- Mar 3, 2017
- 1,747
- 6,598
- 136
Why it is AMD's fault when the performance doesn't meet some random forum user's expectation/fantacy ?not really , its called analysis of data points.
like it really is either AMD completely messed this core up ( i would argue more then bulldozer ( and i would be right ) ) or the extrapolations people are doing are largely correct.
it really comes down to , do your believe the single phrase single data point on a slide , or the foot note that contradicts that slide plus all the other data points that contradict that slide.
Why it is AMD's fault when the performance doesn't meet some random forum user's expectation/fantacy ?
I was actually thinking about Liam Neeson after yesterday...Some people just never learn. They can't be taught.
Absolutely, that 16% number, power draw change on Granite Ridge and several little things don't make much sense. My headcannon is that AMD did their usual penny pinching thing and instead of coming out with a large core that would smash the competition, they went for something cheaper that aimed at great power efficiency. That puts everything Kepler/Adroc put as dead wrong, but I don't think they would just randomly throw >40% improvement for kicks. So for me the guesswork is where will we land between 16% and 30% with GNR.What matters is the actual result. And there seems more to the 16% figure.
Because if you completely ignore the leakers, Zen 5 is still a breaking point in Zen's history.Why it is AMD's fault when the performance doesn't meet some random forum user's expectation/fantacy ?
The slides AMD shared had a lot of MT workloads in the 16% claim but the gaming slides are already indicating a larger gain than that IPC figure. They put Zen5 a tiny bit ahead of the Zen4 X3D which indicates a ~25% performance uplift in gaming.
Immediately something does not mesh so with that in mind I am waiting for Anandtech to do the SPECint 2017 1T tests to work out the IPC uplift in that benchmark. If it is less than 32% then adroc/Kepler were unambiguously wrong, if it matches or exceeds they were correct.
Given the gaming uplifts it suggests the 1T IPC uplift is greater than nT which is something that adroc has said.
Also, AMD sandbagged hard with Zen4. Pre reviews it was looking like it would be a lot slower than Z poen3X3D in gaming but it was actually on par and is now frequently ahead by a small margin.
So a long way of saying wait for benchmarks if you want definitive answers. In the interim just have fun, there is no need to lash out at people, it is just a tech thread, it is not that serious.
AMD did make the Ryzen AI model number last minute change too so they've been adapting, hopefully for better success. I think they've worked too hard for Zen 5 to get lambasted in reviews for something like 95C max temp, when Intel TJmax is 100C and their laptop chips almost always throttle, especially the H and HX ones. Guess this time they want to project an image of cooler running chips.Except that manifest was 40 days ago and the announced power was like 5 days ago.
Except this 12-core prototype never made it to market
Except this 12-core prototype never made it to market
Oh we are patiently waiting for iso-clock comparisons.RE: All the silly "muh IPC" posts, I found this gem, this is how you are supposed to compare CPU's to get the real performance improvement, fixed clocks.
Originally from AMD Computex 2021 presentation:
Computex 2021: AMD's Keynote, a Live Blog (10pm ET)
www.anandtech.com
Screen grab from here:
View attachment 100787
We want all the iso's!!!Oh we are patiently waiting for iso-clock comparisons.
Most of the tasks most users do are still lightly threaded and not bandwidth limited.Yeah. It’s just fanobyism. Same thing happened with Intel, people for years said 4 cores is all you need. Then Zen came out and suddenly 8, 12, 16 cores became desired.
After 16 and Zen 2, the only higher core parts are thousands of dollars, nevermind more expensive mobos.
Same thing for higher bandwidth. We’ve been stuck on dual channel forever.
AMD, it turns out, likes to make money just as much as Intel and Nvidia.
Volume and cost, PPA is a nice metric to win.what is the benefit of shrinking the cores/CCD altogether?
why dont they fill a big die area as much as possible? (i guess they will do that with strix halo)
This particular ES has been doing rounds lately. I'm interested in how high it clocks.most useless benchmark but here is another datapoint.
AMD Ryzen AI 9 HX 370HX "Strix Point" engineering sample is 25% faster than 54W 7940HS in CPU-Z test - VideoCardz.com
Ryzen AI 370 HX ES tested AMD Strix Point in leaked benchmarks. At Computex, a attendee got hands-on experience with ASUS’ upcoming A14 TUF Gaming laptop, showcased at the ASUS booth. This laptop, powered by the Ryzen AI 9 HX 370, was confirmed through the datasheet as displayed next to the...videocardz.com
It’s from an Asus laptop from computexThis particular ES has been doing rounds lately. I'm interested in how high it clocks.
I think it's VS 7940H, and if the clocks are 5.2Ghz for Zen 4 vs 5.1Ghz for Zen 5, the Cpuz 1T is around 15% better per clock. Not bad, given the fact that Zen 4 had it in the IPC mix from AMD with only 1% (vs Zen 3) and still managed to pull 11% in SpecINT.most useless benchmark but here is another datapoint.
AMD Ryzen AI 9 HX 370HX "Strix Point" engineering sample is 25% faster than 54W 7940HS in CPU-Z test - VideoCardz.com
Ryzen AI 370 HX ES tested AMD Strix Point in leaked benchmarks. At Computex, a attendee got hands-on experience with ASUS’ upcoming A14 TUF Gaming laptop, showcased at the ASUS booth. This laptop, powered by the Ryzen AI 9 HX 370, was confirmed through the datasheet as displayed next to the...videocardz.com
13% faster than 7945hs in cpu-Z
You mean the ProArt one? I think that particular laptop had its chip detected as HX 170. Or was there another one?It’s from an Asus laptop from computex
So it boosts up to 4.4 under ST workload if I see it correctly? That's a pretty good score for 4.4 GHzI downloaded the source video from bilibili (because the quality in the player is horrible for me) and reuploaded it to youtube:
I don't know how reliable task manager clocks are, but multi seems to run at 3.7ghz?
View attachment 100794
EDIT: Also the GB6 score at the start of the video is this one: https://browser.geekbench.com/v6/cpu/6420909
matches the eng sample name as well so that geekbench score should be legit.