- Mar 3, 2017
- 1,747
- 6,598
- 136
Eh.I feel like there is huge disparity when AMD fails and don’t reach expectations compared to when Intel, Nvidia and Apple fail. Not directing this at you but it’s something I noticed here and other online forums/twitter.
Should it? That 9.71% you cite is int IPC, and nothing in Zen 5 desktop is going to increase int IPC. If anything, it could end up lower since AMD is reusing the IOD on desktop, and Strix may contain uncore improvements.
Floating point uplift will be higher on desktop when AVX 512 is in use, of course.
David Huang was not using an ES. It was a retail sample.
I don’t get how having a worse generation uplift than Zen 4 is a good result when you take everything into account.
Should it? That 9.71% you cite is int IPC, and nothing in Zen 5 desktop is going to increase int IPC. If anything, it could end up lower since AMD is reusing the IOD on desktop, and Strix may contain uncore improvements.
Floating point uplift will be higher on desktop when AVX 512 is in use, of course.
Disingenuous. Over like 15 quarters.Apple got for that 7% IPC improvement over 7 months
I don't think Zen 5's perf/watt uplifts are that good, based on what we have seen on desktop.I thought it was an ES, my mistake if not.
A ~10% IPC improvement with a greater improvement in efficiency and battery life on a very minor node tweak is pretty good. Not great and a definite step back from the previous few generations of improvements, but it's still pretty good.
Leaked GB scores for desktop Zen 5 indicate higher IPC, but we'll see once we get better tests.
RDNA3 was mocked relentlessly. However, people see AMD as the underdog, still remember Bulldozer, and will naturally be less harsh with them. Intel are also seen in a bad light in general because of business practices and lies.I feel like there is huge disparity when AMD fails and don’t reach expectations compared to when Intel, Nvidia and Apple fail. Not directing this at you but it’s something I noticed here and other online forums/twitter.
Apple is the biggest company in the world with over 3 trillion in market cap. AMD has a market cap of 220 billion. Apple's worst quarter's earnings is about the same as AMD's yearly revenue. They are not seen as the same thing, don't have the same resources, and don't have the same expectations.The amount of crap Apple got for that 7% IPC improvement over 7 months mind you not even 22 months like AMD is certainly constitutes favouritism. Comments like Apple’s CPU team dead and ARM is dead but when AMD doesn’t reach expectations then its “design is hard”, they focused resources somewhere else, Zen 6 is the one to look out for etc.
10% increase in IPC is 5% YoY, so is AMDs CPU team dead no it’s not, so maybe Kepler and the “x86 gang” should not make statements like how Intel and AMD increase IPC 10% YoY which is certainly not true.
Here is what David Huang got for Spec 1t and IPC with an ES sample when equalizing clocks.
View attachment 104081
So a 9.71% IPC increase and the GCC test actually improved by 10.4%. I estimated 7% from Anandtech’s data with a rough calculation of actual clock speed difference but actual clock equalized tests are obviously better. Still a bit underwhelming, but combined with the efficiency increase, it’s an overall good result, just a let down after all the hype. The desktop variant should have higher IPC, it’s just a question of how much.
From a customer perspective, why would anyone buy Kraken Point over Hawk Point?Does anyone have rumors or estimate of the die size of Kraken Point?
It is even more cost efficient for AMD to make than Hawk Point?
Same node, double digits performance increase vs 2 nodes jump, 2 new cores design single digits performance increase.Precisely. I guarantee this forum mocks Lion Cove relentlessly when it launches with similar performance increases.
According to David Huang it's not even double digits performance increase in Specint.Same node, double digits performance increase vs 2 nodes jump, 2 new cores design single digits performance increase.
They are not the same.
I was talking about Gen on Gen improvements.Disingenuous. Over like 15 quarters.
Firestorm was the last bump before that (8%).
At least mention ARM, they still maintain their PPC CAGR, even if power went boom-boom since X4.
I was talking about Gen on Gen improvements.
M3 -> M4 was 7 months regardless of how you spin it. Both have different architectures.
Node wise, Intel has the bigger jump, but core wise, Zen 5 seems like a much more radical departure than LNC is over GLC. And tbh, some parts of LNC don't seem to be buffed as much as a usual Intel tock is, like the ROB capacity or the uop cache capacity. I don't think the same can be said for Zen 5.Same node, double digits performance increase vs 2 nodes jump, 2 new cores design single digits performance increase.
They are not the same.
Coll-P had rounding error IPC bumps (and some regressions).M3 -> M4 was 7 months regardless of how you spin it.
No such thing, you do CAGR.I was talking about Gen on Gen improvements
They could have just sold the mislabeled 9700x with $25 discount.Edit : On a side note :
Was Ryzen 9000 delayed over a simple typo? AMD Ryzen 9 9700X spotted - VideoCardz.com
Ryzen 9000 delay, it was a typo all along? Ryzen 9 9700X instead of Ryzen 7 9700X. AMD has not directly confirmed the reason for the Ryzen 9000 launch delay. Officially, the company cited quality issues that did not meet their expectations. However, it seems the issue might have been a simple...videocardz.com
I would wait for Geekerwan testing.I agree that IPC not going up over Zen4 in Specint at all, is dubious.
It could been a collectors item tooThey could have just sold the mislabeled 9700x with $25 discount.
Yeah, misprints are precious stuff, heh.It could been a collectors item too
Will probably make it to Ebay, at some point, I'd bet.It could been a collectors item too
Are you forgetting that LNC is a two node jump from GLC while Zen 5 is a half-node (at best) jump from Zen 4? Thats pretty significant. And where do you get the idea that Intel didnt "try" for a massive core arch refresh? Isnt that what both Lion Cove and Skymont are?I would consider Zen 5 a bigger "meh" than LNC, but as a whole, the entire "which core is more disappointing vs expectations" shtick isn't really important in the end lol. It's not like Intel tried for a massive core arch refresh and got mid results, they literally didn't try at all.
Agreed. Zen6 may bring more than originally intended to the table, just from bugfixing Zen5. Let's not get our hopes up too much though, shall we?I hope they iron out whatever kinks it now has come second gen FWIW. Would be a shame to waste an unconventional design.