Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
AMD vs Intel - Can yall check my math?
#1
I've seen it written here that C2D, C2Q, i3, i5 & i7 Intel CPUs are, on average, about 20% faster than Socket AM3 AMD CPUs with PCSX2.
If that's the case, then the following chart should be accurate (I hope).

AMD 3.5 GHz = 2.8 GHz Intel
AMD 4.0 GHz = 3.2 GHz Intel
AMD 4.2 GHz = 3.4 GHz Intel
AMD 4.3 GHz = 3.6 GHz Intel
AMD 4.8 GHz = 4.0 GHz Intel

If I'm in error, please let me know.
My PCSX2 Rig:Windows XP x32 & Windows 7 x64.
i5 2500k @ 4.0 GHz.
nVIDIA GTX 460 1 GB.
>Dun Click Here<
Reply

Sponsored links

#2
Your math is wrong.

For example 3.2 * 1.2 = 3.84.

Look at the CPU benchmark too.
Reply
#3
I think it is roughly 3-400 compaing an Phenom or Athlon II to a c2d/q and 5-600 to an i3/5/7.
AMD Athlon II 245 @ 3.55ghz, 9600GT @730mhz 1115mhz, vista sp2 32bit
Reply
#4
It's not wonderful Inter reverted the picture copying AMD? Smile

Soon after the 486 era, AMD diverged from Intel creating it's own CPU architecture, one able to do more things by the same clock cycles. That was necessary, AMD couldn't compete in the raw clock speed at those times. It bought Alpha and prepared to start the 64 era sooner than Intel.

For a time that suited the industry and as always favored the users because none could rest on the conquered laurels. Intel began developing the project Larrabee, what's focused in the GPU model, the project was canceled and the i7 architecture emerged. i7 does for Intel the same Athlon did for AMD in the beginning, it lets aside the raw clock race for efficiency.

Again it is the concurrency acting in the background, AMD had bought ATI, the direct implication of it was AMD would compete in that speed race.

Now, after Intel launching successfully it's iX family AMD is lagging behind, but surprisingly it is quiet, AMD roadmap is kept very conservative. To me that roadmap is telling more for what it's not being talked. Something must be being cooked, AMD is already nearing to be 2 whole generations behind. In marketing terms it could mean the death.

That's the reason I think it's not Intel the main concern for AMD now, it's Nvidia and that thought makes me wonder if it's not in something resembling the Larrabee idea is what in the pressure pan. For I don't think the near future points to continue trying to squeeze the CPU architecture that is near the physical, quantum limit. The future points to integrating the controlling and logical capacity from CPU with the raw power provided by the parallel architecture from GPU with heavy two-way physics processing in the between. AMD roadmap already tells something about it, just I think it's not telling everything.

Just comparing clocks don't tell very much about performance. The I5 family is the concurrent for AMD that has nothing to compete with i7 other than it's 'fantastic' price preventing full adoption. I5 is a capped version, that proverbial poor cousin. I3 aims the mobile and very low end market.

In terms of cost/benefit I stick with AMD yet, mainly because I'm constantly upgrading the machine, something painful in the myriad of models, sockets and sheer incompatibilities from Intel. Besides, my goal is not to have the most powerful machine over there, it's having the most balanced machine at the moment.

PS: Athlon owners shouldn't forget it's an outdated architecture now, lacking too many things in relation to Phenom and mainly with the phenom II. The athlon II does not changes the fact the whole architecture is old, it's meant to supply today's low end market. The good news, it's relatively easy to upgrade to Phenom from them.
Imagination is where we are truly real
Reply
#5
This is my opinion., I could be wrong...

The intel wins out in sse performance.
Even when you compare intruction sets that both cpu's have.
The amd's win out in fpu performance.

The amd's have 2x the l1 cache compared to the intel's.
This I believe is why some report better performance with amd's in some apps compared to the i7.

I believe the reason that intel wins out clock per clock is because of the sheer mem bandwith.
Intel has had a better mem controller for a while now, ever since they beat out nvidia's mem controller on intel boards.
Especially when you compare the i7, with a triple chan config.
This may change in a year or 2, when amd "supposedly" put's out a new core with quad chan.
Finally giving the onboard vga card's the ability to use 16x aa and such on certain games, they gotta be close to that now anyways... with mem bandwith around 20k.

Amd does not like fiddling with chipsets.
They're really not that best at that.
And they certainly don't tune the cpu's at all.
This I know for 100% sure, if they would give me the info to tune the sip rom, I could almost double the clock per clock performance.

I know this because I've done so on the nf2, nf3 and nf4 chipsets where the sip rom and bui where stored un-encrypted with mostly static values.

This crap we see on amd's, like my current system for example.
12k, 10k, and 18k in everest for mem bandwith @ 2000mhz.
I know I could make that darn close to even if I had the info I needed.

I did ask them once for help on the newer board's, for the info on the smbus and jtag info's.
But they told me that was to advanced for there normal programming tech support :\, and that I would have to call some more advanced support line thing.

Prob is there office hours, I've tried twice.
I can never seem to figuer out when they are open.
I needed to rma a phenom x4 9950be and I could never get a hold of them.

I would love to mod my cpu's internal timings, so much bandwith to be had.
This lead that intel has would of never happened if they would of just sat down and tuned a value or 2 per multiplier.

It's something that I miss on these newer boards.
Intel still lets you mod there sip rom, though no one has done it.
I've never had an intel though, not since the pentium 1.
I've had a few p4's and p3's, but I really don't care about them enough to mod the bios'es.


Right now as is.
I think the performance is ruffly the same.
The amd's sag a bit clock per clock performance wise.
And mem bandwith...
But the main performance diffrence between intel and amd is there sse performance and instruction sets.

That's what I think anyways.

You can gt an amd cpu that is ruffly around the performance of an intel.
Or you can get that extra 5% and buy an intel, that cost's you $1000 more if you build the system yourself ^^.

I'll never pay more then $300 for a cpu...
I didn't have to pay no $1000 for a pentium1, 486 or 386, and I won't now :\.

ASUS CH4, 1090t @ 4ghz, 2x1gb @ 2ghz, GT240
2x1tb Modded WD Green RAID0, Optiarc 7241S
Modded Enermax 350w

Win7 Ent x86 VLM / Win2K3 Svr Ent SP2-R2 x86 VLM
Reply
#6
The math is fine if you compare the newest AMD chips against the Intel lineup.
A year ago it looked worse for AMD (Phenom 1 class).
Reply
#7
Chart is not accurate at all, it would only work with specific groups like C2D is not much faster than Phenom/Athlon II (not to mention your math sucks Tongue2).
Core i5 3570k -- Geforce GTX 670  --  Windows 7 x64
Reply
#8
You shouldn't take maths as an exact science so much! ( Tongue2 )
Reply
#9
Anyone know the difference between a Phenom II and Athlon II (basically the L3 cache) in pcsx2?
Reply
#10
It's more, the memory management and many more enhancements in the architecture (including the cache hits)... In a way it's the same thing that places the clarkdale so badly in terms of memory latency comparing with the lynnfield... although they perform well is some tasks they do it poorly in many others.

Games benchmarks shows than for many games the i5 750 is the only that consistently surpass the phenom 2 965 in their classes, by a reasonable mark, and that's only because they overclocks madly, we are talking about Both, the 965 and i750 running at 3.8GHz...

So, if you are going for heavy overclock the i5 750 is an option, else stick with the 965BE for sure. Bellow the i750 the options are always bad for Intel in terms of performance/cost.
Above the i750 the performance clearly favors the Intel but the cost becomes very salty.

PS: The above is meant mainly for multithreaded, multitask applications and so on. I really don't know for sure how all the CPUs performs in limited environment under high core affinity, like in actual PCSX2 for example. In this scenery a 2 cores CPU can perform as good as most 4 cores and sometimes be better even. The same applies for some enhancements in the architecture since they can become weaker or null.
Imagination is where we are truly real
Reply




Users browsing this thread: 1 Guest(s)