AMD Ryzen 7 5800X3D

[Deleted member 24505]

Yea I have to agree with this. This was suspect to me when the hype train came hard about 6 months ago. The performance over the 5800 was overall not earth shattering in gaming while it got gimped doing production. The 5800X got nailed by the 5600 during launch as it was far cheaper to go the 5600 in gaming than buying a 5800X The 5600 sold extremely well so I can see AMD making this a gimmick CPU.

More and more I believe that this was done to remove flawed product that did not pass the grade as well as overstock they have on hand.

For myself it is not worth the price that they are asking for. Save your money and look at other options.

It was good enough to get all the AMD fans slavering, and jabbering again at how crappy ADL is.

 

[SL2]

I can't wait for Raphael to show up, and we hopefully can leave the apples-to-oranges choice of RAM behind (for now).

 

[Dr. Dro]

The only thing the 12700K crucifies is power consumption, it draws significantly more power than the 5800X3D multithreaded:

And its still slower in gaming. This is not a good place for Intel to be; slower in gaming while drawing significantly more power!

I mean, my 5950X also doesn't draw more than 170W or so either. My motherboard won't let it. I'm not entirely sure how this validates your argument - 5800X3D at 166W 8/16 is actually barely half of the 5950Xs efficiency at 179W, if you want to nitpick that way. Which means that the 12900KS remains more efficient if you do the whole wildly inaccurate math thing by simply doing cores*2 math (remember that ADL has both high performance and efficient cores, which have differing levels of power consumption). Apples to oranges, imho.

This chip has excellent perf/watt in games. It seems to be more efficient than anything AMD or Intel currently have on offer:

Source

It's running at the lowest voltage and clock speeds out of the bunch, so it's no wonder, really. I'm not entirely sure how this is being spun as an achievement, given all power-efficient Ryzen SKUs have been doing exactly that thus far - it's less about it being insanely efficient, and more about its default settings not bursting to insanely high clocks/frequencies like the regular SKUs do. Think of high-performance mobile parts (H series) or the desktop GE series, they both run a more conservative v/f curve and have a strict PPT setting to lower consumption. Since this processor is locked, WYSIWYG.

It's a great chip, but I fear it's being praised for something it hasn't earned. It's just how I feel about it, though, I will say that it is an excellent processor - I just think it costs too much for what it offers.

 

[Deleted member 24505]

Oh and we will lock it so you can't OC it and break it as it is so efficient. We made it and don't know how to fully manage it so we had to knobble it so stick it in and run it as it is, which considering it is a enthusiast CPU you will really appreciate.

 

[Lifeless222]

Still wait game version of AL - Raptor Lake with 3D Game Cache.

 

[Mussels]

The price of Intel CPU's are built into the chipset itself. If you look at motherboard prices, they are pretty insane. A clever trick of marketing. Also, a good amount of games have a pretty large disparity. It makes really wonder about a Zen 4 3D stacker in the works, because I would be all over that. Imagine a 20% IPC gain, and additional 15% game performance gain with a cache edition CPU, marketed as Gamer Edition (kinda like Black Edition), where people buying those CPU's don't care about blender. I was really hoping someone would test this on starcraft 2 since it's single thread limited, was curious if cache would help boost the performance.

As much as i'd love SC2 benchmarks, it's pretty much impossible to create a reproducible test, and in MP the other players CPU's slowing down forces yours to slown down too

And for those extra 55 watts, it scores 7544 points higher.

I know thinking hurts, but no pain, no gain.

because it has more cores. See what those 55 watts does for the higher core count AMD chips.
This argument is just moving the goalposts

You are putting a graph of power consumption in cinebench and then talk about gaming. Are you daft dude? You realize according to the graph you just posted the 12700 is more efficient than the 5800x 3d,right? Lol, way to shoot your self on the foot buddy

w1zzard posted his own graph:

and another source quoted above for emphasis:

The higher core count intels are faster in multi threaded apps.
So are the higher core count ryzens.
Duh.


The intel's are higher power consumption in *everything*
That ups the cost of the entire PC, and it's relevant as heck. Intel selling the CPU's cheap is a marketing tactic, when the boards are so expensive.
(Dont forget all the scandals where the cheaper motherboards couldnt handle the CPU's at stock and performance tanked)

The moment you limit the intels wattages as some people keep saying, you lose that performance.
Remember that many sources for the intel systems flip between three states: Intels limits, motherboard limits, and static overclocks.
You cant cherry pick the performance of the CPU with unlimited power, while quoting the wattages and heat of the power limited states.

For that max performance, you need a TDP unlimited CPU, high end cooling, high end motherboard, high end RAM. Skip just one, and you get worse performance than the Ryzens.

The actual cost of a PC changes drastically with all that in mind, even if the motherboards and CPU's were equally priced - because the intel systems need a larger PSU, larger cooling, more case fans, and often a larger case to fit all that in.



The only valid use case for the 12900K/S is for someone who needs top tier gaming performance (165Hz and above) *and* workstation needs at the same time.
Sub 120Hz Gamers can get a 5600x or 12400F and be set for another 5 years.



Why do i care about these distinctions so much?
Because as an end user buying these products from these companies for the last 30 years, i've been burned too many times by hidden catches and gotchas.
It's like trusting a marketing slide
100% faster than last years product!***
*** (If you use DDR4 2667 for last years product but DDR5 6400 in this years, with a 1KW PSU and 360mm AIO with a $500 motherboard with 24phase watercooled VRM's)

 

[Deleted member 202104]

because it has more cores. See what those 55 watts does for the higher core count AMD chips.
This argument is just moving the goalposts

Of course it has more cores. This isn't anything against the x3d so don't get all defensive. The post I was replying to showed a chart with power consumption and the moron was claiming that the 12700k had worse performance while using more power. In that graph, the 12700k did use more power, but it was 50% faster.

 

[Mussels]

I edited my post above and turned it into quite the essay
The 12700k uses more power for more MT performance, absolutely.
It's also not the power whore the 12900K/S is, IMO it's the better chip.

It's just about comparing like for like, because in this case the 5800x3d is the gaming performance of a 12900KS, but it is NOT competing or intended to compete on the multi threaded count.

Imagine a 5600x3D and how people would whinge it's no good, because of a low cinebench score??

 

[JustBenching]

On 12900K/KS DDR5 has worst power efficiency than DD4

View attachment 243627

As much as i'd love SC2 benchmarks, it's pretty much impossible to create a reproducible test, and in MP the other players CPU's slowing down forces yours to slown down too


because it has more cores. See what those 55 watts does for the higher core count AMD chips.
This argument is just moving the goalposts


w1zzard posted his own graph:

View attachment 243654
and another source quoted above for emphasis:
View attachment 243655



The higher core count intels are faster in multi threaded apps.
So are the higher core count ryzens.
Duh.


The intel's are higher power consumption in *everything*
That ups the cost of the entire PC, and it's relevant as heck. Intel selling the CPU's cheap is a marketing tactic, when the boards are so expensive.
(Dont forget all the scandals where the cheaper motherboards couldnt handle the CPU's at stock and performance tanked)

The moment you limit the intels wattages as some people keep saying, you lose that performance.
Remember that many sources for the intel systems flip between three states: Intels limits, motherboard limits, and static overclocks.
You cant cherry pick the performance of the CPU with unlimited power, while quoting the wattages and heat of the power limited states.

For that max performance, you need a TDP unlimited CPU, high end cooling, high end motherboard, high end RAM. Skip just one, and you get worse performance than the Ryzens.

The actual cost of a PC changes drastically with all that in mind, even if the motherboards and CPU's were equally priced - because the intel systems need a larger PSU, larger cooling, more case fans, and often a larger case to fit all that in.



The only valid use case for the 12900K/S is for someone who needs top tier gaming performance (165Hz and above) *and* workstation needs at the same time.
Sub 120Hz Gamers can get a 5600x or 12400F and be set for another 5 years.



Why do i care about these distinctions so much?
Because as an end user buying these products from these companies for the last 30 years, i've been burned too many times by hidden catches and gotchas.
It's like trusting a marketing slide
100% faster than last years product!***
*** (If you use DDR4 2667 for last years product but DDR5 6400 in this years, with a 1KW PSU and 360mm AIO with a $500 motherboard with 24phase watercooled VRM's)

Can we stop with that high end bullshit argument? It gets really boring. Techspot tested a 12700f on an el cheapo b660 with the stock cooler and no power limits. Guess what, it performs identical in games to the 12700k on a 360 aio.

12700f with a 150 euro b660 is cheaper than the 3d and crucifies it in every other workload other than 240p gaming. Period. The rest is just amd fanboying defense.

 

[Mussels]

Can we stop with that high end bullshit argument? It gets really boring. Techspot tested a 12700f on an el cheapo b660 with the stock cooler and no power limits. Guess what, it performs identical in games to the 12700k on a 360 aio.

12700f with a 150 euro b660 is cheaper than the 3d and crucifies it in every other workload other than 240p gaming. Period. The rest is just amd fanboying defense.

Except it's not bullshit. It's rare that a cheaper board has no performance penalty.
A 12700F is not a 12900KS.
Did they test that budget board with budget ram, vs a top tier board with top tier ram, and a high end GPU the same way they do their normal reviews? Did they test it in a normal case, or an open air test bench? Did they still use a bigass expensive cooler on it? cheap JEDEC RAM?

It's very easy to skew results and say "oh only 2% different, they're the same!" except 2% is the entire leaderboard in 4K testing.

 

[gffermari]

The man who buys a 5800X3D does not care if the 12700K is faster in MT tasks.
If he does care, then he shouldn't buy it.
If also he doesn't mind the small difference in games, he could buy any other 5000 cpu or cheap 12000 cpu.
It's just not for them.

It may not make much sense to buy the 3D for general usage, as it doen't make sense to buy the KS.

 

[usul1978]

I was willing to spend the stupid money to get that chip on my gaming AM4 plateform to replace my trusty 5600x just for (pseudo) future proofing, get the bragging rights and because my birthday is 5 days after launch day.

I only game, in 1440p with a 3080 and I aim ingame at 110 fps (if I get more FPS, I up the rez and use downscaling, like in Warzone for instance where I play at 130%) so I'm basically always gpu bottlenecked. I have not much to gain in real life if I get a 5800x3d.

But what bother me, is the things I'll probably lose with the 5800x3D : I'll probably lose my sweet RAM settings as apparently it's not sure the 5800x3d will take 1900 mhz FCLK and makes my DR 3800 cl14 DDR4 fly. I'll lose my low temps and low consumption. I'll lose my (slight) OC on the 5600x...And maybe I'll lose some stability as the 5800x3D looks like something pushed to its limits.

So, after weeks of envy, I'll pass on the 5800x3d, and it makes me sad.

 

[JustBenching]

Except it's not bullshit. It's rare that a cheaper board has no performance penalty.
A 12700F is not a 12900KS.
Did they test that budget board with budget ram, vs a top tier board with top tier ram, and a high end GPU the same way they do their normal reviews? Did they test it in a normal case, or an open air test bench? Did they still use a bigass expensive cooler on it? cheap JEDEC RAM?

It's very easy to skew results and say "oh only 2% different, they're the same!" except 2% is the entire leaderboard in 4K testing.

Well let's agree on something. You care about the fastest, you get a 12900k / ks on a 2 dim mobo some hynix 7000c30 kits. That will cost you 1500,maybe more, but it kisses eveything goodbye...

Now regarding the 12700f, even the cheapest of cheap b660 will handle it just fine for gaming. Of course im not suggesting to buy the cheapest, cause the computer is not a games console, im just stating that it is possible. A 150 euro b660 bazooka maxes it out easily in cbr23 with no power limits, coupled with the 12700f thats 470 euros. And comes with a cooler, which again for gaming is fine (tested by techspot). Of course id suggest to change the cooler as well but still the fact is that for almost the price of the 3d alone you get a cpu a mobo and a cooler that absolutely nails the 3d in everything but 240p games, and even there the difference is what, 5%?

The man who buys a 5800X3D does not care if the 12700K is faster in MT tasks.
If he does care, then he shouldn't buy it.
If also he doesn't mind the small difference in games, he could buy any other 5000 cpu or cheap 12000 cpu.
It's just not for them.

It may not make much sense to buy the 3D for general usage, as it doen't make sense to buy the KS.

Then that man would be buying a 12900 with high end ddr5, no?

 

[gffermari]

@usul1978

yes, but that's by far the best and, most importantly, the last gaming cpu on AM4.

The only reason I'm thinking of pulling the trigger on 5800X3D against the 5900X/5950X (which makes much more sense buying in general) is that when I replace my 2080Ti with a 4080/4080Ti, the difference in games will be even bigger than it is now among the 5000 cpus.

Then that man would be buying a 12900 with high end ddr5, no?

No if he wants to spend half the price of a cpu or maybe less than half of a system.
(although we agree that it's pointless to invest on AM4 now but in theory a ''mad'' man who wants the absolute best and cheapest possible, then the 3D is for him)

 

[JustBenching]

he wants to spend half the price of a cpu or maybe less than half of a system.
(although we agree that it's pointless to invest on AM4 now but in theory a ''mad'' man who wants the absolute best and cheapest possible, then the 3D is for him)

I beg to differ. The best and cheapest is the 12700f. Everything above that is within 5% pretty much. The only reason I got the 12900k over the 12700 is because i wanted better binned chip for ocing

 

[W1zzard]

https://www.techpowerup.com/forums/attachments/1649970428652-png.243621/

Impressive results because graphs dont start at zero.
(Not trying to shit on the OC achievement, Pieter is a master at overclocking and has done amazing work)

 

[Deleted member 24505]

Why would people who spend $1k+ on just their GPU quibble on the price of any part for their PC, either you can afford it or not. I understand the power thing kind of, but i don't give a crap what power any part uses as long as i can cool it. If you have/want a high end rig, it should be high end. You don't buy a Ferrari and switch the engine for a 1000cc 4cyl 60mpg.

 

[InVasMani]

Are we going to see this same end of life platform support banter the other way with Alder Lake's successor from Intel!!? Everyone considering buying AM4 chip now already knows this and anyone upgrading already on AM4 doesn't care at the same time. We don't even know what to expect with Alder Lake's successor because it's f*cking Intel we're talking about here remember Kaby Lake that was literally just Skylake with another 100MHz thrown on top and the IMC slightly ironed out with a tiny pittance of additional L3 cache. Is that what we should expect!?

Who knows because Intel the same company that often locks down overclocking of their chips and sure AMD does sometimes as well, but usually only for good reason not because hey I want to sell it as a unlocked chip for more money while I lock down everything below it to be less appealing. Hell Skylake has less L3 cache than LGA 775 C2Q like honestly years of node advancements and here you go this is your scrap pieces of L3 cache enjoy it. They also tried to sell you on the chips "security" features initially that then turned out to be one of the most vulnerable chips in recent history. Intel pulled the rug out on everyone with their fantasy chip "security" superiority over the competition. Look at all these unsecure IPC gains of 1-2% year over year and +0 core count. Sorry, but Intel didn't give a f*ck for a decade I have a decade of not giving a f*ck in turn to give back to them.

 

[BHS1975]

I was willing to spend the stupid money to get that chip on my gaming AM4 plateform to replace my trusty 5600x just for (pseudo) future proofing, get the bragging rights and because my birthday is 5 days after launch day.

I only game, in 1440p with a 3080 and I aim ingame at 110 fps (if I get more FPS, I up the rez and use downscaling, like in Warzone for instance where I play at 130%) so I'm basically always gpu bottlenecked. I have not much to gain in real life if I get a 5800x3d.

But what bother me, is the things I'll probably lose with the 5800x3D : I'll probably lose my sweet RAM settings as apparently it's not sure the 5800x3d will take 1900 mhz FCLK and makes my DR 3800 cl14 DDR4 fly. I'll lose my low temps and low consumption. I'll lose my (slight) OC on the 5600x...And maybe I'll lose some stability as the 5800x3D looks like something pushed to its limits.

So, after weeks of envy, I'll pass on the 5800x3d, and it makes me sad.

I have a 5600x too slightly OC to 4.7 and the 5800x3d would be more efficient as mine uses about 70w gaming while the 5800x3d uses 60w so it should run about the same temps. And what ram sticks do you have? I have the trident z 32gb 3600 cl16 running at 3800 cl15 flat 2T at 1.47v as they won't run at cl14 unless I use a much higher voltage.

 

[usul1978]

I have a 5600x too slightly OC to 4.7 and the 5800x3d would be more efficient as mine uses about 70w gaming while the 5800x3d uses 60w so it should run about the same temps. And what ram sticks do you have? I have the trident z 32gb 3600 cl16 running at 3800 cl15 flat 2T at 1.47v as they won't run at cl14 unless I use a much higher voltage.

I have Trident Neo 3800mhz cl14 2x16 gb kit (F4-3800C14D-32GTZN), @3800, CL14-16-16, 1T. That was the only B Die in dual rank I found. They suck 1.5V!

Actually my 5600x OC is really small and I guess that a slightly lower FCLK wouldn't be a real problem...so who knows, I may get one anyway : )

I would like to try a 5800x3D, and return it if it doesn't take 1900 FCLK, but that seems a bit complicated.

 

[SL2]

I'm actually surprised that AMD is as popular as it is right now, with Ryzen being top 5 on Amazon best sellers, and 8 of them being in top 10 as well.
I would have expected more like a 50-50 situation. Oh and only one model is from 2022, so not much of an impact from those models yet, although they're still pretty fresh.
I guess having fewer models helps AMD here.

Only 4 Alder in top 20, I thought Alder was so good that it would do better than that. Then there are other stores, of course..

 

[LupintheIII]

Well written summary and appreciate the detailed benchmarks. I was expecting more than WinRAR to really appreciate the cache but this is still quite interesting.

I'm really curious if Microsoft Flight Sim will benefit from the cache..

It does

 

[BHS1975]

I have Trident Neo 3800mhz cl14 2x16 gb kit (F4-3800C14D-32GTZN), @3800, CL14-16-16, 1T. That was the only B Die in dual rank I found. They suck 1.5V!

Actually my 5600x OC is really small and I guess that a slightly lower FCLK wouldn't be a real problem...so who knows, I may get one anyway : )

I would like to try a 5800x3D, and return it if it doesn't take 1900 FCLK, but that seems a bit complicated.

Get it through Amazon prime do you can return easily.

 

[usul1978]

Get it through Amazon prime do you can return easily.

Even if the box is opened?

 

[lexluthermiester]

Impressive results because graphs dont start at zero.
(Not trying to shit on the OC achievement, Pieter is a master at overclocking and has done amazing work)

To me, graphs like that boarder on the dishonest as they don't show the actual differences, even if they show the correct numbers. The fact that you folks don't do that is one of the many reasons TPU is so respected and people are willing to put a great deal of stock in the benchmarking results and reviews you folks offer.