A Broadwell Retrospective Review in 2020: Is eDRAM Still Worth It?
by Dr. Ian Cutress on November 2, 2020 11:00 AM ESTBroadwell with eDRAM: Still Has Gaming Legs
As we crossover into the 2020s era, we now have more memory bandwidth from DRAM than a processor in 2015. Intel's Broadwell processors were advertised as having 128 megabytes of 'eDRAM', which enabled 50 GiB/s of bidirectional bandwidth at a lower latency of main memory, which ran only at 25.6 GiB/s. Modern processors have access to DDR4-3200, which is 51.2 GiB/s, and future processors are looking at 65 GiB/s or higher.
At this time, it is perhaps poignant to take a step back and understand the beauty of having 128 MiB of dedicated silicon for a singular task.
Intel’s eDRAM enabled Broadwell processors accelerated a significant number of memory bandwidth and memory latency workloads, in particular gaming. What eDRAM has enabled in our testing, even if we bypass the now antiquated CPU performance, is surprisingly good gaming performance. Most of our CPU gaming tests are designed to enable a CPU-limited scenario, which is exactly where Broadwell can play best. Our final CPU gaming test is a 1080p Max scenario where the CPU matters less, but there still appears to be good benefits from having an on-die DRAM and that much lower latency all the way out to 128 MiB.
There have always been questions around exactly what 128 MiB of eDRAM cost Intel to produce and supply to a generation of processors. At launch, Intel priced the eDRAM versions of 14 nm Broadwell processors as +$60 above the non-eDRAM versions of 22 nm Haswell equivalents. There are arguments to say that it cost Intel directly somewhere south of $10 per processor to build and enable, but Intel couldn’t charge that low, based on market segmentation. Remember, that eDRAM was built on a mature 22 nm SoC process at the time.
As we move into an era where AMD is showcasing its new ‘double’ 32 MiB L3 cache on Zen 3 as a key part of their improved gaming performance, we already had 128 MiB of gaming acceleration in 2015. It was enabled through a very specific piece of hardware built into the chip. If we could do it in 2015, why can’t we do it in 2020?
What about HBM-enabled eDRAM for 2021?
Fast forward to 2020, and we now have mature 14 nm and 7 nm processes, as well as a cavalcade of packaging and eDRAM opportunities. We might consider that adding 1-2 GiB of eDRAM to a package could be done with high bandwidth connectivity, using either Intel’s embedded multi-die technology or TSMC’s 3DFabric technology.
If we did that today, it could arguably be just as complex as what it was to add 128 MiB back in 2015. We now have extensive EDA and packaging tools to deal with chiplet designs and multi-die environments.
So consider, at a time where high performance consumer processors are in the realm of $300 up to $500-$800, would customers consider paying +$60 more for a modern high-end processor with 2 gigabytes of intermediate L4 cache? It would extend AMD’s idea of high-performance gaming cache well beyond the 32 MiB of Zen 3, or perhaps give Intel a different dynamic to its future processor portfolio.
As we move into more a chiplet enabled environment, some of those chiplets could be an extra cache layer. However, to put some of this into perspective.
- Intel's Broadwell's 128 MiB of eDRAM was built (is still built) on Intel's 22nm IO process and used 77 mm2 of die area.
- AMD's new RX 6000 GPUs use '128 MiB' of 7nm Infinity Cache SRAM. At an estimated 6.4 billion transistors, or 24% of the 26.8 billion transistors and ~510-530mm2 die, this cache requires a substantial amount of die area, even on 7nm.
This would suggest that in order for future products to integrate large amounts of cache or eDRAM, then layered solutions will need to be required. This will require large investment in design and packaging, especially thermal control.
Many thanks to Dylan522p for some minor updates on die size and pointing out that the same 22nm eDRAM chip is still in use today with Apple's 2020 base Macbook Pro 13.
120 Comments
View All Comments
Billy Tallis - Wednesday, November 4, 2020 - link
Ian already said he tests at JEDEC speeds, which includes the latency timings. Using modules that are capable of faster timings does not prevent running them at standard timings.Quantumz0d - Tuesday, November 3, 2020 - link
Don't even bother Ian with these people.Nictron - Wednesday, November 4, 2020 - link
I appreciate the review and context over a period of time. Having a baseline comparison is important and it is up to us the reader to determine the optimal environment we would like to invest in. As soon as we do the price starts to skyrocket and comparisons are difficult.Reviews like this also show that a well thought out ecosystem can deliver great value. Companies are here to make money and I appreciate reviewers that provide baseline compatible testing over time for us to make informed decisions.
Thank you and kind regards,
GeoffreyA - Tuesday, November 3, 2020 - link
Thanks, Ian. I thoroughly enjoyed the article and the historical perspective especially. And the technical detail: no other site can come close.eastcoast_pete - Tuesday, November 3, 2020 - link
Ian, could you comment on the current state of the art of EDRAM? How fast can it be, how low can the latency go? Depending on those parameters and difficulty of manufacturing, there might be a number of uses that make sense.One where it could is to possibly allow Xe graphics to use cheaper and lower power LPDDR-4 or -5 RAM without taking a large performance hit vs. GDDR6. 128 or 256 MB EDRAM cache might just do that, and still keep costs lower. Pure speculation, of course.
DARK_BG - Tuesday, November 3, 2020 - link
Hi , what I'm wondering is where the 30% gap between the 5770C and 4790K in Games came from , compared to your original review and all other reviews out there of 5770C. Since I'm with a Z97 platform and 4.0GHz Xeon , moving to 4770k or 4790K doesn't make any sense given their second hand prices but 5770C on this review makes alot of sense.So is it the OS,the drivers , some BIOS settings or on the older reviews the systems were just GPU limited failing to explore the CPU performance?
jpnex - Friday, January 8, 2021 - link
Lol, no, the I7 5775c is just stronger than an i7 4790k, this is a known fact. Other benchmarks show the same thing. Old benchmarks don't show It because back then people didn't know that deactivating the iGPU would give a performance boost.DARK_BG - Wednesday, July 20, 2022 - link
I forgot back then to reply back , based on this review I've sourced 5775C (for a little less than 100$ this days going for 140-150$) coupled with Asus Z97 Pro and after some tweaking (CPU at 4.1GHz , eDRAM at 2000MHz and some other minor stuff that I already forgot) the difference compared to the Xeon 4.0GHz in games was mind blowing.Later I was able to source and 32GB Corsair Dominator DDR3 2400MHz CL10 just for fun to make it top spec config. :)It is a very capable machine but this days I'll swap it for Ryzen 5800X3D to get the final train on the fastest Windows 7 capable gaming system.Yeah i know it is OLD OS but everything I need runs flawessly for more than a decade with only reainstall 7 years ago due to an SSD failure. It is my only personal Intel System for the past 22 years since it was the for a first time the best price performance second hand platform for a moment , all the rest were AMD based and I keep them all in working condition.
BTW I was able to run Windows XP 64bit on the Z97 platform , I just need to swap the GTX 1070 for GTX 980/980 Ti to be fully functional everything else runs like a charm under XP i was able to hack the driver to install as an GTX 960 so I have a 2D hardware acceleration under XP on GTX 1070 since nvidia havent changed anything in regard to 2D compared to the previous generation
dew111 - Tuesday, November 3, 2020 - link
Rocket lake should have been the comet lake processor with eDRAM. Instead they'll be lucky to beat comet lake at all.erotomania - Tuesday, November 3, 2020 - link
Thanks, Ian. I enjoyed this article from a NUC8i7BEH that has 128MB of coffee-flavored eDRAM. Also, thanks Ganesh for the recent reminder that Bean > Frost.