AMD finally delivers dual 3D V-Cache on Zen 5 with the 9950X3D2, but does twice the cache translate into real gains? We test performance, power, and value to see if this flagship makes sense.
AMD Ryzen 9 9950X3D2 Review: Twice the 3D V-Cache
- Thread starter Steve
- Start date
Sounds to me like AMD might be trying to play the superior product so superior pricing game. Maybe people will buy this CPU, but maybe not. If Intel is significantly cheaper on a price/performance ratio, then AMD only stands to loose to their own corporate hubris.
Corporate hubris, IMO, bit Intel, and many other companies, in the butt; however, AMD, or so it seems, has failed to learn the lesson.
We saw this kind of back and forth between AMD and Intel back in the days of Intel's Core 2. For me, its a big turn off even though I'm primarily an AMD buyer. ATM, I'm not compelled to buy this latest iteration of the part, and I hope lack of interest in the consumer realm causes AMD to finally see that name does not outweigh price/performance ratio.
Corporate hubris, IMO, bit Intel, and many other companies, in the butt; however, AMD, or so it seems, has failed to learn the lesson.
We saw this kind of back and forth between AMD and Intel back in the days of Intel's Core 2. For me, its a big turn off even though I'm primarily an AMD buyer. ATM, I'm not compelled to buy this latest iteration of the part, and I hope lack of interest in the consumer realm causes AMD to finally see that name does not outweigh price/performance ratio.
Squid Surprise
Posts: 7,774 +8,061
"As things stand today, Intel leads for productivity, AMD leads for gaming, and that is going to do it for this review."
Maybe change to "Intel leads for productivity VALUE". AMD DOES still lead for productivity - just charges a heck of a lot more...
At $900 for a CPU, I'd be expecting a lot more... at least it's still "cheap" compared to Threadripper...
Maybe change to "Intel leads for productivity VALUE". AMD DOES still lead for productivity - just charges a heck of a lot more...
At $900 for a CPU, I'd be expecting a lot more... at least it's still "cheap" compared to Threadripper...
MasterMace
Posts: 400 +414
I'm disappointed with 3 things in the 9950X3D2: the price, the name, and the fact they didn't put in 256 MB of L3 cache to make it actually double that of the 9950X3D
I have 6 5900xs in my rack and I feel like I got them at a bargain on ebay for 183 each used. They are power hogs. Compared to my Xeons or Epyc systems I do get more bang for buck out of them but they still lose in raw compute. I'd love a threadripper and dump a terabyte of ram in it. Maybe I'll be able to get my hands on one used at my local ewaste recycler eventually
themastergoose
Posts: 259 +846
Is all the discussion at the end about the Intel Core Ultra 7 270K Plus really necessary? It's hardly a competing product to the CPU at the centre of this review. Even more bizarre is the recommendation to declare that Intel processor the productivity champion. It is several steps down the results for productivity except for the sole Cinebench single threaded test. Even lost in several of them to the 14900K.
I was curious if there were any meaningful gains in any workload that could explain why this processor exists, so I read the Phoronix review. It looks like it is quite a bit faster than the 9950X3D in some HPC workloads, like computational fluid dynamics. It's not the kind of thing consumers or even most professionals are using, but at least there's some use case for it.
Shirley Dulcey
Posts: 139 +94
If you want top gaming performance with a side of good productivity performance, the 9950X3D is still there for you. $200 less, essentially identical in gaming, a small drop in productivity performance.
This new processor is for bragging rights. It's not cost-effective for anybody. Though if you end up spending $2,000 or more on RAM, as you might for an engineering or AI workstation, what's a couple of extra hundred for the CPU?
This new processor is for bragging rights. It's not cost-effective for anybody. Though if you end up spending $2,000 or more on RAM, as you might for an engineering or AI workstation, what's a couple of extra hundred for the CPU?
Neatfeatguy
Posts: 2,246 +4,537
Technically the 5900X did have a 3D cache version, it however, was never green lit for manufacturing. The 5900X3D was demonstrated and tested for that one long, bushy, unkempt hair youtuber tech guy (I don't know his name). He had a video of his trip to AMD and they got to see the 5900X3D get tested.
Shame it never came out, I would have picked one up and used it over my 5900X. Most likely the 3D version would have been slightly slower in multitasking, but a bit faster in some games with the 3D cache. I'd have been okay with that.
Is the 9950x3d2 in a market all by itself, or are other products sold that compete with it?
And did you miss they said "productivity VALUE"? Not just "productivity". That "value" is crucial context. The 9950x3d2 is triple the price but doesnt offer anywhere near triple the performance in productivity for that price.
I think the "calls" for the dual vcache CCD config were more a hangover from the prior gen with its various clock tradeoffs and annoying scheduler issues, that seem to be already fixed in Zen5
Sorileus86
Posts: 44 +15
Glad I picked up a 9850X3D instead of AMDs new chip, a literal 1% difference with the 9800X3D, and I have the 9850X3D so that might mean it has a lead in games over the 9950X3D2. Also its cheaper, less power draw, runs cooler. My 420mm Phantek's AOI will have no issue cooling my 9850X3D, but the 9950X3D2 at nearly pulling 300Watts is Insane, I know that's not normal but still lol.
I haven't finished building my new PC, but not getting the new chip allowed me to buy the ram I wanted....yah...the ram....@)%^&! lol
I haven't finished building my new PC, but not getting the new chip allowed me to buy the ram I wanted....yah...the ram....@)%^&! lol
CBTex
Posts: 202 +546
I always wanted to know what frame rates I would get at 1080p medium if I spent $900 on a cpu and $3000 on a gpu. /sarcasm
Can we get some actual real world resolution results?
Real story idea: Test what cpu becomes the bottleneck for each gpu at 1440 and 4k. This would help people know if their current cpu or gpu need upgraded first.
Can we get some actual real world resolution results?
Real story idea: Test what cpu becomes the bottleneck for each gpu at 1440 and 4k. This would help people know if their current cpu or gpu need upgraded first.
hahahanoobs
Posts: 5,784 +3,513
Paul Dougherty
Posts: 24 +12
I appreciate seeing the 265K getting included in the testing. I bought one a while ago (got the same G.skill RAM kit used in testing too) from MC, before the AIpocalypse, cheap.
Sometimes I think I should just pony up a few hundred dollars more the 9-series just to be able to see direct comparisons a year later when most reviewers have dropped the 7-series. Then I remember that's dumb and do something else dumb instead.
Sometimes I think I should just pony up a few hundred dollars more the 9-series just to be able to see direct comparisons a year later when most reviewers have dropped the 7-series. Then I remember that's dumb and do something else dumb instead.
daffy duck
Posts: 1,075 +989
Really really poor form from HUB testing like it's a normal consumer cpu. This is a poor man's HEDT replacement for TR.
If you want sensible testing look at Phoronix and and see what software actually thrives on cache. For example OpenFOAM fluid sim software. The people that buy these know exactly why they need more cache and could care less one bit about utterly useless tests on Cinebench, compression and gaming.
If you want sensible testing look at Phoronix and and see what software actually thrives on cache. For example OpenFOAM fluid sim software. The people that buy these know exactly why they need more cache and could care less one bit about utterly useless tests on Cinebench, compression and gaming.
I feel this is a niche chip that most people will not appreciate, especially when you consider the price. The extra cache is nice, but it is expected that it is not going to materially improve games and most productivity software. In latency critical software, the extra cache will help. What I do think that extra cache may be helpful is to keep a game or software perform more consistently over a longer period of time. Cache is limited, so I do think it may run out over time, so having more means it can sustain longer in this case?
Starfals
Posts: 334 +302
Try 5500 bucks. 3000? What is this? 2025?
Melkor Unlimited
Posts: 341 +316
It has exactly double amount of L3 (and L2 and L1) of 9800x3d/9850x3D
It has triple L3 (but the same L2 and L1) than 9950x.
