- Oct 30, 2011
- 3,719
- 0
- 23,460
So I have 3 questions for a build I am planning for a friend that I need some help on. Need all your guys expert help for this 
The potential builds will be
2500k + Z68 $250 ish used (Buying used from a friend, proven OC to 4.8ghz on stock voltages)[at 4.8, it should match the 3570k and 4670k in terms of OC][mb only has 1.5 year warrant left]
3570 + z77 $290+tax (OC this to 4.0/4.1/4.2/4.2 using turbo)
4670 + B85 $300+tax (IPC roughly same as 3570 OC)[3.6/3.7/3.8/3.8]
3570k + z77 $310+tax (OC will be unknown, should at least match the 4670 IPC given, and likely higher in multi thread and with more OC)[4.2+ all 4 cores]
4670k + Z87 $350+tax (OC to 4.2 should be doable, highest performing option but also most expensive)[non z87 OC won't be an option I think][4.2 all 4 cores]
I will be getting a CNPS10X Optima regardless of which build as I get get them on sale for under $10+MIR and I detest the stock Intel cooler. Any budget saved will go to other parts. Chipset features are largely irrelevant. CFX and SLI are not under consideration for now.
Questions
1. Will the potential of new instructions such as AVX2 and FMA3 be utilized in games coming out within the next 7 years? (Replacement build likely around then)
I don't know about the adoption of AVX and SSE4/4.1/4.2 in games or do they use older code like SSSE3 and SSE3 still? So if there is a lengthy time period till adoption by devs it boosts the argument for not getting Haswell. If there will eb quick adoption, then Haswell be the better option.
2. Extrapolating from past GPU speeds in relation to PCIe, going the route of PCIe 2.0 instead of PCIe 3.0 should not be too much of a concern as the top end cards have barely saturated PCIe 1.1 x16 so PCIe 2.0 x16 should be good for PCIe 4.0 and potentially obsoleted at PCIe 5? This means future GPU upgrade should not be bottle necked by PCIe 2.0 on the z68 for the life time of this build?
3. Given that Haswell has removed the 4 free bins on top of the highest turbo bin, any non K cannot be over clocked. For the 4670, is there the option on B85 boards to change the turbo multi for 4 cores active from 36x to 38x or will it require an z87 board or is it simply impossible?
My conclusion for best price/perf assuming new instructions are not used and pcie 2.0 is still viable.
1. 2500k
2. 3570
3. 3570k
4. 4670
5. 4670k
If PCIe 3.0 is needed then choose on of the Ivy options. If new instruction sets will be rapidly adopted then Haswell wins. So the answer to those two questions will be key.
The potential builds will be
2500k + Z68 $250 ish used (Buying used from a friend, proven OC to 4.8ghz on stock voltages)[at 4.8, it should match the 3570k and 4670k in terms of OC][mb only has 1.5 year warrant left]
3570 + z77 $290+tax (OC this to 4.0/4.1/4.2/4.2 using turbo)
4670 + B85 $300+tax (IPC roughly same as 3570 OC)[3.6/3.7/3.8/3.8]
3570k + z77 $310+tax (OC will be unknown, should at least match the 4670 IPC given, and likely higher in multi thread and with more OC)[4.2+ all 4 cores]
4670k + Z87 $350+tax (OC to 4.2 should be doable, highest performing option but also most expensive)[non z87 OC won't be an option I think][4.2 all 4 cores]
I will be getting a CNPS10X Optima regardless of which build as I get get them on sale for under $10+MIR and I detest the stock Intel cooler. Any budget saved will go to other parts. Chipset features are largely irrelevant. CFX and SLI are not under consideration for now.
Questions
1. Will the potential of new instructions such as AVX2 and FMA3 be utilized in games coming out within the next 7 years? (Replacement build likely around then)
I don't know about the adoption of AVX and SSE4/4.1/4.2 in games or do they use older code like SSSE3 and SSE3 still? So if there is a lengthy time period till adoption by devs it boosts the argument for not getting Haswell. If there will eb quick adoption, then Haswell be the better option.
2. Extrapolating from past GPU speeds in relation to PCIe, going the route of PCIe 2.0 instead of PCIe 3.0 should not be too much of a concern as the top end cards have barely saturated PCIe 1.1 x16 so PCIe 2.0 x16 should be good for PCIe 4.0 and potentially obsoleted at PCIe 5? This means future GPU upgrade should not be bottle necked by PCIe 2.0 on the z68 for the life time of this build?
3. Given that Haswell has removed the 4 free bins on top of the highest turbo bin, any non K cannot be over clocked. For the 4670, is there the option on B85 boards to change the turbo multi for 4 cores active from 36x to 38x or will it require an z87 board or is it simply impossible?
My conclusion for best price/perf assuming new instructions are not used and pcie 2.0 is still viable.
1. 2500k
2. 3570
3. 3570k
4. 4670
5. 4670k
If PCIe 3.0 is needed then choose on of the Ivy options. If new instruction sets will be rapidly adopted then Haswell wins. So the answer to those two questions will be key.
Facebook
Twitter
Instagram