- Mar 3, 2017
- 1,626
- 5,910
- 136
RDNA3 would be best served with a 30% price cut to move units if it will come to that. With Intel improving their dgpu drivers ever so regularly and being committed to the next gen hardware I don't see AMD being the budget or mid choice anymore and they cannot compete at the high end with the 4090. AMD's lackadaisical approach to getting RDNA3 out and improving the experience with their software stack competing with Nvidia has not been a great show to consumers such as myself who were considering getting an RDNA3 card. I don't want to run the risk of a house fire getting an Nvidia card now. I'll wait for RDNA3.5 next year.N3E vs N5 provides 32% reduction in power draw at ISO performance.
This is pretty good If you think about It.
If we say that 7800XT had 225W just for the GPU alone, then 2/3 of units would be 150W, further reducing It by N3e process would result in 102W at 2430MHz.
Then there is still the possibility of higher clocks by fixing what went wrong in RDNA3, It really could end up faster than N22 while consuming a lot less.
No it's just doing more effort for negligible packaging area gains.
PHY swaps aren't quite that.
No you stick a fat PHY in and let it run lowest link states outside of heavy loads.
OEMs are cautious with AMD only because if they can't supply enough stock to them.Even if it helped only marginally to the OEMs, in order to increase adoption,
It won't increase the adoption since the upfront part cost is the main barrier to entry.Even if it helped only marginally to the OEMs, in order to increase adoption, it could end up worth it
They're doing more than well for that.Like I said, this is an area where AMD is lacking...
Masks aren't the expensive part and the design effort is minimal.But that would need a full mask set, wouldn't it?
Unrelated to this but do you think AMD would follow Intel with on package LPDDR5X or is the cost risk too high?It won't increase the adoption since the upfront part cost is the main barrier to entry.
Not in the near future, that's for 10W parts AMD isn't building.AMD would follow Intel with on package LPDDR5X or is the cost risk too high?
Ok. What's your opinion on Intel trialing it out? I know this is the zen thread but would this maybe be viable with low 10 watt packages or even higher power packages in future to compete with Apple without going the dismal apple route? Would you think there would be a breakthrough in packaging that allowed the LPDDR5X to be replaced if it failed instead of being soldered?Not in the near future, that's for 10W parts AMD isn't building.
OEMs are cautious with AMD only because if they can't supply enough stock to them.
and what is a stacy.
That collapse was due to consumer demand falling flat on its face. Anyone who needed a new computer got one between early 2020 and early 2022. It wasn't just amd who suffered here.Just as those concerns evaporated (H2 2022), when all the constrained disappeared, AMD client shipments collapsed. So that "cautiousness", on part of OEMs does not apply.
Given what adrock thuston has said Zen 5 seems like a real winner here. What I am more curious about it how fast processor releases will be now given that the pandemic is behind us and AMD can be on time but the whole cowos debacle is a new issue that needs to be nipped n the bud.Let's see how things develop on the client side, now that the inventory correction also seems to be behind us. If AMD market share does not shoot up, then we can conclude it is something other than AMD's ability to supply, and if the product is equal or better than Intel, but market share does not reflect it, we will also know it is something other than that.
I'm going to assume this is one of those old school names some men have.That's a he.
Yeah selling stuff at substandard margins bad.AMD client shipments collapsed
No it's just doing more effort for negligible packaging area gains.
PHY swaps aren't quite that.
No you stick a fat PHY in and let it run lowest link states outside of heavy loads.
Given what adrock thuston has said Zen 5 seems like a real winner here. What I am more curious about it how fast processor releases will be now given that the pandemic is behind us and AMD can be on time but the whole cowos debacle is a new issue that needs to be nipped n the bud.
That really depends if USR on the SoC side gets more than 1 SDP.So would be possible, if given enough power, that Strix Halo could outperform the equivalent desktop part.
Depends on the SDP count.and effectively 8 half channels or 4 full channels of memory bandwidth could be supplied to a single CCD.
It has SLC for a good reason.Memory latency is typically a bigger problem than bandwidth, and we will see how it works out with Strix Halo
It still needs a gigantic CoWoS-L carrier to integrate all the HBM.By Mi400, AMD will fully transition to the silicon bridges.
Oh right you may be onto something there. I keep attributing covid to the late z4 launch but forgot that it would have affected rdna too. that may be why someone in another thread was comparing rdna5 to black well and why someone asked and then my inept reply because I'd overlooked that issue.It seems like RDNA 4 and Zen 5 will shorten the release cadence.
As far as CoWoS, it is not affecting AMD in the client space or server CPU. Only in datacenter GPU and some Xilinx products.
Intel has a similar 2.5D approach in its Meteor Lake and future server products. I am not sure how the industrywide bottleneck affects Intel specific implementation.
Going forward, AMD is bypassing CoWoS like 2.5D. Apparently with a combination of silicon bridges, that may come to RDNA5 and Venice and Fan Out implementations in client.
So it is really only Mi300 that is in the same boat. By Mi400, AMD will fully transition to the silicon bridges.
When the AI boom is over, there may be companies out there holding a CoWoS bag, with too much capacity and not enough demand.
That looks very nice tbh. was convinced it was plastic until I read the review's initial first few paragraphs. I do not like the offset track pad and the arrow keys jutting out.I would love Strix Halo in Lenovo Legion Slim 7(i) Gen 8. That laptop looks great. review
I actually like that arrow keys are jutting out, not sure If It's usable or not.That looks very nice tbh. was convinced it was plastic until I read the review's initial first few paragraphs. I do not like the offset track pad and the arrow keys jutting out.
there isn't enough palm space on the left.I actually like that arrow keys are jutting out, not sure If It's usable or not.
I don't understand the meaning of offset track pad.
What I don't like is that small tilde key and the combined keys.
Oh well, you can't have It all.
N3E vs N5 provides 32% reduction in power draw at ISO performance.
Here you go.How does N3E compare to N4P? Or, if that direct comparison is not available, how does N5 stack up versus N4P? Then we can compare.
Thanks. I was too lazy to look that up myself. Napkin math when considering N3E 3-2 (it's the most-performant of the N3E variants) versus N4P:Here you go.
BTW, why do you need N4P exactly?
Not sure how you can get improved performance at isopower but higher power consumption at isoperformance . . .? Too bad we'll probably never know if specific N3E-based products will use N3E 3-2 or 2-2.
Potentially.Can't they mix up all of these in a single design now?
I also remember they can with N3.Can't they mix up all of these in a single design now?