RTX 4090 & 53 Games: Ryzen 7 5800X vs Ryzen 7 5800X3D

[zlobby]

I can only imagine people are living in poorly insulated massive houses in cold climates or something

People living in the conditions you are describing can't usually afford a 3090.

 

[Mussels]

People living in the conditions you are describing can't usually afford a 3090.

I'm a disabled single dad and i could afford one.

 

[Gica]

With a much more expensive X3D, it should be taken into account that it is weaker than the 1800X in applications that do not react intensively to the cache memory (3.8/4.7 GHz versus 3.4/4.5 GHz).

 

[oxrufiioxo]

This is pretty much why none of the new gen cpus are all that appealing for gaming.... Yeah they are probably slightly ahead of this when tweaked to the max but this chip will perform well with crap ddr4 and a sub 150 usd board possibly even a sub 100 usd board while being easy to cool.

 

[The King]

Dual rank sams b-die on 5800x3d vs "questionable" ram on 5800x.
Would really like to see the opposite, ie 5800x with good b-die vs 5800x3d with trash ram.

@W1zzard if possible, can you also add 0.1% lows to the graph?

I completely missed that!

If those are 8GBX2=16Gb then the 5800X is at a big disadvantage vs the 5800X3D 16GBx2 dual rank setup.

Another possible issue is that many on ZEN 3 see some performance loss running 2000 FCLK and many people don't even run that.

I would have also preferred to see the same 2X16GB 3600 CL14-14-14 1T on the 5800X vs that config that was run in this test.

Don't expect the outcome to change but the gap may not be has big then. Also 4000 CL20 RCD 23

 

[Space Lynx]

I'm less interested in the ram testing stuff and more interested in weird stuff like how much more wattage does a 13600k draw vs a 13600kf, even when integrated graphics are not being used? Is it the same, of 74 watts when gaming, or is it truly fully dormant, or does the igpu sip a few watts simply because of its proximity on a monolith design?

I understand it doesn't really matter though, so I wouldn't ask anyone to test this for me lmao

 

[zlobby]

I'm a disabled single dad and i could afford one.

Hence the 'usually' part in my previous post.

 

[gffermari]

**HU used a 3090Ti, not a 4090.**
Excellent job!
I wanted to see a 1% lows chart using the 4090 but ok.
We can imagine that the difference would be massive.

 

[Richards]

@W1zzard we need a 13900k vs 7950x oc'd on a 4090 and 7900 xt when you get it

 

[Mussels]

@W1zzard we need a 13900k vs 7950x oc'd on a 4090 and 7900 xt when you get it

I get the feeling he's going to do a few of these comparisons, and when he has enough done he'll be able to use the results in the new reviews from then on


When you first start with a new review setup, you can't compare it to the old results

 

[Nopa]

View attachment 268093
View attachment 268094

**HU used a 3090Ti, not a 4090.**
Excellent job!
I wanted to see a 1% lows chart using the 4090 but ok.
We can imagine that the difference would be massive.
View attachment 268095

Really intriguing to see how many % the next X3D's gonna demolish 13600K & 13700K. I wager anywhere from 15-30% faster.

 

[W1zzard]

Days Gone

2021​

DX12

UE4

Number #47514 when I ask you to stop including Days Gone under DX12. This meta joke has to STAHP.

Lol i cant believe i failed this again… fixed now and hopefully for future articles, too

 

[Vayra86]

Impressive.

That X3D is a monster CPU. I need one. I don't need one. But I do. Shit

 

[ratirt]

With a much more expensive X3D, it should be taken into account that it is weaker than the 1800X in applications that do not react intensively to the cache memory (3.8/4.7 GHz versus 3.4/4.5 GHz).

You have no idea what you are talking about. 1800x is much slower than any 5000 series processor. Here you are comparing 8 cores to 8 cores. 5800x3d even with slightly slower clocks than a 5800x, eats the 1800x whole. The 3000 series CPUs where miles ahead of the 1st Ryzen.

 

[kapone32]

With a much more expensive X3D, it should be taken into account that it is weaker than the 1800X in applications that do not react intensively to the cache memory (3.8/4.7 GHz versus 3.4/4.5 GHz).

What are you talking about? The IPC improvement alone between 1st Gen and 4th Gen belies what you are saying. Then let us realistically look at clock speeds. If you can get a 1800X to 4.7 GHZ (On air) you have the ultimate unicorn. The 5800X is faster in most applications than the 5800X3D but if you are coming from anything before that outside of the 3900/3950X the 5800X3D is a common sense upgrade for anyone with a previous Gen AM4 chip that wants to Game. Put another way the 5800X3D is like the next compelling CPU that the 3300X was. I can also confirm once again that the 5800X3D uses less power than the 5800/5900/5950X and blows them away in 1% lows and FPS in some Games.

 

[HTC]

If you can get a 1800X to 4.7 GHZ

I'm pretty sure dude meant a 5800X but wrote 1800X instead: his statement makes absolutely NO SENSE otherwise.

 

[HD64G]

@W1zzard many kudos for your great job! This 3D-cache will be standard on many future CPU gens from AMD since not also make fast RAM less important but also it stop the CPUs being the bottleneck of many unoptimised game engines out there. Adding the success 5800X3D has in the market atm while both Intel and AMD released the latest CPU series, it will become the normality for the desktop (mostly) sooner that most expected.

 

[Deleted member 185088]

All games are tested in custom bench scenes as the integrated benchmarks often paint a completely inaccurate picture compared to actual gameplay. Also, the GPU vendors actively optimize their drivers to achieve good results in integrated benchmarks

Apparently even replays are not reliable according to Ian (former anandtech) part of the game engine doesn't work like physics.

 

[Nopa]

@W1zzard many kudos for your great job! This 3D-cache will be standard on many future CPU gens from AMD since not also make fast RAM less important but also it stop the CPUs being the bottleneck of many unoptimised game engines out there. Adding the success 5800X3D has in the market atm while both Intel and AMD released the latest CPU series, it will become the normality for the desktop (mostly) sooner that most expected.

5800X3D's probably the best AMD CPU since Athlon 64 Era tbh. I imagine the profits from it must be so good to the point we'll see X3D being featured in every generations from now on.

 

[gffermari]

@W1zzard many kudos for your great job! This 3D-cache will be standard on many future CPU gens from AMD since not also make fast RAM less important but also it stop the CPUs being the bottleneck of many unoptimised game engines out there. Adding the success 5800X3D has in the market atm while both Intel and AMD released the latest CPU series, it will become the normality for the desktop (mostly) sooner that most expected.

And that's the problem with the 3D versions. They make the normal SKUs.....somewhat not worthy.
Which shouldn't be the case.

 

[ODOGG26]

At this point going forward, AMD should release X3D of all 4 chips first then drop non X3D versions later. Make X3D their main sku's then drop X versions later with less wattage and price and if they want they can go a step further and drop non X versions with even less wattage and price. Maybe those can go in productivity laptops.

 

[theglaze]

Thanks for the benchmark analysis!

It would be very interesting to learn more about WHY the 3D V-Cache provides a massive gain in some titles (even at 4K, there are 10 titles that still show a +15% FPS improvement).

From all the tech talk about 3D V-Cache, I have decoded in layman's terms that the 5800X3D accelerates when:

1. Workloads cannot fit in a 'normal' sized cache
2. Workloads are extremely CPU limited
3. Workloads are memory sensitive
4. Workloads reuse data and calculations

It would take some investigative journalism to reach out to the game developers, but I'm curious to learn how those 10 titles (and others) align into different categories, such as those 4 above.

AMD has clear intentions to move forward with 3D V-Cache on AM5, and the package will (hopefully) be improved upon. But it's still an assumption to think the same game titles will benefit (in the same way) from future X3D CPUs.

 

[Xuper]

can we have RUST in bench?

 

[Vayra86]

Thanks for the benchmark analysis!

It would be very interesting to learn more about WHY the 3D V-Cache provides a massive gain in some titles (even at 4K, there are 10 titles that still show a +15% FPS improvement).

From all the tech talk about 3D V-Cache, I have decoded in layman's terms that the 5800X3D accelerates when:

1. Workloads cannot fit in a 'normal' sized cache
2. Workloads are extremely CPU limited
3. Workloads are memory sensitive
4. Workloads reuse data and calculations

It would take some investigative journalism to reach out to the game developers, but I'm curious to learn how those 10 titles (and others) align into different categories, such as those 4 above.

AMD has clear intentions to move forward with 3D V-Cache on AM5, and the package will (hopefully) be improved upon. But it's still an assumption to think the same game titles will benefit (in the same way) from future X3D CPUs.

Its not a baseless assumption.

Bigger cache means you can keep more information on hand to be (re-) used faster. Larger cache simply works like lubricant in a realtime application; the rest of the gears in the processing pipeline run more smoothly, or 'barely have to wait' to process the next bit of data required to produce the next frame. Put differently, because gaming is real time, what you want is the fastest possible throughput of code, you're not really working on a massive problem, but thousands of tiny ones that each provide input for your next frame in the game. Cache helps that throughput. This explains 1,2, 3 and 4, pretty much.

CPU limited is a broad term - you can be short on threads in applications that require parallelism. If an application requires (or uses) peak overall 5800X3D performance, extra cache won't help it much; if it requires 16 physical cores for optimal performance, cache won't get it closer to that either. With gaming, the most influential though is latency to provide the next bit of information - cache reduces latency. And simply put, the slowest link in a pipeline can single handedly limit your FPS even if everything else is super fast. The less you need to move data around, or the shorter the distance (actual distance, yes), the more efficient all parts in the pipeline get to work, as they keep getting 'fed'. Cache is ON the CPU - and has several Levels even to show how close (L1, L2, L3). RAM is on the board. The GPU is a whole lot further down the board, furthest away from the CPU. If data takes longer to get transported, or if there is congestion somewhere, you lose frames. Cache makes the flow of traffic less congested.

 

[SL2]

I'm pretty sure dude meant a 5800X but wrote 1800X instead: his statement makes absolutely NO SENSE otherwise.

Yeah everyone got so emotional about the 1800X statement that they missed the mismatching clock speeds in the same post..

Anyway, Gica is right, but a 4 % clock speed difference isn't much, and the price difference is smaller these days.