What's new

Surface Pro 2017 Benchmarks (and Anomalies)

3DMark 2006:
3dm06.png


3DMark Vantage:
3dmvantage.png

3DMark11:
3dm11.png


3DMark Ice Storm:
3dmicestorm.png

3DMark Cloud Gate:
3dmcloudgate.png


3DMark Sky Diver:
3dmskydiver.png


3DMark Fire Strike:
3dmfirestrike.png


SuperPI:
superpi.png


wPrime:
wprime.png


HDTune SSD 1:
hdtune1.png
 
OP
M

mimarsinan

Member
HD Tune SSD 2:
hdtune2.png


Now, the anomalies:

1. With HDTune, at least this favorite version of mine, when you try to benchmark the Storage Spaces drive, the transfer rate remains the same (which I found surprising if this is really a raid stripe), but the process fails with a read error at exactly the middle of the operation. Probably some kind of false positive due to incompatibility with the Storage Spaces tech and the legacy HDTune version.

2. The benchmarks are an improvement...until a certain point, which is the main anomaly:
Untitled1.png

3DMark 11, Cloud Gate, Sky Diver, and Fire Strike are all major regressions in performance over the Surface Book 4, while the other numbers and benchmarks (including SuperPI and wPrime and HDTune, which are not featured on this chart) are substantially better.

Can anyone else test and post their own numbers? Maybe I have a funny CPU or something and need to RMA the unit? Although I am skeptical of that, given improved performance in other areas.

Truly a strange scenario. I had seen similar benchmark results on other sites comparing the Pro 4 and the Pro 2017, but some of these numbers are really bad - and what I'd seen elsewhere, IIRC, featured better numbers for the more recent benchmarks but not the legacies (which is actually the exact opposite of the pattern I have here).
 

GreyFox7

Super Moderator
Staff member
HD Tune SSD 2:
View attachment 8634

Now, the anomalies:

1. With HDTune, at least this favorite version of mine, when you try to benchmark the Storage Spaces drive, the transfer rate remains the same (which I found surprising if this is really a raid stripe), but the process fails with a read error at exactly the middle of the operation. Probably some kind of false positive due to incompatibility with the Storage Spaces tech and the legacy HDTune version.

2. The benchmarks are an improvement...until a certain point, which is the main anomaly:
View attachment 8636
3DMark 11, Cloud Gate, Sky Diver, and Fire Strike are all major regressions in performance over the Surface Book 4, while the other numbers and benchmarks (including SuperPI and wPrime and HDTune, which are not featured on this chart) are substantially better.

Can anyone else test and post their own numbers? Maybe I have a funny CPU or something and need to RMA the unit? Although I am skeptical of that, given improved performance in other areas.

Truly a strange scenario. I had seen similar benchmark results on other sites comparing the Pro 4 and the Pro 2017, but some of these numbers are really bad - and what I'd seen elsewhere, IIRC, featured better numbers for the more recent benchmarks but not the legacies (which is actually the exact opposite of the pattern I have here).
First retest, then double check your setup to make sure you don't have background processes interfering with benchmarks. I've seen sites publish benchmark results from a unit just out of the box when it's clearly doing updates, indexing, optimizations, and maintenance.
 
OP
M

mimarsinan

Member
As the HDTune and raw CPU benchmarks indicate, no background tasks were ongoing at the time of benchmarking - everything is fully updated, indexed, and so forth. The raw CPU benchmarks also indicate that the performance of the CPU is above the previous generation.

However the latter GPU tests are worrysome indeed. Does anyone else have any numbers they may publish? This might help triangulate the issue. Thanks!
 

Mike007

Member
Your 3Dmark06 score is way lower than mine is. I am testing the others right now and will post back with results shortly

y4mxVR8exVrjxll0KrpT0Fd-qAp2QDuZ5fPbmRoCWOUYc_hSqyGKH0ndd-OrlBl5Uwe1g_vXmtQrujZgOTQ3DnViTyrjrDroD-yWT4TpaG8aoquavxCwB33tBjliJ9Omedk-GS-pWSKGvqgw3SRSY309e12XF97PX9gFSWnnwP_vgFl4jBYRmz3tkTv7Ux6lp7rx5bKIfwGvHJQ4vNEGpWoCw
 

jabbasoft

New Member
@Mike007 does PassMark (my post) use the same test routine as 3DMark? I was about to download 3DMark to run a comparison but its 3.5GB and I have to get back to work, will try later ;)..
 

Mike007

Member
@Mike007 does PassMark (my post) use the same test routine as 3DMark? I was about to download 3DMark to run a comparison but its 3.5GB and I have to get back to work, will try later ;)..

I think its an entirely different program. I haven't actually used passmark yet, will check it out.

You can get the free versions of 3Dmark06, Vantage and 3dmark11 from futuremark's site. (Look under legacy benchmarks)
These have a much smaller filesize than the newer one.
 
Top