Video summary

AMD Ryzen 9 3950X Review: Premiere, Blender, Overclocking, & Gaming CPU Benchmarks

Main summary

Key takeaways

Product Review

Product reviewed

AMD Ryzen 9 3950X (16-core / 32-thread) CPU, benchmarked stock and overclocked. The review emphasizes:

  • Premiere
  • Blender
  • Photoshop
  • General productivity workloads
  • Gaming
  • Thermals and power

Key features highlighted

  • 16C / 32T CPU, with advertised boost up to 4.6 GHz
  • “GPU-like” boost behavior: frequency varies with thermals and power budget, so “boost” can’t be assumed as sustained under all conditions
  • TDP rating: 105W, but AMD updated cooler guidance to an 80mm CLC, making the TDP number less directly useful for real-world thermal expectations
  • Overclocking capability
    • Most charts use an AIO (Corsair “X62” class 280mm CLC)
    • Mentions liquid nitrogen (LN2) testing later (stream goal): study frequency scaling under extreme temperatures

Benchmarks / performance findings (main takeaways)

1) Frequency behavior (stock behavior)

  • Blender all-core (stock): stabilizes around ~3.9 GHz once steady state is reached, with small fluctuations thereafter
  • Cinebench single-thread max spikes:
    • Typically ~4.0–4.1 GHz
    • Brief spikes up to ~4.5 GHz
    • Not sustained at the marketing-style 4.6 GHz level

Marketing claim skepticism: reviewers question how meaningful “4.6 GHz” is if it occurs only under highly limited, short-lived conditions.

2) Overclocking results (with 280mm CLC)

  • Max stable all-core result: ~4.4 GHz (limited by temperature/voltage with their cooling)
  • Voltage used: ~1.3625V set, with motherboard-reported “actual” Vcore closer to ~1.35V
  • Suggestion: with better cooling (custom loop / stronger 360mm AIO), users might reach ~4.5–5.0 GHz, but not necessarily:
    • All cores
    • 24/7
    • At the same voltage

Comparison note: the 3950X needed less voltage for similar clocks than the 3900X in their testing.

Productivity workload results

Adobe Premiere (render benchmarks @ 1080p)

  • Stock 3950X: 2.8 minutes
  • Stock 3900X: 3.3 minutes
  • Result: ~15% faster than the 3900X

Overclocked 3950X (~4.4 GHz):

  • ~2.4 minutes (about ~14% faster than stock 3950X)

Intel comparisons (positioning/value):

  • i9-7980XE (~$980, 18-core): 2.9 minutes stock
  • i7-9900K (~$260 cheaper): slower in this test than the 3950X

Verdict for Premiere

Reviewers strongly endorse the 3950X for money/value for Premiere work.

Adobe Photoshop (time/score; frequency important)

  • Photoshop is framed as frequency-first, with less benefit from additional threads
  • The 3950X lands close to the 9900K:
    • It can technically win in their results
    • But the 9900K may be better value due to price

Review note: the 3950X is expected to perform well in Photoshop relative to older AMD struggles, reflecting improvement vs previous Ryzen generations.

Blender (production rendering; non-synthetic, tile-based scaling)

  • Blender scales with more threads, but not perfectly linearly
  • GM logo render:
    • 3950X stock: 12 minutes
    • 3900X stock: ~22% slower (implied relative gap per review)

Overclocking:

  • 3950X at 4.4 GHz: about ~16% less render time than stock

Intel positioning:

  • i9-9900K is not competitive in this thread-heavy rendering thread
  • i9-7980XE is “above” but described as not worth it by value unless you pay for it for other reasons

Verdict for Blender

The 3950X is positioned as a great productivity chip for tile/thread-dependent rendering.

Additional production workloads

  • Adobe Media Encoder:
    • 3950X stock: ~8.2 minutes
    • 3900X stock: ~9 minutes
    • 7980XE stock: ~7.7 minutes
    • 9900K stock: ~11.9 minutes
  • 7-Zip decompression:
    • 3950X stock: 186 MIPS (about ~86% of 2990WX)
    • Review also notes the 3950X leads similarly in the stack relative to an i9-7980? at 4.6 GHz (per subtitles context)
  • V-Ray: skipped/insufficient context from subtitles

Gaming results (limited; often GPU-limited)

Overall: at 1080p/1440p, gaming is commonly GPU-limited, so CPU differences shrink.

  • General observation: 3950X looks similar to other high-end CPUs, not a clear FPS monster
  • Read Dead Redemption 2:
    • GPU-limited
    • 3950X: ~140 FPS average
    • i9-9900K: ~151 FPS max under their settings
    • Intel leads, but it’s framed as not meaningfully different for most people
  • Civics (time/turn):
    • 3950X: 30.1s stock (subtitles suggest ~30.7s overclocked, and compare vs 3900X)
    • 9900K: 29.7s
  • GTA V:
    • 3950X and 3900X are close
    • Intel still leads in some scenarios

Game-specific scheduling/behavior quirks:

  • Shadow of the Tomb Raider: possible latency/core scheduling issues; “experience better” on 3900X or 9900K
  • Total War Warhammer II campaign: excess cores can hurt performance; workaround suggested:
    • disable cores via software/BIOS (aligns with “manual core disabling” theme)

Quirk summary

Higher-core Ryzen chips can have game-specific problems, where the fix is sometimes disabling cores per title.

Power + thermals

Power draw

  • Blender render (stock):
    • 3950X: ~137W
    • 3900X: ~147.6W
  • Overclocked 4.4 GHz:
    • 3950X: ~246–262W (tuning dependent)
    • 3900X: ~170W at 4.3 GHz / 1.34V (as described)

Implication: efficient stock behavior, but overclocking becomes power-hungry quickly.

Thermals (280mm AIO)

  • Stock Blender hottest CCD: ~60–62°C
  • Overclock: ~95°C sustained in their setup (strongly implying 280mm AIO may be insufficient for sustained high clocks)

Pros / cons mentioned

Pros

  • Strong productivity performance
    • Premiere and Blender show clear improvements vs 3900X
    • Photoshop is competitive, with improved AMD positioning vs older generations
  • Excellent value-per-performance for the workflows reviewed
  • Efficient stock behavior (better power than 3900X in their Blender test)
  • Overclocking headroom to ~4.4 GHz stable with AIO; potentially higher with stronger cooling

Cons

  • Gaming isn’t a standout win
    • CPUs often converge due to GPU limitations
  • Some games dislike high core counts
    • Example: Total War campaign; workaround may be needed
  • Boost marketing ambiguity
    • 4.6 GHz is framed as mostly opportunistic/single-peak, not realistic sustained behavior
  • Overclocking can be inefficient
    • high power and high temps with only 280mm CLC cooling

Comparisons explicitly made

  • Against AMD Ryzen 9 3900X
    • Major Premiere/Blender improvements
    • 3950X often achieves similar clocks with less voltage
  • Against Intel Core i9 / HEDT parts
    • i9-9900K: frequently best as a value-focused alternative (especially gaming and Photoshop)
    • i9-7980XE / similar HEDT: sometimes competitive or better in some places, but expensive
    • Mentions i9-10980XE / upcoming ~10th gen refresh tier (higher price tier; price implied around ~$1000)
  • Against power rivals like 2990WX
    • 3950X doesn’t always beat it (e.g., decompression/render timing specifics), but remains competitive

Overall verdict / recommendation

Buy the Ryzen 9 3950X if you do serious productivity work—especially:

  • Adobe Premiere
  • Blender / tile-based rendering
  • Heavy multi-threaded workflows where rendering time matters

Don’t buy it primarily for gaming, because:

  • Gains over cheaper high-end CPUs are often small (GPU-limited scenarios)
  • Some titles may perform worse due to core-count scheduling/behavior quirks

Practical positioning from the review

  • Best fit: users “making money” / production workloads
  • Better gaming pick for most: commonly i9-9900K
  • Photoshop-only-ish users: 9900K may be better value

Unique points list (distinct claims extracted)

  1. 3950X presented as a high-end desktop functionally: 16C/32T
  2. No meaningful mainstream competitor at its tier; comparisons flank it with Intel 9900K and AMD 3900X
  3. Planned LN2 overclock stream to study frequency vs temperature scaling
  4. Clarifies boost behavior: 4.6 GHz advertised is mostly opportunistic/single-peak
  5. TDP 105W; AMD cooler guidance changed to 80mm CLC, reducing real-world usefulness of TDP
  6. Stock Blender all-core frequency stabilizes after ~100s with fluctuations around ~±10 MHz
  7. Cinebench single-thread max: ~4.0–4.1 GHz with brief ~4.5 GHz spikes; not sustained
  8. BIOS differences affect Cinebench R15 frequency stability; used F10C
  9. With 280mm CLC, stable OC reached ~4.4 GHz; 4.5+ possible with better cooling
  10. Overclock tuning described; 3950X needs less voltage than 3900X for similar clocks
  11. Premiere: 2.8 min vs 3.3 min (~15% faster than 3900X)
  12. Premiere: 7980XE stock slower than 3950X; Intel value positioning noted
  13. Premiere overclock improvement: ~14% vs stock
  14. Photoshop: frequency matters; 3950X close to 9900K and can slightly beat it, but 9900K may be better value
  15. Blender: 3950X completes 12 min; 3900X is ~22% slower
  16. Blender overclock reduces render time by ~16%
  17. GM logo Blender: 3950X competitive vs higher Intel parts, though Intel can still lead on value
  18. Blender scaling isn’t linear (example: 2990WX isn’t “half the time”)
  19. Media Encoder: 3950X 8.2 min, 3900X 9 min, 9900K 11.9 min
  20. 7-Zip: 3950X 186 MIPS (~86% of 2990WX)
  21. Gaming at tested settings is heavily GPU-limited
  22. RDR2: 3950X ~140 FPS average; 9900K has higher max in their limits
  23. Civics: 3950X slightly behind 9900K in turn-time
  24. GTA V: 3950X and 3900X close; Intel leads in some scenarios
  25. Hitman: 3950X ahead of 3900X; Intel still leads vs 3950X in some metrics
  26. Shadow of Tomb Raider: possible AMD/core scheduling issues; experience may be better on 3900X/9900K
  27. Assassin’s Creed Origins: 3950X edges averages; Intel leads but deltas are small
  28. Total War: extra cores can hurt; campaign benchmark aligns with 2990WX “game mode” behavior; suggests disabling cores if problematic
  29. Power efficiency: Blender stock 137W vs 147.6W
  30. Overclock power: 3950X OC 246–262W vs 3900X OC
  31. Thermals: stock ~60–62°C; OC sustained around ~95°C on their AIO
  32. Final guidance: not for gaming-only; strong for Premiere/Blender/Photoshop
  33. Motherboard VRM matters: avoid weak boards meant for low-core CPUs (example given: “don’t buy a weak board meant for 3600”)
  34. Mentions upcoming Intel Core i9-10980XE / ~10th gen refresh, likely around ~$1000
  35. Ending tone: generally positive, but points out specific games with worse behavior and workarounds

Speakers / views

  • Subtitles appear to reflect one main presenter voice.
  • No clearly distinguishable alternate speakers contribute separate perspectives beyond the same reviewer’s testing narrative.

Original video