- Mar 3, 2017
- 1,684
- 6,227
- 136
why does IBM run SMT 8 , there are always reasons , as soon as your data flow/workload needs to go distance SMT is all win. An example would be Database's.This all makes sense but makes me wonder.. why keep SMT then? If there’s a meager 15% uplift from SMT, is it really worth the trade offs at that point? Getting rid of it reduces a lot of security and validation hurdles.
But...but... Redgaming tech said SMT-2 for Zen5 with high confidence!Expect that both, Intel and AMD will ditch the SMT from their mainstream CPUs.
it does
Then AMD would say "different TSMC 7nm finfet technology as Zen 2"
Better arch
Intel already has with Arrow Lake.Expect that both, Intel and AMD will ditch the SMT from their mainstream CPUs.
Intel might but no way amd will, might be 4+ yearsExpect that both, Intel and AMD will ditch the SMT from their mainstream CPUs.
Without smt intel projects a <10% ipc uplift. Is it worth it though?Intel already has with Arrow Lake.
I suspect one of the reasons Zen 5 kept SMT is because Turin customers wouldn’t be happy with half the threads.
That’s not IPC, that’s 1T performance. We don’t know what LNC will bring in IPC gains. Considering it’s probably going to lose ~800mhz of frequency and appears to be on a less than optimal node it’s got to have a decent IPC bump just to hit 5% 1T performance increase.Without smt intel projects a <10% ipc uplift. Is it worth it though?
You were wrong.
And for Zen 4 they mention that they use a customised process, we can eventually expect the same for Zen 5 :
AMD: We’re Using an Optimized TSMC 5nm Process
www.anandtech.com
The process was tweaked or optimized between the initial Zen 2 release and Zen 3, but it was extremely minor. They used the same tweaked process for the "XT" Zen 2 skus and they gained very little from it (~2-3% higher performance at iso power if memory serves). The vast majority of Zen 3's improvement was from architecture and a more optimized layout.
I don’t see how N5P->N4P will be enough to get Zen 5 the same efficiency as Zen 4 isofrequency. It could have same or better perf/watt but it almost certainly won’t be at isofrequency.Edit : On the same token we can expect AMD to use the most refined 4nm for Zen 5 as if MLID s slides are legit there s no way that such a monster could be as efficient as Zen 4 at isoprocess and isofrequency.
I don’t see how N5P->N4P will be enough to get Zen 5 the same efficiency as Zen 4 isofrequency. It could have same or better perf/watt but it almost certainly won’t be at isofrequency.
This is what the article saysYou were wrong.
And for Zen 4 they mention that they use a customised process, we can eventually expect the same for Zen 5 :
AMD: We’re Using an Optimized TSMC 5nm Process
www.anandtech.com
Also, why did AMD say they will use the best "high performance" libraries for the target market? Standard cell for literally every core since Zen 2 (or even further back) is HD for logic, and Zen 3 moved to HD in SRAM as well for lower leakage.In order to avoid confusion, AMD is dropping the ‘+’ from its roadmaps. In speaking with AMD, the company confirmed that its next generations of 7nm products are likely to use process enhancements and the best high-performance libraries for the target market, however it is not explicity stating whether this would be N7P or N7+, just that it will be ‘better’ than the base N7 used in its first 7nm line.
MLID would've given an equally wide range but it would still turn out to be wrongMLID, is that you.??....
This is what the article says
Also, why did AMD say they will use the best "high performance" libraries for the target market? Standard cell for literally every core since Zen 2 (or even further back) is HD for logic, and Zen 3 moved to HD in SRAM as well for lower leakage.
I would take the words of the literal engineering slides from ISSCC presentation on Zen 3 in an actual technical conference over AMD marketing material lol.
MLID would've given an equally wide range but it would still turn out to be wrong
I agree. There are an abnormal amount of users defending him. I did not bring it up because I know what it is like to be singled out, but really...Also seems like mlid has friends on this forum (or his own accounts) promoting his videos.
They can create new models all they want. Half the X670/X670E boards have been discontinued. There is still some time before a Zen 5 launch.The motherboard vendors probably want a new model name. If they do, I can't see AMD not supporting their partners with one.
The reason is that Zen4 can currently support much higher memory clocks than it could at launch. Memory clock speed is a selling point printed in large type on MB packaging, and most current models only have 6400+, which was what it was possible to test when the motherboards were released. So MB vendors want to refresh their lineups, and when they do that, they usually want to have a new chipset name.
It's entirely possible that the chipset in question is literally PROM21, just rebranded.
This is the silicon world, not the software world. Once your designs are released, you can't iterate (for hardware changes) without launching a new product. There are no takebacks. Period. Oh and did I mentioned it costs several million dollars, EXCLUDING development (or marketing, or anything else) costs, to launch new hardware?Rarely isn't the same as never. Especially when a company focuses on agility, many things can be left of the table. Zen4c is a poster child for that. And as an engineer, you know there is such a thing as a free lunch--they just happen to be extremely rare.
But I don't think your categorization is correct.
Anything to do with cache and memory can increase IPC while reducing power consumption. That's a category you can pretty much always do something with.
And anything to do with layout can increase IPC (timings) while reducing power consumption (distances, voltages). That's a category that is never 100% optimal in this day of billions of transistors.
For the former, the most significant example I can think of is Maxwell memory compression: Huge increase to IPC with a massive decrease in power.
For the later, as an extreme example, take an Epyc processor (or even desktop Zen) and make it monolithic (or stacked).
It shows a mix, mostly because Zen3 runs at a higher frequency but has better voltage scaling. Taken as a whole, running clock-for-clock, volt-for-volt, Zen3 probably runs a bit hotter, but that's taking into account the aggregate changes. Unified L3 likely does both increase IPC and reduce power consumption, and there are most likely smaller optimizations that have some positive effect on both, documented or not. If you count changes that less directly improve power consumption by facilitating lower voltage, there are probably others of that kind also.
No, it won't, and AMD isn't changing lane configs. Such a change would require at minimum a new IO die, and more likely a socket change.If they actually had a new chipset, it could significantly better. The links from the CPU are all PCIE 5.0, a better chipset could provide twice the throughput.
But what Ajay and I are proposing is that the reason is just literally marketing. New x770 motherboards will seem fancier than last-gen x670 ones. Even if it's the same chip. There are reasons why vendors do rebrands.
Still strutting your stuff eh?Its a only a few AM5 motherboards that can run 8000MT/s stable in 2:1 mode, pretty much only two 1DPC from Asus and Gigabyte atm...
A new generation motherboards could improve improve memory layout/traces for the 2DPC boards.
At the moment they cap out at ~7400-7600MT/s if you want stability every reboot. (kinda behave like tuning memory on raptor lake, stability changes each reboot)
Below i did on my 1DPC GENE
View attachment 86626
That looks like a prerelease slide. It is inaccurate. Zen 3 was 7nm DUV with no process changes. Yes. They did that. It is impressive. 7nm+ is 6nm btw, and the 5xxx series did not use it. Later Zen 3 iterations, however, did.You were wrong.
And for Zen 4 they mention that they use a customised process, we can eventually expect the same for Zen 5 :
AMD: We’re Using an Optimized TSMC 5nm Process
www.anandtech.com
The Arrow Lake thing is just an unsubstantiated rumor at this point, however, SMT is going away EVERYWHERE. When it goes away is another question, The REASON it is going away is security. SMT opens up a ton of side channel attacks.Intel already has with Arrow Lake.
I suspect one of the reasons Zen 5 kept SMT is because Turin customers wouldn’t be happy with half the threads.
No, it was not, and the XT chips also weren't 'tweaked'. They were binned chips.The process was tweaked or optimized between the initial Zen 2 release and Zen 3, but it was extremely minor. They used the same tweaked process for the "XT" Zen 2 skus and they gained very little from it (~2-3% higher performance at iso power if memory serves). The vast majority of Zen 3's improvement was from architecture and a more optimized layout.
How do you know lnc will have a 800mhz clock deficit? As far as I know,, the intel projection slides only showed <10% is ST at 250W.That’s not IPC, that’s 1T performance. We don’t know what LNC will bring in IPC gains. Considering it’s probably going to lose ~800mhz of frequency and appears to be on a less than optimal node it’s got to have a decent IPC bump just to hit 5% 1T performance increase.
No, it was not, and the XT chips also weren't 'tweaked'. They were binned chips.
AMD specifically called out process refinements for increased boost clocks on the XT chips. When Zen 3 launched, they specifically called out they were using the same optimized process as the XT chips versus the original Zen 2 release.
View attachment 86642
N7+ is also said to deliver 10% higher performance at iso-power or, alternatively, up to 15% lower power at iso-performance. On paper, N7+ appears to be marginally better than N7P, albeit that comes at the cost of re-implementing the design.
Agree on this point, that s stated at wikichips that Zen 2 refresh used 7nm+, but there s some obscure reason as to why the XT version didnt have significantly better perf/watt because here what is said at wiki :
Obviously that s a tremendous improvement for a same node, yet only Zen 3 seemed to benefit from those numbers, unless of course that Zen 2 at the start used N7P.
Or possibly that AMD kept the same voltage/frequency curve for the XT to simplify things even if 7nm+ allowed a lower voltage at same frequency.
That's a mistake on wikichips part due to AMD originally listing Zen3 on their roadmaps as 7nm+ which they later recanted (or clarified) that they didn't mean to say that they were using the actual 7nm+ process, merely that it would be an enhanced process over the 7nm used on the original Zen2. Again, the process enhancement was minor, leading to a small (2 - 4%) increase in boost clocks as seen in the XT processors. Zen3 increased upon this with an improved architecture and complete rework of the layout.
We don't know for sure yet. There are also rumors that say Intel only disabled it because it wasn't working on the new architecture. And there was no time to fix it without massive delay. It might be fixed on the successor.Intel already has with Arrow Lake.