Recent content by EduardoS

Or testing artifact? CPU workloads usually doesn't need to much bandwidth so memory controller is organized (and optimized) to minimize latency, not to maximize bandwidth, the testing software will check CPUID and branch to path where code is written and memory organized in a way supposed to...

No, someone else already had the Fusion trademark registered.

I'm not sure, but I expect memory latency to be near 100ns, at 1GHz about 100 base clocks, not 256. And then, there are the caches and prefetching...

If I intentionally execute a computer virus it doesn't becomes another type of software, if it does something I want or not is not a definition for virus. The only definition for computer virus I have ever seen is of a software being able to self replicate, the "power virus" is a exception to...

Definitions... In this case a very informal one, I'm not a marketting guy, yet, everytime I hear or talked abou power virus the only requirement is making the chip draw a really lot of power, maybe the name is silly, but if everyone understand all is fine.

You speak (I mean, write) as if what slow downs Thaiti in games is too expensive to fix¹ and doesn't affect compute performance². 1) Maybe, if it was easy I assume it would already be fixed, but I still believe this wasn't intentional. 2) I'm not sure if this is true there still so few GPU...

Sure you are right, but why comparing a hypothetical GPU to a real GPU? I'm not sure GTX680 is the perfect balance. BTW, I don't think HD7870 is the perfect balance either, the problem Thaiti have Pitcairn have as well just more limited, and the most interesting topic about GCN to me still...

Because it is already a "gamble"? Performance may vary with OS version, 3rd party software, montly drivers releases, etc. For the market it is targeted this variance is acceptable.

Good question, they are closeser to each other than ever, apparently GCN have some sharp edges to be polited, so let's wait. It is usefull, you don't waste signifcant resource on redundant operations, but I can't say if it is used, by looking at previous drivers...

When using some software, it tells nothing about memory bandwith to gaphics because: 1) The poor result from those benchmarks are likely becuase the specific software fails to understand the interleaved memory rows only reading from one channel each time, a more random pattern is more likely...

What if, I choose a random factor for each chip being a multiple of 3 for Cape Verde, a multiple of 6 for Pitcairn and a multiple of 8 for Tahiti? I know this isn't very scientific but... There should be something more than just ROPs, TMUS, SPs and memory bandwidth there...

Actually, since I included the HD7*50 models the CUs/TMUs ratio increase by 25%, with no meaningful increase in performance. Since Tahiti have 3-way associative ROPs and the other chips doesn't (direct mapped?) this may increase Taihiti ROP's efficiency. Initially I read this as 4 partitions...

Maybe I missed thumbnail?

I put my lazyness to sleep, open techpowerup and Excel and plotted an interesting chart about % 7870 performance at 2560 per pixel fill rate: Maybe ROPs are a bottleneck, but then why Taithi performs so well? Are the ROPs and memory controllers tied in other parts? What associativity for...

The workload may lacks parallelism too, then the higher clock of HD7870 helps.