9800x 3d vs 12900k - Battle of the Century

[x4it3n]

Have you checked the effective clocks? aggressive curve might not make system unstable but the performance is actually lowered due to lower effective clocks.

I'll check when I have time but I'm pretty sure it's 5400MHz effective yes

Have you checked the effective clocks? aggressive curve might not make system unstable but the performance is actually lowered due to lower effective clocks.

I'll check when I have time but I'm pretty sure it's 5400MHz effective

You guys seen this new bios update?think its for asus only for now.Supposed to improve latency for gaming.

Is it the same as the Latency Killer on MSI ?

 

[wheresmycar]

So in summary...who won in this battle of the century: 9800X3D vs 12900K?

 

[ShrimpBrime]

So in summary...who won in this battle of the century: 9800X3D vs 12900K?

9800X3D 3 years before it even released.

 

[wheresmycar]

9800X3D 3 years before it even released.

I gotto be honest, in all my Intel-years, i didn't see it coming. Right up to Kaby Lake I wouldn't even give AMD a second glance other than the odd quick glimpse at benchmarks. And here I am in 2024 sitting on a 5800X3D.

 

[phanbuey]

I gotto be honest, in all my Intel-years, i didn't see it coming. Right up to Kaby Lake I wouldn't even give AMD a second glance other than the odd quick glimpse at benchmarks. And here I am in 2024 sitting on a 5800X3D.

I don't think they saw it coming either.

 

[JustBenching]

So in summary...who won in this battle of the century: 9800X3D vs 12900K?

The one im going to keep, which looks like is the 12900k

But the 9800x 3d is faster in 90% of games

 

[AVATARAT]

You guys seen this new bios update?think its for asus only for now.Supposed to improve latency for gaming.

*To add a option that return the latency to the level it was before it was broken by AMD.

 

[Cowboystrekk]

I gotto be honest, in all my Intel-years, i didn't see it coming. Right up to Kaby Lake I wouldn't even give AMD a second glance other than the odd quick glimpse at benchmarks. And here I am in 2024 sitting on a 5800X3D.

There has been a few milestones which made a very big difference in hardware the last years. Most noteably in my opinion is ssd-storage, g-sync/adaptive sync, dlss and 3D cache. They made huge improvements that surprised me at how much difference they made.

 

[kDude]

*To add a option that return the latency to the level it was before it was broken by AMD.

When did they break latency?

 

[HD64G]

Future proof has been thrown around for decades.

In these modern times, you will never build a future proof computer.

I built a 12 core monster in 2021, today an 8 core beats it in everything.

Future proof's meaning is just that what you buy today will be able to satisfy your needs in X years down the road, not expecting the tech progress with stall and the same component will remain competitive with the newest and greatest after those X years.

 

[AVATARAT]

When did they break latency?

Before 2 AGESA version the latency jump with up to 10ns.
(AGESA 1.2.0.2A add "Turbo Game Mode"/"Gaming Mode"/...)

 

[ShrimpBrime]

Before 2 AGESA version the latency jump with up to 10ns.
(AGESA 1.2.0.2A add "Turbo Game Mode"/"Gaming Mode"/...)

As if 10ns would be a big deal considering the amount of bandwidth to work with..

It feels sluggish?

That's merely 10 Billionths of 1 second. You (not you you, everyone you) wouldn't know the difference if you couldn't measure it with aida64, if that's even accurate measurement tooling....

 

[AVATARAT]

As if 10ns would be a big deal considering the amount of bandwidth to work with..

It feels sluggish?

That's merely 10 Billionths of 1 second. You (not you you, everyone you) wouldn't know the difference if you couldn't measure it with aida64, if that's even accurate measurement tooling....

Fast CPU <=> RAM communication, as you know, is very important.
My AIDA64 result jumped from ~56 to ~66ns, so:

FPS decreases.
Anything related to latency will run slower.

If you do a Google search, you'll read a lot about it

 

[ShrimpBrime]

Fast CPU <=> RAM communication, as you know, is very important.
My AIDA64 result jumped from ~56 to ~66ns, so:

FPS decreases.
Anything related to latency will run slower.

If you do a Google search, you'll read a lot about it

I bench it. Makes most difference if you compare to your default settings probably in access of 100ns latency.

But as far as frame rates go, 1% differences 56 to 66ns, margin of error.

I run a Cas of 38 and AIDA says 60.3ns latency. According to what I read, I should have really poor performance.

I also know this latency can be reduced by cache and cpu frequency as well.

But 10ns isn't a lot to worry about. Only for those that take the time to measure it.

 

[phanbuey]

I bench it. Makes most difference if you compare to your default settings probably in access of 100ns latency.

But as far as frame rates go, 1% differences 56 to 66ns, margin of error.

I run a Cas of 38 and AIDA says 60.3ns latency. According to what I read, I should have really poor performance.

I also know this latency can be reduced by cache and cpu frequency as well.

But 10ns isn't a lot to worry about. Only for those that take the time to measure it.

I've seen 81ns rigs run smoother with higher lows than 53ns rigs, on the same system lower ns in AIDA is typically better, but even then, not always. Comparing across systems is completely pointless IMO - a 33ns 10900k will get absolutely smoked by a 70ns-90ns 12900k/5700x3d+ in games.

Benching the games is really all that matters. Anyone who thinks they can tell a difference between 50ns and 100ns across different systems doesn't understand nanoseconds work. There are 1,000,000 nanoseconds in a millisecond. The NS latency on the same system typically correlates to faster memory settings, and faster memory settings = higher FPS, but comparing latency across different architectures as a gauge of gaming performance is a mistake.

 

[AVATARAT]

OK.
Everyone can take it how they want, I just said that the new AGESA won't improve anything, it will fix what it broke.

 

[JustBenching]

Fast CPU <=> RAM communication, as you know, is very important.
My AIDA64 result jumped from ~56 to ~66ns, so:

FPS decreases.
Anything related to latency will run slower.

If you do a Google search, you'll read a lot about it

Did you actually test it though? I think it's the Aida reading that's off, gaming performance is the same.

I bench it. Makes most difference if you compare to your default settings probably in access of 100ns latency.

But as far as frame rates go, 1% differences 56 to 66ns, margin of error.

I run a Cas of 38 and AIDA says 60.3ns latency. According to what I read, I should have really poor performance.

I also know this latency can be reduced by cache and cpu frequency as well.

But 10ns isn't a lot to worry about. Only for those that take the time to measure it.

That's not true either. Memory latency makes a huge difference. The thing is that in this particular instance with the x3d it's just the Aida reading that's off, not the memory latency itself.

The reason my 12900k is only 15 - 20% off from the 9800x 3d is because the 47ns of latency compared to 60+ you'll get with xmp.

 

[AVATARAT]

Did you actually test it though? I think it's the Aida reading that's off, gaming performance is the same.

Latency Killer reduces latency to match previous AGESA microcode updates

This feature was introduced to combat purported memory latency degradation that started with AGESA 1.2.0.2a on AM5 motherboards—1.2.0.2a was the microcode update that added support for Ryzen 9000X3D CPUs. An AIDA64 benchmark run comparing the latency booster to default operating reveals an 8ns improvement in latency when turned on.

 

[rv8000]

In regards to latency on the new bios and certain MSI/ASUS features you can read here.

Looks like AIDA doing AIDA things, results may vary game to game but apparently the 1.2.0.2b based bios and additional features may on average run better in games despite showing worse latency in AIDA

 

[phanbuey]

Did you actually test it though? I think it's the Aida reading that's off, gaming performance is the same.


That's not true either. Memory latency makes a huge difference. The thing is that in this particular instance with the x3d it's just the Aida reading that's off, not the memory latency itself.

The reason my 12900k is only 15 - 20% off from the 9800x 3d is because the 47ns of latency compared to 60+ you'll get with xmp.

You're comparing a tuned 12900k that took you years to tune to a stock XMP on the 9800X3D that you spent a week fiddling with (on Beta bioses) - and still it's still losing by 15-20%. Clearly memory latency isn't the deciding factor.

In regards to latency on the new bios and certain MSI/ASUS features can read here.

Looks like AIDA doing AIDA things, results may vary game to game hut apparently the 1.2.0.2b based bios and additional features may on average run better in games despite showing worse latency in AIDA

This!

 

[JustBenching]

You're comparing a tuned 12900k that took you years to tune to a stock XMP on the 9800X3D that you spent a week fiddling with (on Beta bioses) - and still it's still losing by 15-20%. Clearly memory latency isn't the deciding factor.



This!

Both are megatuned though, im running 6400c28 on the 9800x 3d - 63ns of latency before the new agesa.

 

[AVATARAT]

In regards to latency on the new bios and certain MSI/ASUS features you can read here.

Looks like AIDA doing AIDA things, results may vary game to game but apparently the 1.2.0.2b based bios and additional features may on average run better in games despite showing worse latency in AIDA

Of course the result will vary, some games are memory dependent and others not so much.

This!

You can always try Shadow of the Tomb Rider because it scales quite well with RAM tuning.
One of the x3D things to be a beast there.

 

[phanbuey]

Both are megatuned though, im running 6400c28 on the 9800x 3d - 63ns of latency before the new agesa.

Yeah but you're on immature bioses; the 12900K took like a year to get the latency down when it launched since even 5600 ram was impossible to get a hold of - the average latency people were getting was 65-75ns, the first boards/versions with lowest latency were DDR4 and the initial MSI bioses didn't hit anywhere near 47ns -- it took a while and new bioses and dimms to get there. My 7600mhz kit didn't get a proper bios on until like June of 2023 that let me break below 55ns.

And AIDA is self-admittedly buggy -- they will likely release new AIDA version and you will 'miraculously' be sitting at 52ns on the AM5 system with no change in fps in games.

 

[AVATARAT]

And AIDA is self-admittedly buggy -- they will likely release new AIDA version and you will 'miraculously' be sitting at 52ns on the AM5 system with no change in fps in games.

There is a problem and this time it is not in AIDA, but additionally we have a fix in BIOS for that.

 

[rv8000]

There is a problem and this time it is not in AIDA, but additionally we have a fix in BIOS for that.

If AIDA is the outlier… it IS the problem.