I doubt he typed his alias there (if that even is his alias). 32% is too low. Could be one of the lowest gains of Zen 5 in a benchmark though (when it is completely crippled in terms of memory latency and bandwidth).if he confirms it
It's information overload lately.
But even if you average the 5-10% IPC and the 50-100% IPC, it'll arrive at ~50% IPC improvement. Then add the MT perf increase higher than ST perf increase claim on top. I'll be willing to pay the $999 that has been promised by some in this thread for the top 16C SKU in that case.
I'm not making any predictions. I'm an observer. But I'm missing predictions from several of the main leakers / self claimed insiders in this thread, who has been most vocal about their claims, in that spreadsheet. Would be nice if they could add their concrete numbers to the spreadsheet, for clarity and accountability.
Spreadsheet says "IPC % over Zen 4" and for row 7 "32%". Should it not be interpreted as average IPC increase? Otherwise I think there should be a comment column specifying a specific benchmark, if it only applies to that.I doubt he typed his alias there (if that even is his alias). 32% is too low. Could be one of the lowest gains of Zen 5 in a benchmark though (when it is completely crippled in terms of memory latency and bandwidth).
What about the 5-10% IPC increase prediction then, is that also not daft? Then perhaps averaging the two daft predictions of 5-10% and 50-100% ends up with something truly genius and precise. 🤣if you are talking about my daft comments I am just being silly with Igor. You don't have to be serious all the time.
I could add some.Are you taking the people responding to Igor "Voice in my head" Kavinski seriously? Or Hans "RDNA3 uses inferior silicon" Gruber?
Then you must also take the people saying clock regression with 5-10% IPC seriously.
The aggregate spreadsheet, to which I asked you to add your predictions numerous times but remains absent, still has an average of 25% (a bit too high in my opinion)
"I.P.C." is always an average. It's just that the sample from which this average is obtained can consist of a single executable and input data, or multiple of those.Spreadsheet says "IPC % over Zen 4" [...]. Should it not be interpreted as average IPC increase? Otherwise I think there should be a comment column specifying a specific benchmark, if it only applies to that.
Can I get a nickname?And
I could add some.
Same as one cited. Just more non sense.Can I get a nickname?
Henceforth I shall be known as Mah "just more nonsense" BoiSame as one cited. Just more non sense.
At this point it's well established that AMD themselves always use SPECInt Rate 2017 as a measuring stick for IPC, so whenever we say "IPC" in our predictions we mean "clock-normalised SPECInt Rate 2017 score"."I.P.C." is always an average. It's just that the sample from which this average is obtained can consist of a single executable and input data, or multiple of those.
Regardless, I agree that without specifying this sample, these "I.P.C." figures are completely meaningless. Some don't specify it because they have no idea what they are talking about, others don't because they prefer to remain vague.
ARM do as well. It’s an industry standard.At this point it's well established that AMD themselves always use SPECInt Rate 2017 as a measuring stick for IPC, so whenever we say "IPC" in our predictions we mean "clock-normalised SPECInt Rate 2017 score".
By "we", you don't include the entire AnandTech forum membership (at least the part which involved itself in iso-clock core performance related discussions), do you?At this point it's well established that AMD themselves always use SPECInt Rate 2017 as a measuring stick for IPC, so whenever we say "IPC" in our predictions we mean "clock-normalised SPECInt Rate 2017 score".
O.K., I did indeed forget to add: …and others don't because they imply IPC = "SPECrate® 2017 Integer iso-clock performance".without specifying this sample, these "I.P.C." figures are completely meaningless. Some don't specify it because they have no idea what they are talking about, others don't because they prefer to remain vague.
I'm fairly sure he means the AMD Clan Warriors of AT.By "we", you don't include the entire AnandTech forum membership (at least the part which involved itself in iso-clock core performance related discussions), do you?
By "we", you don't include the entire AnandTech forum membership (at least the part which involved itself in iso-clock core performance related discussions), do you?
Those frequenting this thread.By "we", you don't include the entire AnandTech forum membership (at least the part which involved itself in iso-clock core performance related discussions), do you?
oh snap, we have a productive member of society among us. Rise up, anandtech!I am working on a benchmark similar to Geekbench (in that it uses similar workloads), but I have no timeline for completion. I'm also thinking about making it open source and cross platform, but we will see. I'm undecided as of yet. I will probably release initial versions for free and decide after. (note I'm NOT trying to pump this project with these posts, my project may ppossibly never see a release due to my ridiculous schedule and ADHD, I'm simply pointing out the need for such a benchmarking tool)
Well, sorry, over all the weird Geekbench and CPU-Z and whatnot discussions here (which I mentally tried to skip but failed to) I did indeed forget that there are people who follow the one convention which you pointed out.Look, you can keep being difficult about it, that's not going to change that this topic was discussed multiple times at this point and the outcome of said discussions is always the same.
We'll see if it makes it to release. 🤣 2 of my kids have special needs, so they consume almost all my free time. There is hope, however.oh snap, we have a productive member of society among us. Rise up, anandtech!
Being an IT forum and having ADHD I'm gonna guess Audhd, I have two of them myself, handfull they are.We'll see if it makes it to release. 🤣 2 of my kids have special needs, so they consume almost all my free time. There is hope, however.
32% less than zen 5What is the Specint number for the 7950X?the total avg