Is the RSX even based on the 7800?
- Thread starter BenQ
- Start date
groper said:So what about Cell - Xcpu Comp/sec ?
Oh noes! Can of worms!
If you're asking this question then I assume you didn't understand the calculations nor the architecture?
Okay, firstly, the Anand comparison neglects the SSE and 3DNow vector execution units. Taking these into account,
P4 = (2+4)x3.8 GHz ~ 23 GigaComps/sec
A64 = (3+4)x2.8 Ghz ~ 20 GigaComps/sec
Xenon = (2+4)x3coresx3.2 GHz ~ 58 GigaComps/sec
CELL = (2+4)x1PPEx3.2GHz + (4)x7SPEx3.2GHz ~ 109 GigaComps/sec
And before anyone asks,
Xenos ~ (5)x48x0.5Ghz ~ 120 GigaComps/sec
GTX512/RSX ~ as above ~ 128 GigaComps/sec
An important metric missing is branch prediction capability.
Out-of order CPUs (P4, A64)> in-order CPUs (Xenon, CELL) > SM3.0 GPUs (Xenos, GTX512/RSX)
groper said:Thanks Jaws.
Your assumption is correct.
So if Cell is 90% more powerfoul than xcpu in both Gflops & GComp/sec i guess that is safe to assume that is 80-90% more powerfoul overall ?
No, it isn't safe to assume it is 80%-90% more powerful overall, if by that you mean encompassing all scenarios. To know overall we would need to be able to benchmark the Cell and X360 the same way we benchmark any other GPU and CPU. This is all theory. Until someone actually makes code that uses all of the Cells power(because those are peak values of GCOMP and GFLOPS), you can say its that powerful, but we will never get to see it be that powerful if that power is never used. I must remind you that this isn't real world performance and I wouldn't expect either system to get close to their GFLOP or GCOMP benchs.
ihamoitc2005
Veteran
Interesting Metric
This is interesting metric you have provided my friend. One tech demo of P4 and CELL comparison showed 5x improvement by CELL and this is not very different from number you have provided and very different from simple flops comparison especially since depending on use sometimes scalar is meaningless.
However another experiment of 7800GTX with CELL had different result than measurable by this method or by flops so question of processor rating seems very complicated. Thank you for your provided numbers as it is very instructive for my understanding.
Jaws said:
This is interesting metric you have provided my friend. One tech demo of P4 and CELL comparison showed 5x improvement by CELL and this is not very different from number you have provided and very different from simple flops comparison especially since depending on use sometimes scalar is meaningless.
However another experiment of 7800GTX with CELL had different result than measurable by this method or by flops so question of processor rating seems very complicated. Thank you for your provided numbers as it is very instructive for my understanding.
overclocked
Veteran
Xcpu has what about some 75 Gflops peak if i remember reading MS docs right. The 115 Gpflops number apparently if i understood it right came from that they include shuffling as instructions in the PR numbers.
From Ibm docs they dont count shuffling on Cell so i think the ones spreading most PR shit is MS.
Whitout saying much more i think Cell will be the best cpu and if many could get their bias away i think most would agree with that.
From Ibm docs they dont count shuffling on Cell so i think the ones spreading most PR shit is MS.
Whitout saying much more i think Cell will be the best cpu and if many could get their bias away i think most would agree with that.
Overclock you have a good point. I too believe the Cell is a very powerful CPU. I also believe the RSX will be a powerful GPU. I don't question that at all. My skepticism lies in , as xbdestroya said, if it will be tapped and if it the SPEs will be limited in their use. Yes, some of this is just speculation, but I think we will truly see if it can or can't be done when the PS3 comes out in all its glory next year. Actually I think we will know before then, well at least my patience hopes it is before then, because as time passes we will get to see more and more of the PS3 and its games. I am looking forward to the gaming conferences that will reveal the true power of the Cell. Once I see some playable demos my skepticism will go away.
fulcizombie said:
Not sure why people don't take these games as being early builds into consideration when making comments. Not trying to rag on anyone but at least realize what we have seen from 90% of the PS3 has been on very early kits. All 360 stuff is running on nearly complete dev kits. I think I-8 looks great, GoW great? No of course not, but certainly has the potential to get to that level with ease IMO.
And as a side note, Too Human? Wasn't that a PSX game? No a Cube game? Seriously, I won't even acknowledge it until I see and actual review of it. Look up vaporware in the dictionary and you'll see a picture of Too Human.