Phenom vs Ryzen: Hexa-cores a decade apart (Part 4)

CPU Benchmarks

Before we get into the benchmarks, here is a specification comparison of the 3 processors (the dual core Phenom II X2 550 will be compared in a separate section).

Starting off with the popular Cinebench benchmarks, Ryzen 5 5600X gets off to a flying start (more than flying start I should say) with a 319% lead over Phenom II X6 1055T in Cinebench R15. This goes on to show how misleading CPU comparison by core count can be.

Though both the processors are hexa-core parts, the performance disparity of over 4 times tells a different story which cannot be discerned by core count alone. Even with Simultaneous Multi-threading (SMT) disabled, I recon 5600X will still score more than double that of 1055T.

Moving on to Cinebench R20, we see more of the same. Single-core score of 5600X is much closer to the multi-core score of 1055T than its single-core score. That statement alone summarizes the graph quite nicely. 5600X was 223% faster than 1055T in single-core with its multi-core score was in a league of its own (nearly 4.5 times faster).

While the newer Cinebench R23 is said to measure performance of more recent processors better, it was still included in the test suite, for the sake of curiosity. As with R20, 5600X blows its fellow hexa-core processor out of water in this test, besting it by 226%. As for multi-core score, well lets just say if I didn't know any better, I would say we are comparing a dual-core processor to an octa-core one. 5600X was a whopping 356% faster than the 1055T.

Next up with Geekbench 4, we see the same trend follow. 5600X outpaced the 1055T by 198% in single-core test and 259% in multi-core. If you notice in this and the previous graphs, the single-core score doesn't scale linearly with multi-core score i.e. for a hexa-core CPU, its multi-core score isn't 6 times its single core one. That's partly because of the boost frequency. Processors can generally boost higher on 1 or 2 cores than on all cores/threads. You can expect to get higher clock speeds with better cooling, more so with the all-core boost frequency.

The single-core score of 5600X comes dangerously close to the multi-core score of 1055T in Geekbench 5, although it is bound to favour newer hardware. Zen 3 is a work of art as much as it is an engineering feat. With Zen 3 Ryzen got a massive gen-on-gen boost to IPC, second only to the gains made on 1st generation of Ryzen over FX series. This is all the more impressive considering AMD didn't transition to a smaller node that will allow for higher frequencies, transistor count and efficiency.

A point of note, Geekbench 5 and 4 as well takes into account memory transfer speed and latency in its single and multi-core tests. This system having only a single 8 GB RAM module is running in single channel (Not that it necessarily means a single memory module will always be single channel. More on that later), which will net in lower memory transfer speed than running in dual-channel. This lowers the overall Geekbench scores in both the tests across both versions. In Geekbench 5, good multi-core scores of 5600X even on stock cooler is about 7500.

So while having higher percentile score among all 5600X scores in the Geekbench 5 database, the 5600X in this system is still 251% and 247% faster than the Phenom II X6 1055T in single-core and multi-core respectively, which is nothing to sneeze at.

PCMark 10 benchmark covers a wide range of activities from everyday productivity tasks to demanding digital media content. Here the 5600X is 97% faster than the 1055T, which doesn't sound like a whole lot considering the performance deltas we saw in the previous benchmarks. This is another benchmark that takes a hit with the slower single-channel memory as stock scores with 5600X is around 7200-7500.

Individual categorized scores saw 5600X being 68% faster in Essentials, 115% faster in Productivity and 113% faster in Digital Content Creation vs. the 1055T.

Performance Test from PassMark is another benchmarking tool to measure and compare performance online with millions of PCs in their online database. The 5600X eclipses the 1055T by a massive 550%. that's the largest performance delta we have seen yet (spoiler alert! there is higher still). The landslide victory for Ryzen continues in this battle of 2 hexa-cores.

Looking at another synthetic benchmark, 3DMark 11 Fire Strike Physics score sees 5600X outdo the 1055T by 305%.

Blender is a very popular cross-platform open-source 3D content creation tool. Blender tests has always been one of the strengths of Ryzen Series processor and this one is no different, though it would be more insightful when compared to more modern processors of its class.

Testing with the bmw37 scene, 5600X provides a sizable performance uplift over 1055T at 350%. In other words, the Ryzen chip cuts down total render time per scene from 18 minutes to a mere 4 minutes! Scaling linearly, an hour long workload on 1055T of comparable complexity might take only about 13 minutes for 5600X to render. For the asking price, that's quite the value you get in time saved from a mainstream hexa-core processor.

While Ryzen 5 5600X has its strengths in gaming, it proves to be no slouch in productivity and multimedia tasks either. You can dabble into every sort of workloads with this one it would seem.

Handbrake is an open-source cross-platform transcoder. I primarily use Handbrake to compress videos with h.265 codec (also referred to as High-Efficiency Video Coding (HEVC)) to reduce file size while preserving quality. While inferior to the newer standard, h.264 (AVC) is still popular, specially for 1080p or lower video encoding.

5600X handily beats down the 1055T by 242% Though a larger video file would have made for a more insightful benchmark, I didn't quite have files of proper size that wouldn't take up 30 minutes or more to convert.

With 4K resolution becoming the new standard for videos, file size needs to be manageable. HEVC, the successor to AVC or h.264 is able to compress information more efficiently the trade-off being higher compute requirement. Most of the modern processor will have hardware support for HEVC codec, accelerating encoding and decoding.

Phenom II X6 1055T belongs to a much older era where AVC was the best option out there and thus doesn't have hardware acceleration support for newer standards. This disparity is clearly evident from the graph. If this was a race and 5600X turned to check how far ahead she is, 1055T would be nowhere in sight! Ryzen completed the race within half a minute while 1055T took over 4 minutes to reach the finish line. The match was rigged from start but this staggering 728% difference is just brutal.

This still isn't the final indicator of 5600X's full potential in HEVC workloads. HEVC workloads tends to stress the CPU threads to its utmost limits which isn't obvious from the CPU utilization percentage in task manager or other softwares like HWInfo. Both Cinebench R20 multi-core and Handbrake HEVC encoding had the CPU utilization at 100% but that doesn't show the full picture of exactly how much of the processors capabilities is being used.

In loose terms, utilization percentage value is a measure of how long threads in a kernel(sequences of instructions in a block of code) are currently being processed by the CPU, as a percentage of the time allocated for that thread. Threads might be active ('busy') or waiting for data from cache/memory ('stalled') or finished ('idle'). Busy and stalled threads get classified in the same way, so a high utilization figure does not necessarily mean the processor is being worked very hard, and the % value does not tell you how much of the processor's capabilities being used. More on that here in this article from techspot.

During both h.264 and h.265 test, 5600X had a sustained temperature of around 92C with peaks touching 95C. This was also one of the reasons for the choice of shorter video file in the tests. AMD does say temperatures up to 95C is in operable range with the stock cooler. Still, I can't have my daily driver CPU frying up on me accidentally. So, there's that.

Next up we have compression and decompression tests using in-built benchmark of 7-zip. As in life, packing is a lot harder than unpacking here too. Ryzen is an absolute beast when it comes to decompression, easily verifiable from multiple professional reviewer's takes. High decompression performance is particularly useful when installing large softwares and PC games as a faster decompression rate will result in faster unpacking and installation.

The 5600X takes a comfortable 399.5% lead in compression speed and 288% in decompression.

Coming up next are the AIDA64 cache benchmarks. Before that, some context first.

Modern CPUs are incredibly fast. They take just one clock cycle to add two 64 bit integer values together, and for a CPU running at 4 GHz, that would be just 0.00000000025 seconds or a quarter of a nanosecond (reference here). Let that sink in. To keep the CPU constantly fed with data and instructions we need a fast enough storage system, which just doesn't exist yet. Even the DRAM, boasting transfer speeds north of 25 GB/s is slow as it still takes about 100 ns to find data, never mind hard disks or even Solid State Drives(SSDs). This necessitates another stage of memory - cache. Cache consists of multiple stages of SRAM which are degrees of magnitude faster than DRAM but as with every technology, there are trade offs - that here being storage and cost.

Before diving into the results, a note. This test was originally done with BIOS v4.00 AGESA 1.2.0.0 which had a L3 cache performance issue. The test was redone on BIOS v4.10 AGESA 1.2.0.2 that fixed the issue. With the new BIOS, L3 cache bandwidth almost doubled, as we will see later.

Again we find 5600X taking the upper hand by quite the margin in L1 cache test. We are looking at a 226% improvement in L1 cache read speed, a 386% improvement in L1 cache copy speed and a 229% improvement in L1 cache write speed.

At this point we have come to expect large differences between the two processors. Even so, although we expect a landslide win for 5600X, the performance delta here is astounding to say the very least. We are talking about a 623% uplift in read speed, 650% uplift in copy speed and an insane 1338% uplift in write speed! Cache memory has come a long way since 2010 evidently.

AMD completely overhauled the cache system in Ryzen 5000 series. With the new unified cache system, each of the 8 cores in a CCX will now have access to the entirety of 32 MB L3 cache as they are grouped into a single unit of resource. This dramatically reduces core-to-core and core-to-cache latencies. Latency-sensitive tasks like running games especially benefit from this change, as tasks now have direct access to twice as much L3 cache versus Zen 2.

Though these improvements could only be discerned by comparison with previous generation, it adds to the performance improvement over 1055T nonetheless.

The L3 cache test turned out to be an absolute stomp. 5600X relegated the 1055T to imperceptible level by establishing a tremendous lead of 862% in read speed, 872% in copy speed and a gigantic 1644% in write speed.

Cache is intimately intertwined with IPC of a CPU- its size, bandwidth and latency plays a significant role in gen-on-gen increase in IPC. Much of the 19% IPC gain of Ryzen 5000 series over 3000 series came from reworked cache design. Lower latencies definitely help.

Taking a look at cache latencies, there is virtually no difference in L1 cache latency between the two processors. We see a significant 246% reduction in L2 cache latency of 5600X over 1055T while L3 cache latency is lower by 173%.

Summarizing, we saw Ryzen 5 5600X outshined Phenom II 1055T across the entire test suite with anywhere from 2 times to a massive 17 times performance improvement. Things get really interesting when we break it down to absolute margins.