Monday, August 7th 2017

Alienware Area 51 Ryzen Threadripper System Benchmarked

LinusTechTips has received an Alienware Area 51, the boutique system which Alienware has developed through its exclusive partnership with AMD for the 16-core, 32-thread Threadripper CPUs.

The system was tested using a proprietary Dell/Alienware motherboard, as well as 32 GB (4x 8 GB) of 2666 MHz DDR4 RAM. Graphics were handled by a Founders Edition GTX 1080 Ti, and storage was etched on a SanDisk A400 256 GB NVME drive (Up to 2050/700 MB/s Sequential Read/Write). The nature of the benchmarked systems, with their myriad of hardware parts, means that the benchmarks are not 100% representative of CPU performance deltas between the tested processors; however, they do put out some interesting numbers. The other HEDT system to be tested was an Intel-based X299 system with a stock Intel Core i9-7900X (10 cores, 20 threads) on an ASUS Prime X299 Deluxe, 4x 8 GB of 3200 MHz DDR4 memory, the same GTX 1080 Ti Founders Edition, and a 512 GB Samsung 950 Pro NVME SSD.
With the tested hardware, the Threadripper system delivers 68 maximum and 63 minimum FPS, which puts its slightly ahead of the Ryzen 7 system, and slightly behind, in minimum framerates, compared to the Intel systems (67 minimum for the X299 platform, and 65 for the i7-7700K powered Z270.)
3D Mark is a benchmark which doesn't scale linearly according to the number of cores - not even on its CPU test scenario. Instead, this benchmark is optimized for the highest throughput of a given architecture on the maximum number of threads it can actually spawn. This is probably why the Threadripper system delivers slightly better performance than the Ryzen 7 system on the CPU score (though not as much as one might imagine), while the X299 Core i9-7900X steals the show with higher scores across the board and  ~2,200 points extra score on the CPU benchmark - despite having six less cores and 12 less threads than the Threadripper system.
There's not much of a surprise on the Cinebench benchmarks - this is a benchmark where the Ryzen architecture shines through and through, and it shows here - not even considering the thread advantage of the Threadripper 1950X.
The same case can be made for the Blender benchmark, naturally.
What do you think of these Threadripper 1950X benchmarks? I'd say these are certainly competent enough to allow users to look at the Ryzen architecture's other strengths.

Check out the LinusTechTips video:


Source: Linus Tech Tips
Add your own comment

35 Comments on Alienware Area 51 Ryzen Threadripper System Benchmarked

#1
btarunr
Editor & Senior Moderator
Yeah, run the Threadripper machine at DDR4-2666 and the CoreX machine at DDR4-3200 (instead of the other way around). Expected of Linus.
Posted on Reply
#2
ShockG
Getting ThreadRipper to run 3200 Quad Channel is not as straight forward right now as it is on AM4.
The most reliable DRAM frequency right now is actually 2666MHz. Higher frequencies on the IMC also lower the CPU clocks somewhat (and depending on cooling, drastically). Memory frequency and fabric frequency also don't scale performance much after 2,666MHz in test like Cinebench (or at least Cinebench specifically)
Posted on Reply
#3
the54thvoid
btarunr said:
Yeah, run the Threadripper machine at DDR4-2666 and the CoreX machine at DDR4-3200 (instead of the other way around). Expected of Linus.
Perhaps the TR board couldn't handle faster frequency memory? It's not unknown that Ryzen still, even with new BIOS is temperamental with frequency.
Posted on Reply
#4
dicktracy
GPU bottleneck for a CPU test. Heh screw that AMD review guideline.
Posted on Reply
#5
Ubersonic
ShockG said:
Getting ThreadRipper to run 3200 Quad Channel is not as straight forward right now as it is on AM4.
According to the Dell/Alienware product page for the Area 51 it's ThreadRipper board only supports dual channel anyway.
32GB Dual Channel DDR4 at 2667MHz; up to 64GB (additional memory sold separately)
Posted on Reply
#6
uuuaaaaaa
Ubersonic said:
According to the Dell/Alienware product page for the Area 51 it's ThreadRipper board only supports dual channel anyway.
This Alienware system is overheating, underperforming and overpriced, what a joke.
Posted on Reply
#7
Ubersonic
uuuaaaaaa said:
This Alienware system is overheating, underperforming and overpriced, what a joke.
Sorry can you be specific? You just described pretty much every Alienware product of the last 20 years :P

(Except the M11x, that was awesome).
Posted on Reply
#8
wiak
Ubersonic said:
According to the Dell/Alienware product page for the Area 51 it's ThreadRipper board only supports dual channel anyway.
thats why blender score is lower than what amd has shown, ie memory run at lower feq and dual instead of quad channal
Posted on Reply
#9
Rehmanpa
Why all the hate on linus?
Posted on Reply
#10
uuuaaaaaa
Rehmanpa said:
Why all the hate on linus?
This has been touted as the first ThreadRipper review, but it really is an Alienware product review that is not representative of the real ThreadRipper performance, i.e. 2666MHs ram in dual channel instead of 3200MHz quad channel + overheating cpu (single 120mm AIO LC was used).
Posted on Reply
#11
Ubersonic
uuuaaaaaa said:
This has been touted as the first ThreadRipper review, but it really is an Alienware product review that is not representative of the real ThreadRipper performance, i.e. 2666MHs ram in dual channel instead of 3200MHz quad channel + overheating cpu (single 120mm AIO LC was used).
So, basically all the stuff Linus complained about, derided Dell/Alienware over and said made this non indicative of an actual Threadripper review?
Posted on Reply
#12
uuuaaaaaa
Ubersonic said:
So, basically all the stuff Linus complained about, derided Dell/Alienware over and said made this non indicative of an actual Threadripper review?
I meant that there is the danger of people taking this Alienware product as a true 100% representation of the actual 1950X true performance while in reality is a worst case scenario, i.e. dual channel 2666MHz ram + overheating.
Posted on Reply
#13
TheDeeGee
$1000 extra for the Brandname Alienware?
Posted on Reply
#14
Sempron Guy
How is 65c overheating again? 3dmark Timespy CPU score seems suspect though unless Timespy can use up to 16 threads?
Posted on Reply
#15
uuuaaaaaa
Sempron Guy said:
How is 65c overheating again? 3dmark Timespy CPU score seems suspect though unless Timespy can use up to 16 threads?
Wasn't it up to 85ºC?
Posted on Reply
#16
Sempron Guy
says 65c on the graph or am I missing something?
Posted on Reply
#17
Ubersonic
Reported TCTL temp sensor was 92c, they guessed at 65c "real" temperature as they had no way of testing it.
Posted on Reply
#18
uuuaaaaaa
Sempron Guy said:
says 65c on the graph or am I missing something?
I was referring to the temperatures before offset, my bad... However, Linus said that the cooler struggled to keep up with the cpu even after they changed thermal paste.
Posted on Reply
#19
Sempron Guy
Ubersonic said:
Reported TCTL temp sensor was 92c, they guessed at 65c "real" temperature as they had no way of testing it.
guessed or AMD informed them about the offset? either way the video's rushed and riddled with flaws. For a "tech channel" if that's what you call it these days, with millions of subscribers that's inexcusable.

uuuaaaaaa said:
I was referring to the temperatures before offset, my bad... However, Linus said that the cooler struggled to keep up with the cpu even after they changed thermal paste.
yes but it says otherwise in the graph unless the scriptwriter had it all wrong lol
Posted on Reply
#20
DarkHill
uuuaaaaaa said:
This has been touted as the first ThreadRipper review, but it really is an Alienware product review that is not representative of the real ThreadRipper performance, i.e. 2666MHs ram in dual channel instead of 3200MHz quad channel + overheating cpu (single 120mm AIO LC was used).
Its interesting though,

the AIO in question is a rebrand of the Arctic Cooling Liquid Freezer 120 which usually is comparable to most other 240mm due to its thick radiator (49mm vs 26mm for most other asetek 240mm clones).

This points towards that a normal 240mm asetek generic AIO will NOT be sufficient to cool the threadripper platform (atleat the 1950x). The picture might change somewhat if push-pull are used, but i doubt it.

The real problem is more likely the heat-exchanger area might be to small to efficiently draw the heat away from the 4x4 CCX's
Posted on Reply
#21
uuuaaaaaa
DarkHill said:
Its interesting though,

the AIO in question is a rebrand of the Arctic Cooling Liquid Freezer 120 which usually is comparable to most other 240mm due to its thick radiator (49mm vs 26mm for most other asetek 240mm clones).

This points towards that a normal 240mm asetek generic AIO will NOT be sufficient to cool the threadripper platform (atleat the 1950x). The picture might change somewhat if push-pull are used, but i doubt it.

The real problem is more likely the heat-exchanger area might be to small to efficiently draw the heat away from the 4x4 CCX's
That might be the case, custom cooler base plates are needed for Threadripper, that thing is immense.
Posted on Reply
#22
Raevenlord
News Editor
DarkHill said:
the AIO in question is a rebrand of the Arctic Cooling Liquid Freezer 120 which usually is comparable to most other 240mm due to its thick radiator (49mm vs 26mm for most other asetek 240mm clones).
One of the best price/performance AIOs out there. Can't complain at all.
Posted on Reply
#23
Rexter
I find their graphs hilarious!



Just look at how inaccurate it is.
Posted on Reply
#24
DarkHill
Raevenlord said:
One of the best price/performance AIOs out there. Can't complain at all.
for the price its absolutely exceptionel, i use one on my G12 bracket for my Asus 1070 strix
Posted on Reply
#25
springs113
Ubersonic said:
Sorry can you be specific? You just described pretty much every Alienware product of the last 20 years :p

(Except the M11x, that was awesome).
My 15r n 13r3 begs to differ.
Posted on Reply
Add your own comment