Thoughts on Haswell-E, X99 and the Slowing Refresh Cycle

With the recent release of both X99 and the 8 core 5960x, the old familiar upgrade itch started up once again.  Currently still on 3960x 6 core @ a healthy 4.7Ghz and an Asus P9x79WS x79 gen 2 board with 1TB Crucial M4 SATA3 in RAID 0 and tri-SLI GTX Titans in PCI-E 3 mode, I can say that I never feel like the system is resource limited anywhere with the exception of possibly very large stream video encoding and gaming at ultra high resolutions as always.  I recently settled in on a single 1440P 27″ IPS panel (QNIX) overclocked to 120Hz and 1 accessory display.  Even with the above rig 120fps cannot be sustained at maximum details.  So there is room for increased performance.  Given this reality, is it time to sell a kidney and make the giant move to an entirely new platform (including motherboard and RAM this time)?  At first it all signs seemed to point to YES.  From Anand, to Toms to [H]ard, to PCPER, to… well everyone really, the reviews were glowing.  And rightly so!  Haswell-E/x99 is absolutely a great platform, fully modern, and as “future proof” as any PC can be (“not very” essentially, but good enough).  So why isn’t this a build thread then?

Well it probably comes as no surprise given the intro and title, but I’ve made the executive decision on X99/Haswell-E and… its a skip! Very interesting that even for a lunatic like me hardware refresh is slowing. Its a combination of mobile devices and consoles curving the development target down from the absolute pinnacle and hardware capabilities having outstripped most workloads for a couple of generations now.

For gaming at 1440P and above we remain heavily GPU bound so that’s a wash. The 8 Core package is great, but very limited desktop workloads are that threaded, so value over a 6 core is limited. Heat and power have gone up for Haswell-E vs IvyBridge-E and SandyBridge-E so even though IPC has gone up in increments of 5-8% through these 3 gens, overclock ceiling has gone *down* by about 10%. A net gain of 3-5% aggregate CPU performance isn’t terrific.

On the chipset front late edition x79 brought us PCI-E 3 (40 lanes on SB-E and IB-E), SATA3 and USB3 (OEM integrated but it works fine) and support for up to quad x16 GPU via PLX chips (something you would still need on x99 as the Haswell-E chips also top out at 40 lanes). Compounding the issue is the move to DDR4 which brings higher bandwidth and future potential, but also higher latency. The current DDR4 modules are running with CL15 timing vs the CL9 I run on x79 (and thats very conservative). M.2 and SATA Express are nice to haves, but with SATA3 RAID 0 pushing 1.2GB/s real world perf for me, I rarely feel like disk is any kind of bottleneck.

Final conclusion is that for new system builds, or for folks on X58, the platform is fantastic, but for Sandy and Ivy upgraders, unless you need 8 cores (*really* need), I’d say skip it.


One thought on “Thoughts on Haswell-E, X99 and the Slowing Refresh Cycle

  1. Pingback: Project Destiny: Upgrading the 20GB PlayStation 3 Phat HD | Complaints Incorporated...

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s