They used different tech processes with different IPs, as for PPA N7+ has +10% speed and x1.2 density.AMD managed to keep roughly the same clocks at the same voltage while delivering a ~20% IPC uplift (on the same node), which was pretty impressive, with zen 3 vs zen 2.
what's IPC/Cac? 1:1 IPC/Cac meaning for every 1% improvement in IPC, you get a 1% increase in dynamic power?They used different tech processes with different IPs, besides, it has little to do with IPC / Cac ratio.
Zen3 is vanilla N7p.as for PPA N7+ has +10% speed and x1.2 density.
It's not IPC/Cac ratio, but the ΔIPC/ΔCac ratio. Isn't it a bit different thing?AMD lists the IPC/Cac ratio of Zen 3 to be >1
that's the wikichip statementZen3 is vanilla N7p.
On Investor's Day in May 2017 Jim Anderson, AMD Senior Vice President, confirmed that Zen 3 is set to utilize 7nm+ process.
Isn't that we were talking about in reference to Zen 3 vs Zen 2 in context? Someone brought up that Zen 5 is likely to have lower 'architectural efficiency iso frequency, someone else mentions that AMD tends to deliver 1:1 IPC/Cdyn, which makes it sound like for every new 1% increase in IPC is met by a 1% increase in Cdyn, which based on AMD's Zen 3 slides, appear to be true. It's possible Zen 5 would follow that path and do the same thing.It's not IPC/Cac ratio, but the ΔIPC/ΔCac ratio. Isn't it a bit different thing?
Isn't that we were talking about in reference to Zen 3 vs Zen 2 in context? Someone brought up that Zen 5 is likely to have lower 'architectural efficiency iso frequency, someone else mentions that AMD tends to deliver 1:1 IPC/Cdyn, which makes it sound like for every new 1% increase in IPC is met by a 1% increase in Cdyn, which based on AMD's Zen 3 slides, appear to be true. It's possible Zen 5 would follow that path and do the same thing.
True, although Anandtech got a 19% improvement in SPEC in their Zen 3 review, didn’t they?Actually if we remove games from the 19% IPC estimation and keep only a set of significant apps then Cdyn, i.e core size, increase by 1.36%/1% IPC, that s still a remarkable achievement since the scaling is more generaly a root square function, that is 2% bigger size for each 1% better IPC.
True, although Anandtech got a 19% improvement in SPEC in their Zen 3 review, didn’t they?
One thing to keep in mind about Zen 3 is that the TAGE branch predictor from Zen 2 was originally planned to appear in Zen 3. So, according to the original plan, the IPC increase from Zen 2 to Zen 3 was supposed to be over 19%, but ended up being smaller because of that. I expect to see an official figure of at least 20% when AMD unveils Zen 5.
Price of the CPU is the least problem for Zen5 DIY CPUs, IMHO. AMD will pick a price that delivers good margins and go from there. If the market is still weak, prices will go down, if not, they will stay up for a while. People always go on about price. Clearly, in the current market, and with the complexity of leading edge semiconductor design/manufacture these days, initial prices will keep going up for the top chips.if amd can launch zen 5 at current zen 4 prices and close the gap of the reg vs x3d down from $100 to $70-80 I think they'll have a wnner on their hands. they won't get that initial big wow in revenue but not like that worked out for them with zen 4, but they'll make it up in volume. at those prices they'll be flying off the shelves. by next year ddr5 will be cheaper than it is now. and it's cheap now.
That's what amd tried to accomplish this last fall but sales were stagnant. Am4 still delivered phenomenal performance for the price. ddr5 was expensive. for many it is still expensive today even if you can get a decent 32gb kit for under $120. am4 is still very good. For many there is no reason to upgrade to zen 4, probably won't be for zen 5. for amd the optics of launching as you say and then reducing prices because units aren't moving isn't going to be a good look.Price of the CPU is the least problem for Zen5 DIY CPUs, IMHO. AMD will pick a price that delivers good margins and go from there. If the market is still weak, prices will go down, if not, they will stay up for a while. People always go on about price. Clearly, in the current market, and with the complexity of leading edge semiconductor design/manufacture these days, initial prices will keep going up for the top chips.
I believe Granite Ridge is destined to come out in q4 of 2024 not q2 like old leaks suggested. It did seem very likely it would be q2. I'm not too concerned about projection based on preliminary data more than a year out. in either case the july leaks suggest a 6-21% average gain over the 13900k. Zen 5 is allegedly around that same figure over zen 4. It'll come down to what performs better in edge cases and power usage. That figure range is worst to best based on current sampled data the leak is from. I'm not at all concerned about hardware that's a rough sample that boots.I wouldn’t expect very attractive prices from Granite Ridge, initially. It should only be competing with Raptor Lake refresh at first, and even Arrow Lake, which should launch quite a bit later, seems to be somewhat underwhelming according to Intel’s own leaked projections. So we’ll probably have to pay up for premium performance. No dense cores to pad the MT performance like Strix Point in laptops, either. Betting there will be outrage about the launch pricing of the 6 core model.
I guess we'll see. I'd imagine that AMD is making sure Zen3 will be largely out of the market in time for Zen5. They left the option open given that they knew platform prices for Zen4 were a bit insane at release. If I stayed on Zen2, Zen5 would likely have looked very good - but Zen3 is good enough for me and probably will be for several years to come. If someone has a 5800X3D - they are probably better off saving for a better video card if they are gamers. Guess that's a problem for AMD. Why put more money into your CPU (which requires a platform upgrade), when video card prices are stratospheric (and will be staying that way)? Too bad RDNA3 didn't pan out better, they could be pulling in better revenues from graphics cards (although the whole PC market is in a slump right now).That's what amd tried to accomplish this last fall but sales were stagnant. Am4 still delivered phenomenal performance for the price. ddr5 was expensive. for many it is still expensive today even if you can get a decent 32gb kit for under $120. am4 is still very good. For many there is no reason to upgrade to zen 4, probably won't be for zen 5. for amd the optics of launching as you say and then reducing prices because units aren't moving isn't going to be a good look.
Amazon's best sellers for the top 50 processors has a scant showing of zen 4, mostly zen 3 and more Intel than Zen 5.
5600x and 5800x make up 1st and 2nd, 3rd and 4th are the 13900k and 13700k respectively.
Even if zen 5 brings 20% ipc over zen 4 it's still going to be a hard sell if amd charges $800 again for their 16 core, again.
if for some magical reason intel can compete on performance and thermals and cost less, they'll take the crown. it's hardly like newegg's top sellers page is zen 4, there's only 3 zen 4 processors, the rest is am4s zen 3 and intel 12th and 13th gen.
I wouldn't trust estimates of dates any more than I would rumors of performance at this point.I believe Granite Ridge is destined to come out in q4 of 2024 not q2 like old leaks suggested. It did seem very likely it would be q2. I'm not too concerned about projection based on preliminary data more than a year out
Too bad RDNA3 didn't pan out better, they could be pulling in better revenues from graphics cards (although the whole PC market is in a slump right now).
RDNA3 perform well in tests but badly according to the usual urban legends rehashed ad nauseam, what they lack is cards that cover the mid range segments, at this point they only have the 7900XT/XTX at the upper range and the 7600 at the bottom, money is made between thoses two segment for a big part.
Grafikkarten-Rangliste 2024: GPUs im Vergleich
Welche Grafikkarte kaufen? Die GPU-Rangliste 2024 liefert Empfehlungen auf Basis von Tests mit GeForce RTX 4000, Radeon RX 7000 & Arc.www.computerbase.de
So what did badly panout according to these numbers..?
bit like telling a prego woman when she's due.I wouldn't trust estimates of dates any more than I would rumors of performance at this point.
That at least has a hard estimate based on embryo size and established growth rate with health of mother or foetus being the main factor in the estimate being off.bit like telling a prego woman when she's due.
something was hard 9 months prior yes.That at least has a hard estimate based on embryo size and established growth rate with health of mother or foetus being the main factor in the estimate being off.
Semiconductor product annoncements are much more nebulous with so many dependencies involved including previous gen channel stock, market conditions, current company finances and who knows what else before you even consider the hardware and software readiness problem.
😱 I feel like this emoji doesn't convey the ewww from reading that 😅something was hard 9 months prior yes.
It will depend on supply.Edit: Will say that I am wondering how much interest from OEMs AMD expects from Strix Halo.