com/uk/reviews/amd-ryzen-9-3900-review-eco-mode

 

Thanks for responding and running the test. +rep
Yes its overclocked a bit using msi afterburner.

 

If I may complete your excellent points, there is also the 16 core 3950x. It comfortably beats 10900K in CB20 in this 15 inch laptop, despite operating at just 65W:

https://www.eurogamer.net/articles/digitalfoundry-2020-xmg-apex-15-review-ryzen-9-3950x-in-a-laptop

Perhaps the main caveats are very minimal battery life, and the lack of TB3. None of the AMD laptops have Thunderbolt. That's fine if one doesn't care about TB3 docking stations (most probably don't) and eGPUs (ouch). The price of the top spec 3950X config starts at around 2000 euros. Not sure what the starting price of the 10900K variant of 51M R2 is, but I heard they started taking kidneys as payment.

Edit: the 10900K 51M starts at £2850 so about 3100 euros. It comes with a non-downgradeable 2070S though, vs 2060 in the Apex. For pure CPU loads, the XMG Apex looks like an amazing deal.

 

Oh I remember that bench, @Prema had ~30mV left on GPU core but had hit the 2100Mhz core wall. 31K graphics was a nice uplift though.

 

If like anyone you do decide on purchasing a Dell system I highly recommend you use the Chrome Extension "Honey" at checkout that can automatically test multiple promotional codes. This in addition to any personalised codes you may get from the EPP or Advantage schemes if your eligible.

That and I recommend speaking to Dell direct as they may be able to customise the build so you could have the 10900K with a lower spec GPU not normally advertised.

I bounce between consumer, military buyer and a registered partner reseller (family firm) account on top of speaking to Dell Direct and our official Supplier to tease out the best deals for my family firm, work or myself. That and compound deals with codes with reward vouchers.

Speaking of which if anyone has any money saving tips on Dell purchases be great to share and help many purchase the R2 without selling a kidney!

 

Also, especially if one decides to buy th 51M for scientific applications, I would advise not to buy more than the minimum 8GB of the slow overpriced RAM, and separately buy HyperX 2933 RAM instead - cheaper and 20% faster.

 

Mine with a OC seems to boost to 2130 it doesn't seem to make much difference to the score though.

 

Try running GPU-Z in the background and see if the core fluctuates or holds the set speed under the 'sensors' tab.

 

Looks to stay around 2100-2130 I wonder if the slow dell memory is making such a big difference then?

 

your memory cannot be slower than my crappy 2666 mhz Micron memory which runs @ 2400 MHz. In your Alienware Command Center under GPU, is there a GPU temp slider you could adjust? what's the current temp value set to?

 

How does the card behave if you are to lock it to 2100Mhz at 1.063mV/1.050mV through Afterburner?

 

The desktop version can draw 150 Watts. The laptop one is firmware limited to 65W sustained and it boosts to like the high 80s

It still crushed the competition in multi-core benchmark, and was close enough in single core. And this is only with a Ryzen 3900. You can get the 3950X in it!

Check out the tests on NotebookCheck. They do all the CPU ones from:
Cinebench R15 | Blender 2.79 | 7-Zip 18.03 | Cinebench R20 | Geekbench 5.1 / 5.2 | HWBOT x265 Benchmark v2.2 | LibreOffice | R Benchmark 2.5

https://www.notebookcheck.net/Schen...w-Extremely-fast-extremely-loud.464178.0.html

As not to invade the Ailienware forum, you can discuss the owner's impressions and improvements to noise and whatnot here:
http://forum.notebookreview.com/thr...zen-3950x-clevo-nh58af1-owners-lounge.832568/

 

So my R2 has shipped from DELL and is on the way to HIDevolution now to do the upgrades (Silicon Lottery 10900K + Delid + thermal pads). I got my laptop's service tag and after checking the service tag, I finally was able to find out what SSD is Dell using for the 2TB (2x 1TB PCIe M.2 SSD) RAID 0 [Boot] + 2TB (2x 1TB PCIe M.2 SSD) Non-Raid [Storage] configuration which comes under part number 400-BJHO. It is a Toshiba XG6 NVMe SSD. Review: https://www.anandtech.com/show/13254/the-toshiba-xg6-1tb-ssd-review-first-96l-3d-nand

Not the greatest but way better than any crap they used from SK Hynix before such as the SK Hynix PC400 NVMe SSD that came with my R1

@etern4l @Fire Tiger @Melvin Rousseau @Normimb @normand668 @ssj92 @MogRules

The RAM brand I still can't figure out but it's this one: https://www.dell.com/en-my/work/sho...-2rx8-ddr4-sodimm-3200mhz/apd/aa937596/memory

 

Unfortunately, the CL is not stated in the RAM "specification" section, which means it will be whatever is cheapest to deliver :/

 

Expect JEDEC timings.

 

What does that mean though?
For 2933, I can see V, W, Y, and AA variants ranging from CL19 to 22:

https://en.m.wikipedia.org/wiki/DDR4_SDRAM#JEDEC_standard_DDR4_module

 

My guess is the suffix is regarding the module size, ie. 8GB or memory IC.

 

Nope, see these two "V" variants:

https://www.serversupply.com/MEMORY/PC4-23400/32GB/SAMSUNG/M393A4K40CB2-CVF.htm
https://www.serversupply.com/MEMORY/PC4-23400/16GB/SAMSUNG/M393A2K40CB2-CVF.htm

Maybe they just specified different "standard timings" to accommodate a broader range of underlying RAM ICs. Dell's CL21 would appear to be is DDR4-2933Y: second from the bottom, so the good news is they could have done even worse. Yay!

 

Whether the 2933 or 3200 XMP Dell stock RAM, it will be interesting again to know whenever buying HyperX is warranted (almost certainly for the 2933 I guess). On another note just bought a 1TB Samsung Evo 970 Plus and basic small form ssd nvme heatsinks. That, my spare 330W adaptor and 17inch Vindicator backpack are looking forward to the beast arriving!

 

Or they might be JEDEC standarts for different dimm types ie. RDIMM (ECC) // UDIMM (Unregistered) // LRDIMM (Load-Reduction/High Density) // SODIMM

 

Dunno bro. Could be also that the JEDEC timings are the minimum required by the standard. Here are HyperX Impact timings

TIMING PARAMETERS:
JEDEC/PnP: DDR4-2933 CL17-19-19 @1.2V
XMP Profile #1: DDR4-2933 CL17-19-19 @1.2V
XMP Profile #2: DDR4-2666 CL16-18-18 @1.2V

http://www.kingston.com/dataSheets/HX429S17IB_32.pdf

So, the way I read that is: will run at CL17 in JEDEC mode.

 

Didn't you get Frank's last memo: it's bad boy!