ThunderX: From Small & Simple to Wide & Complex

As a brief recap, the original ThunderX was an improved version of the Octeon III: a dual-issue in-order CPU core with two short pipelines.

The advantage of the original ThunderX design is that such a simple core can be very energy efficient, especially for "low ILP" (instruction level parallelism) workloads such as web servers and most database servers. Of course, such a short pipeline limits the clockspeed, and such a simple in-order design offers low single threaded performance in medium and high ILP workloads, whereas more advanced out-of-order processors can extract significant parallelism.

Cavium's "New" Core: Vulcan

Relative to the original ThunderX, the Vulcan core of the ThunderX2 is an entirely different beast. Announced in 2014 by Broadcom, it is a relatively wide core that runs 4 simultaneous threads (SMT4). As a result, the wide back end should be quite busy even when running low-ILP server workloads.

To make sure that all 4 SMT threads can be sustained, the ThunderX2 front-end can fetch up to 64 bytes from the 8-way set associative 32 KB instruction cache, which is outfitted with a simple next line prefetcher. However, fetching 8 instructions is only possible if there is no taken branch inside those 64 bytes. In that case, the fetch breaks off at the taken branch.

That means that in branch intensive code (databases, AI...) the fetcher will get +/- 5 instructions per clock cycle on average, as one out of 5 instructions is a branch. The fetched instructions are then sent to a smoothing buffer – a buffer where the fetched instructions are held for decoding.

The decoder will then work on a bundle of 4 instructions. Between the decoder and the rename phase each thread has "skid buffer" which consists of 8 bundles. So between the 4 threads up to 32 bundles (128 instructions) can be skid buffered at any one time..

Those 4 instructions – a bundle – travel together through the pipeline until they reach the unified issue queue of the scheduler. Just like Intel has implemented in Nehalem, there is also a loop buffer and predictor, which Intel used to call a "Loop Stream Detector". This loop buffer avoids branch mispredictions and contains decoded µops, which "shortens" the pipeline and reduces the amount of power spent on decoding.

Overall, up to 6 instructions can be executed at the same time. This is divided into 2 ALU/FP/NEON slots, 1 ALU/branch slot, 2 load/store slots (16 bytes), and 1 pure store slot that sends 16 bytes to the D-cache. There is a small (Cavium would not disclose how small) L1 TLB for zero latency translation from Virtual to physical addresses. There is no hardware prefetcher for the L1 D-cache, but the L2 cache has a rather complex hardware prefetcher which is able to recognize patterns (besides being able to stride or fetching the next line).

This is enough to feed the back-end, which can sustain 4 instructions per cycle from 4 different threads.

Micro Architecture Differences

Ultimately Cavium has only published a limited amount of information on the ThunderX2 cores, so there are some limits to our knowledge. But we've gone ahead and summarized some of the key specifications of the different CPU architectures below.

Feature Cavium ThunderX2
L1-I cache
32 KB
(+ 24 KB L0)
32 KB
L1-D cache
32 KB
32 KB
32 KB
32 KB
Load Bandwith 2x 16B 2x 16B 2x 32B 2x 16B
L2-cache 256 KB
256 KB
1 MB
512 KB
Fetch Width 8 instructions 4 instructions 16 bytes (+/- 4-5 x86) 32 bytes (+/- 6-8 x86)
Issue Queue 60 76 97 unified 6x14
Sustainable Instructions/cycle 4 4 5-6 4-5
Instructions in Flight 180 (ROB) 128 224 (ROB) 192
Int. Pipeline Length


15 stages 19 stages
14 stage from µop cache
19 stages?
TLB Instructions
TLB Data
"Small L1" + 2048 unified L2 ?
+1536 Unified

A detailed analysis is out of the scope of this article. But you can read Ian's analyses of the Falkor, Skylake and Zen architectures here at AnandTech. We limit ourselves to the most obvious differences.

It is pretty clear that Intel's single-threaded performance remains unchallenged: the Skylake core is the widest core, keeps the most instructions in flight, and most importantly runs at the highest clockspeed. The ThunderX2 core is the one that fetches the most instructions per cycle, as it has to be able to keep 4 threads running. The fetch unit will grab 8 instructions from one thread, than grab 8 from the second thread and it will keep cycling between threads. A bad prediction could thus lower the performance of single thread significantly.

Sizing Things Up: Specifications Compared The ThunderX2 SKUs: 16 to 32 Cores


View All Comments

  • JohanAnandtech - Thursday, May 24, 2018 - link

    I have been trouble shooting a Java problem for the last 3 weeks now - for some reason our specific EPYC test system has some serious performance issues after we upgraded to kernel 4.13. This might be a hardware/firmware... issue. I don't know. I just know that the current tests are not accurate. Reply
  • npz - Thursday, May 24, 2018 - link

    Large Pages should be used whenever possible on Intel. You do waste some more memory, but it's worth it for most workloads as can be seen in your Intel Java benchmark. We've tested it for IO devices and enabling large pages in drivers to do DMA to shows a big difference for some high throughput devices. Reply
  • junky77 - Thursday, May 24, 2018 - link

    What? A 2.5GHZ ARM core is around 60-70% of a 3.8GHZ Skylake core?? For 3.8GHZ, the ARM is probably at least as fast? Reply
  • Wilco1 - Thursday, May 24, 2018 - link

    Probably around 90% since performance doesn't scale linearly with frequency. Note these are throughput parts so won't clock that high. However a 7nm version might well reach 3GHz. Reply
  • AJ_NEWMAN - Thursday, May 24, 2018 - link

    If Caviums tweaked 16nm hits 3GHz - it would to be unreasonable to aim for 4GHz for a 7nm part.

    With 2.3 times as many transistors available - it will be interesting to see what else they beef up?

    HIgher IPC? 64 cores? 16 memory controllers? CCIX - or perhaps they will compete with Fujitsu and add some Supercomputer centric hardware?

  • meta.x.gdb - Thursday, May 31, 2018 - link

    Wonder why the VASP code limped along on ThunderX2 while OpenFOAM saw such gains. I'm pretty familiar with both codes. VASP is mostly doing density functional theory, which is FFT-heavy... Reply
  • Meteor2 - Tuesday, June 26, 2018 - link

    All I want to say (all I can say) is that Anandtech has some of the best writers and commenters in this field. Fantastic article, and fantastic discussion. Reply
  • paldU - Saturday, July 7, 2018 - link

    A typo in Page 2. "it terms of performance per dollar" should be " in terms of performance per dollar". Reply

Log in

Don't have an account? Sign up now