Ivy or Sandy in new build; PCI-E 3.0

dbhelman

Distinguished
Feb 4, 2012
13
0
18,510
Hello. I have been researching a new build. I understand that a SB would be plenty of power, but I like the efficiency of IB. It would seem stupid of me not to wait for IB when it's so close and the prices will be equivalent to SB. I also want to know more about PCI-E 3.0 in IB. I hear it won't be useful for a while, but it seems worth while to go ahead with IB. Thank you.
 
At this point in time if you can wait, then wait. Yes PCI-e 3.0 is not really necessary now and maybe not be necessary until the Radeon HD 9000 series or GTX 800 comes out. Maybe useful if you XFire / SLI high end Radeon HD 8000 or GTX 700 series video cards; just a guess.

Anyway, there should be some increase in CPU processing performance. How much is unknown. I'm guessing around 6% 'cause Intel seems to be focusing more on improving the integrated graphics core performance of the Intel HD 4000 (supposedly up to 60% better than the Intel HD 3000), and focusing on getting the die shrink and tri-gate transistor design correct. Besides, Intel is working on Haswell (next CPU architecture) while all this is going on. Other's think up to 20% because they read an article published back in Jan/Feb 2011 stating that performance increase is expected to be up to 20% better. Of course, during that same time there were a couple of articles stating that Bulldozer would offer 50% more performance than the 1st generation Core i3/i5/i7 CPUs.

 
Radeon 7000 has PCIe 3.0 and Ivy's main performance advantage over Sandy should be how far it will overclock. PCIe isn't necessary but it can be helpful now if you have an X78 system with Radeon 7000 cards. It can be helpful later with Radeon 7000 cards and Ivy. Nvidia hasn't revealed whether or not GTX 600 cards also have PCIe 3.0, but they probably do.

like the above poster said, Ivy isn't much faster than Sandy at stock but it uses much less power so Ivy has great overclocking potential. Ivy will NOT be 20% faster than sandy, Intel has confirmed this many times. Intel claims 10-15% performance increases over comparable Sandy CPUs, but that is probably because of both the improved IPC and slightly higher stock clock rates.

Haswell could be as much as 20% faster than Sandy, maybe even more. It's important to remember that Ivy is mainly just a die shrink of Sandy with updated connectivity and a better IGP. Haswell is a whole new micro-architecture, not a die shrink. It should be like the difference between Westmere/Guftown and Sandy.

There may have been articles claiming 50% performance difference between Bulldozer and Nehalem, but they were probably fan hype, not directly from AMD. Of course it turns out that Intel's chips are about 40-50% faster at the same clock rate so it wasn't to far off if you replace i3/i5/i7 and FX in that claim.
 

triny

Distinguished
Feb 2, 2012
450
0
18,790
If they increase IGP to dx11 a four core IB would probably be less powerful than a 2500k ,In fact I'd say 2500k will still be the best choice for gaming 2600k for bussiness. IB is a about igp for the upper mobile market.

 

triny

Distinguished
Feb 2, 2012
450
0
18,790


FX is proably 50% faster in some benchmarks
 


Completely wrong. What does DX11 compatibility have to do with decreasing CPU performance? Absolutely nothing. The Ivy replacement for the 2500K, I think it's the 3570K, is faster than the 2500K in every way and uses about 20% less power. IB's replacement for the 2600K/2700K is also faster than both in every way and uses about 20% less power.




FX is slower than i5s in everything that uses less than 8 threads and slower than i7s in everything. They are also slower than i3/SB Pentium in everything that uses three or fewer threads. Even though the FX eight core chips are faster than i5s in highly threaded workloads, it isn't by much, well under 50%. FX eight core chips are beaten by quad core i7s. It goes without saying that the hexa-core i7s shred FX, although I wouldn't compare the two because of the huge price differences.
 

TRENDING THREADS