Overclock.net › Forums › Industry News › Hardware News › [hardocp] Ivy Bridge - E review
New Posts  All Forums:Forum Nav:

[hardocp] Ivy Bridge - E review - Page 35

post #341 of 344
Quote:
Originally Posted by Cyro999 View Post

It's a benefit in Crysis 3, AFAIK

So it was said, but my experience with HT and Crysis 3 on a 2600K suggested otherwise.
Broadwell-EP
(11 items)
 
  
GraphicsHard DriveHard DriveHard Drive
MSI GTX 960 2GB Samsung 840 Pro (256GB) Samsung 840 Pro (512GB) Samsung 840 Pro (512GB) 
CoolingOSCaseAudio
Noctua NH-U12S Windows 7 Ultimate x64 (SP1) Silverstone FT02 Asus Xonar Essence STX 
  hide details  
Reply
Broadwell-EP
(11 items)
 
  
GraphicsHard DriveHard DriveHard Drive
MSI GTX 960 2GB Samsung 840 Pro (256GB) Samsung 840 Pro (512GB) Samsung 840 Pro (512GB) 
CoolingOSCaseAudio
Noctua NH-U12S Windows 7 Ultimate x64 (SP1) Silverstone FT02 Asus Xonar Essence STX 
  hide details  
Reply
post #342 of 344
Quote:
Originally Posted by Cyro999 View Post

HT on helps some stuff, hurts others. It's a benefit in Crysis 3, AFAIK
Helps on most.
Quote:
Originally Posted by Oubadah View Post

So it was said, but my experience with HT and Crysis 3 on a 2600K suggested otherwise.

Crysis 3 uses HT effectively as of patch 1.3, so usually an FX beats an i5 but an i7 beats an FX by a much larger amount.
post #343 of 344
Thread Starter 
I see that Crysis 3 and HT argument quite a lot. I'm not so sure about that.
Quote:
Modell Version 1.0 Version 1.1 Version 1.3
Core i7-3770K 56,7 Avg-Fps 51,3 Avg-Fps 74,9 Avg-Fps
FX-8350 58,6 Avg-Fps 58,7 Avg-Fps 57,4 Avg-Fps
Core i5-3570K 50,5 Avg-Fps 52,6 Avg-Fps 54,7 Avg-Fps
FX-6300 46,4 Avg-Fps 46,3 Avg-Fps 45,1 Avg-Fps
So lets do analysis again.

First lets renormalize i5 to i7. These values are wrong because they didn't disable boost, but it helps with analysis.
56.7 vs 51.3 vs 74.9
51.98 vs 54.147 vs 56.3

Now lets do the same for AMD.
58.6 vs 58.7 vs 57.4
53 vs 52.9 vs 51.54

It's interesting that even renormalized FX-6300 falls behind FX-8350. Perhaps it's result of FX-8350 having two more decoders, because it has one more module. Then again it might be caused newer revision of AMD chip.

Now lets look at Intel's results. The first column looks as expected. 3770K should be faster than 3570K, right? Right? Patch 1.1 follows and we seen 3770K to be WORSE than 3570K. Note that AMD speed didn't change. Third column shows massive speed increase for 3770K.
On the other hand AMD shows increase in the second column, and slow down in the third column.

This more looks like they improved something and HT hurt them bad, because patch 1.1 has speed increase for all CPU except the CPU with HT. They corrected it in patch 1.3, but the correction hurt non HT CPUs, because AMD CPU slowed down. I assume there were some speed optimizations which helped more Intel CPUs thus 3570K had net gain, and the AMD slow down can be explained by a net speed loss caused by combination of code that tried to solve critical fault caused by HT and normal speed improvements which were linear through these patches..
post #344 of 344
I was on a long drive yesterday and listening to YouTube. For a little laugh 1:15

watch?v=awYbOQLZnn8

http://m.youtube.com/watch?v=awYbOQLZnn8
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [hardocp] Ivy Bridge - E review