Why not faster?

JasonStevens

Reputable
Jul 21, 2014
9
0
4,510
Hi there,

Why would Build 1 below only be 30% faster than Build 2 at a Visual Studio 2013 compile - 25 seconds vs 35 seconds respectively?

Build 1 has a 20% faster CPU + turbo, 2 more cores + hyperthreading, Sandy Bridge-E rather than Yorkfield, and 4x the memory. In the Passmark benchmark, the 3930K outperforms the Q9550 by 300%.

In both cases disk access doesn't seem to be a problem, all CPU cores/threads are used and busy (70-80% in Build1 and 75-90% in Build2). Build2 has significantly higher page faults, but otherwise they're about similar in Resource Monitor.

I'm about to upgrade from Build 2 and want to see if it's possible to do a little better than Build 1. I don't mind spending a little more but not go too crazy.

Any advice would be much appreciated. Thanks!

Build 1
Intel Core i7 3930K @ 3.20GHz
32GB DDR3 @ 770MHz (11-11-11-28)
P9X79 (LGA2011)
232GB Samsung SSD 840 EVO 250GB

Build 2
Intel Core 2 Quad Q9550 2.83 GHz, 1333MHz FSB
Corsair XMS3, 4x2GB, DDR3-1333 (9-9-9-24)
Asus P5E64 WS, Socket 775, 1600MHz FSB, 4xDIMM DDR3
OCZ-Vertex2 SSD
 

JasonStevens

Reputable
Jul 21, 2014
9
0
4,510
I turned off virus scanning on both machines because I noticed that was doing some serious file intercepting during a compile.

However, this had the effect of slowing down the compile time on each machine! I believe because the drive queue length increased and caused serious congestion. I figure that drive IOPS is (still) very important so I'll focus my build around that by getting a Xeon + PCI-e SSD. I'll post how it goes.

 

JasonStevens

Reputable
Jul 21, 2014
9
0
4,510
Not really an answer so much as the results of my build.

I found that the PCIe SSD does the business for my needs. The Xeon probably contributes because of its on-board PCIe I/O (30% better they say).

i7 3930K (6 core, 3.2GHz), 32Gb RAM, Samsung SSD 840 EVO > 25s
E5-2650 v2 (8 core, 2.8GHz), 16Gb RAM, Samsung SSD 840 EVO > 25s
E5-2650 v2 (8 core, 2.8GHz), 16Gb RAM, OCZ RevoDrive 350 PCIe SSD > 20s

So an improvement of 5s (20%) by switching to a PCIe SSD. The Xeon is running easy during the compile although it does wind up to full turbo. It peaks at 80% but in general it's ~35% over the build. By comparison, the i7 3930K is going like the clappers most of the way through.

--
Fractal Design Define R4
Asus P9X79-E WS
Intel Xeon E5-2650 v2
G.Skill Ripjaws Z Series DDR3 1866MHz CL9 4x4GB
Corsair AX760i
Samsung 840 EVO 250GB
OCZ RevoDrive 350 240GB PCI-e x8