By: Chester (lamchester.delete@this.gmail.com), May 22, 2020 12:06 pm
Room: Moderated Discussions
Doug S (foo.delete@this.bar.bar) on May 22, 2020 11:09 am wrote:
> none (none.delete@this.none.com) on May 22, 2020 3:39 am wrote:
> > Just looking at the picture the L2 labelled part is less than 8 times larger than the L1 one.
> > Isn’t Skylake SP 32KB L1 and 1024KB L2? If so is L2 really more than 4 times denser than L1?
> > Or is the die shot incomplete? Or did I miss something else?
>
>
> It isn't just the number of KB, it also depends on the associativity, tags, amount of read/write I/Os
> for each and so forth. I would expect an L2 to be denser than L1, and an L3 to be denser still.
Different SRAM types are used in L1 and L2. So yeah, L2 would be a lot more dense. Pretty much everyone does this - L2/L3 are a lot denser at the cost of being higher latency.
Also I didn't point it out, but the SRAM blocks near the L2 would be for tags.
> none (none.delete@this.none.com) on May 22, 2020 3:39 am wrote:
> > Just looking at the picture the L2 labelled part is less than 8 times larger than the L1 one.
> > Isn’t Skylake SP 32KB L1 and 1024KB L2? If so is L2 really more than 4 times denser than L1?
> > Or is the die shot incomplete? Or did I miss something else?
>
>
> It isn't just the number of KB, it also depends on the associativity, tags, amount of read/write I/Os
> for each and so forth. I would expect an L2 to be denser than L1, and an L3 to be denser still.
Different SRAM types are used in L1 and L2. So yeah, L2 would be a lot more dense. Pretty much everyone does this - L2/L3 are a lot denser at the cost of being higher latency.
Also I didn't point it out, but the SRAM blocks near the L2 would be for tags.
| Topic | Posted By | Date |
|---|---|---|
| High res SKL-SP die shot | Travis Downs | 2020/04/27 04:19 PM |
| High res SKL-SP die shot | anonymou5 | 2020/04/27 11:15 PM |
| Huh? | Huh | 2020/05/22 10:12 PM |
| Huh? | Freedy | 2020/05/24 07:43 AM |
| Huh? | anonymou5 | 2020/05/24 12:40 PM |
| High res SKL-SP die shot | Chester | 2020/05/01 05:13 PM |
| High res SKL-SP die shot | anon person | 2020/05/01 10:44 PM |
| High res SKL-SP die shot | anon | 2020/05/02 09:11 AM |
| High res SKL-SP die shot | anonymou5 | 2020/05/02 03:17 PM |
| High res SKL-SP die shot | wumpus | 2020/05/03 06:53 AM |
| High res SKL-SP die shot | Travis Downs | 2020/05/21 08:28 PM |
| High res SKL-SP die shot | Brett | 2020/05/22 12:41 AM |
| High res SKL-SP die shot | Chester | 2020/05/22 12:25 PM |
| High res SKL-SP die shot | Brett | 2020/05/22 01:34 PM |
| High res SKL-SP die shot | none | 2020/05/22 02:39 AM |
| High res SKL-SP die shot | Doug S | 2020/05/22 10:09 AM |
| High res SKL-SP die shot | Chester | 2020/05/22 12:06 PM |
| High res SKL-SP die shot | none | 2020/05/23 02:15 PM |
| High res SKL-SP die shot | David Kanter | 2020/05/23 07:57 PM |
| High res SKL-SP die shot | Chester | 2020/05/22 12:01 PM |
| High res SKL-SP die shot | Travis Downs | 2020/05/03 12:01 PM |