I am all for forward progress, but I was certainly hoping to see more of it with these new chips considering the two years they'vehhad since SB-E surfaced.
Then again, it's almost unheard of for a processor to hold its ground as the best enthusiasts CPU for two years as the 3930K has done. Don't be fooled by stock benchmark results, I have been benching SB-E since release and the 3930K is within a single percentage point of the 3960/3970 once overclocked.
Furthermore, the 3930K I have now, the one that was a true keeper, has the strongest IMC I've personally seen on an SBE chip, and on average the 3930K has shown to have an extra 133-200Mhz for the memory than the X chips I've had. Also, the K's have, in general, been the best overclocking of the SBE chips in terms of voltage to speed. My 3930K (RIVE + 16GB Ripjaws Z 2133 9-11-10 or Trident X 2400 9-11-11 or 2666 10-12-12) is running at 4.98Ghz 24/7 w just 1.365v (1.165v for VCCSA/VTT1), will run BCLK as high as 108.4 on water, and has my Ripjaws Z running DDR3-2540 10-11-10-26 1t 24/7 w 1.665v vDIMM. It will hit 5.39Ghz on water with 1.51v, and will push memory as high as DDR3-2688 10-11-11-32 1t for benching w 1.715v vDIMM and 1.185v for both VCCSA/VTT1.
I know how the Silicon Lottery works, and that my chip is not exactly "average", but if you have even a half decent 39xx I see no benefit to the high upgrade cost.
The only way I would have gone for one of these would have been if they released a new, and truly worthwhile, chipset alongside the processors.
At this point in time, that means:
- 4way SLI/CFX @ x16/x8/x8/x8
- 2x x1 slots or maybe an x1 and an x4
- Improved VRM design, using top notch VRMs as well as increasing count by 50-100 percent to eliminate the thermal throttling
- Native USB3.0 supporting up to 10x rear IO ports + 3-4x headers
And most importantly, a storage setup worthy of the fastest consumer chips ever...
That would be something like:
- 6-8x native SATA6
- 4-6x native SATA3
- 2+1x rear/front eSATA6
- 1-2x mSATA or M.2 6Gbps/12Gbps ports
- Dual Intel GbEth with teaming and AP natively
- 4-6x USB2.0 rear
- 6-12x USB2.0 via headers
And if they really wanted to be more than 100 percent certain that their platform will only be beaten by its successor...
- RAID Dual-Core SoC from the new Adaptec PCIe3.0 cards
- 8-12x SAS/SATA6g ports
- 4x SAS12Gbps ports if only 8@6G
- Upgradeable Cache using a single SO-DIMM DDR3-1333/1600/1866 module that slots right in next to the PCH
- Support for up to 8GB cache w write-back caching available without a BBU but instead uses a Capacitor/NAND backup for write dumps
Make it interoperable with the SATA6 ports, specifically for the purpose of using one or more for SSD's caching the array/arrays.
Make mandatory extra power inputs, following Asus' lead, specifically:
- 2x 8pin EPS
- 24pin ATX
- 2x 8/6pin PCIe for extra Lane regulation
- 1x FDD + 1x SATA Power specifically for memory
- 2x SATA Power for the PCH and RAID SoC
Mandatory in that they must be implemented on retail boards, not that they all be connected for the board to function.
The PCIe and memory would be for SLI (or to enhance voltage stability for single cards when overclocking) while the memory plugs each functioning separate (first one to be used if all 8 DIMMs are filled or 4 are filled with 8GB modules, and the secondone is ffor overclocking and provides extra voltage and stability).
The PCH ones are to provide better stability for BCLK tweaking as well as providing a direct power source for the RAID SoC and it's cache. Without the latter, you would not be able to use write-back caching and the array would would obviously lose speed. The plug would also enable a "Turbo-RAID" option if connected, which would automatically increase the speed of the dual core RAID SoC from 650Mhz to 900Mhz, automatically enable W-B caching, and essentially increase the max potential IOPS and transfer rate.
You would want to have some voltage regulation and power cleaning (VRM sections) for each input, keeping crosstalk interference out and, as the point of the connections is to boost VR to components, it only makes sense to ensure that the power coming onto the board itself is immediately made perfect.
Point is to "parallelize" the power coming to the board as much as possible to enhance stability, important for regular use but critical for OC'ing (especially subzero).
Basically, just because Intel is currently producing the first, second, third, fourth....twentieth, twenty-first, etc fastest processors for consumers and enthusiasts, doesn't mean that they shouldn't continue to offer the best possible cuttingest-edge components, and where better to prove that than with the absolute fastest platform on the market?
Unfortunately, we didn't get any of those things. I say unfortunately, but it's actually bitter sweet.
Sure, Z87 has a lot of connectivity, and plenty of gimmicks, but unless you are a RAM bencher (meaning that's all you do really and you are in the top 10 or 20 on the bot), Haswell is frankly disappointing. It's not even close to as big a flop as the AMD FX chips were/are, but SB-E remains the undisputed king, losing to Haswell only in a few, heavily single threaded marks.
Having spent an immense amount of time with every platform and major chip from AMD and Intel since the Core2Duo/Core2Quad/Athlon II/Phenom II days (actually, still have my first batch FX-51 from a decade ago), and gamed, edited video and photos, done a great deal of my audio engineer work (record, edit/mix, master; I master in 24bit 192Khz, and downsample depending on the artist's distribution preferences), and of course done a ton of benching (air, custom water, custom chilled water, and even sub-zero air w the PC outside in as low as -25*F weather)....
I would say that the only people who would find IV-E a worthwhile upgrade are those with a Sandybridge system or earlier on the Intel side (especially if the SB is an i5), anyone running anything AMD, and anyone who is going to run multiple HIGH END GPU's (2+ 580s, 670s, 680s, 770s, 780s, Titans, 7950s, or 7970s) and doesn't currently have a SB-E system.
If you have a good IVB or even a higher end SNB rig, much less a SB-E rig, the gains are limited to either the natural performance increase from a 50pct core count increase and nearly doubled memory bandwidth (SB/IVB), the extra PCIe lanes that, despite what you have heard, actually are vastly, immensely better than PLX pretend-lanes (anyone but SB-E), and the improvements made to the IMC despite being rather disappointing considering the potential to perform much better than what they actually gave (SB/SB-E).
Of course, if you just want to make a really awesome, no holds barred workstation but want to stick with a single CPU, then you can not do better than something like...
4960X @ 4.5Ghz
Asus P9X79-E WS Motherboard
32-64GB Trident X DDR3-2666 10-13-13-32
3x Titans or 3x 780 Classy/Lightning/HOF
Adaptec 8-series 16-port or 24-port x8 PCIe3.0 RAID card w all extras
8x OWC 256GB Enterprise SSD's in RAID0
1x Samsung 840Pro 512GB for OS
4x Samsung 840EVO 1TB SSD's in RAID0
8x WD RE 4TB in RAID5
2x Plextor M5P Xtreme 512GB for SSD caching
4x WD VR 1TB RAID0
Intel PCIe 4x10GbEth NIC
12-bay External RAID Box w SAS expansion board
6x WD10EZEX 1GB RAID0
4x Seagate Constellation 3TB in RAID10
For mere mortals, tho, I'm going to say that IV-E is a "pass"
Facebook
Twitter
Instagram