MSI OCLab Reveals Ryzen 9000X3D 11-13% Faster Than 7000X3D, AMD Set to Dominate "Arrow Lake" in Gaming

[AnarchoPrimitiv]

Dominate? Won't it be within like 10%?
Not a big deal...

For a company with literally a third of the R&D budget....it should considered domination. Intel has LITERALLY every measurable advantage over AMD, so the fact that AMD is beating them in spite of that is impressive to say the least.

 

[Makaveli]

going to need more than 3 games.

And wukong is gpu bottlenecked so also not a good choice.

Zen5%, a total skip for any Zen4 or 5800X3D owner.

not skipping at all.

I'm going 5800X3D to 9800X3D that will be like a 30% gain in performance since the 7800X3D is on average 18% faster than a 5800X3D.

The ST performance improvement alone with be worth it as I do more than just play games on my rig.

 

[Dr. Dro]

I genuinely wonder if Zen 5's lukewarm performance improvements have something to do with the socket AM5 platform design in general. Guess we'll have to wait and see; so far both Arrow Lake and Zen 5 don't seem to bring pretty much anything noteworthy to owners of Raptor Lake and Zen 4 platforms, and in many cases, stretching as far back as the 5800X3D. It's still a great chip, it's just not, well, revolutionary I guess. Kind of like Ivy Bridge and Haswell were to people who had Sandy Bridge chips at the time, I guess?

 

[Makaveli]

I genuinely wonder if Zen 5's lukewarm performance improvements have something to do with the socket AM5 platform design in general. Guess we'll have to wait and see; so far both Arrow Lake and Zen 5 don't seem to bring pretty much anything noteworthy to owners of Raptor Lake and Zen 4 platforms, and in many cases, stretching as far back as the 5800X3D. It's still a great chip, it's just not, well, revolutionary I guess. Kind of like Ivy Bridge and Haswell were to people who had Sandy Bridge chips at the time, I guess?

I think the IF speeds are the problem need to push faster DDR5 at 1:1

but it doesn't look like that will be addressed until Zen 6.

 

[AnarchoPrimitiv]

I genuinely wonder if Zen 5's lukewarm performance improvements have something to do with the socket AM5 platform design in general. Guess we'll have to wait and see; so far both Arrow Lake and Zen 5 don't seem to bring pretty much anything noteworthy to owners of Raptor Lake and Zen 4 platforms, and in many cases, stretching as far back as the 5800X3D. It's still a great chip, it's just not, well, revolutionary I guess. Kind of like Ivy Bridge and Haswell were to people who had Sandy Bridge chips at the time, I guess?

Go look at Phoronix's review of the Zen5 consumer chips and then of the Zen5 Epyc chips....you'll see in professional tasks they had huge performance improvements. Perhaps gamers are unaware, but the enterprise x86 market dwarfs the gaming market in total addressable sales and revenue. Zen5 did have good performance gains where it is most lucrative for AMD.

 

[Frick]

How disappointing... This means that the vast majority of this almost meaningless perf increase is coming from clock speed enhancements only. The Zen 5 architecture really is an utter disappointment. Something went very wrong inside AMD. The exact same monumental f-up as what caused the awful RDNA 3 chip which has all but destroyed the Radeon brand. It's almost as if nobody inside AMD has ever actually seen these chips in a real working system and measured their performance - It's like they are only looking at simulation results.

How this miniscule kind of performance increase is going to generate upgrade sales is beyond me. Hey look, you can have maybe 5-10% more frames than your old and awful 7800X3D for the low low price of $500+!!! Grab 'em while they're hot!

One thing is for sure, this Zen 5 is not a fixable situation, it's almost certainly either an errata in the design, or a deliberate lie by AMD marketing.

Somebody needs firing.

There is some interesting stuff to Zen5 though (see below). I'm intensely interested in seeing how the Zen5 X3D stuff performs in games that like single core performance (meaning simulation style games).

 

[R0H1T]

Yes & no, look at the latest reviews & their scaling's tanked(?) for some reason ~

Go look at Phoronix's review of the Zen5 consumer chips and then of the Zen5 Epyc chips....you'll see in professional tasks they had huge performance improvements. Perhaps gamers are unaware, but the enterprise x86 market dwarfs the gaming market in total addressable sales and revenue. Zen5 did have good performance gains where it is most lucrative for AMD.

The power consumption's not too bad though!

 

[Darmok N Jalad]

I do hibernate/hybrid sleep most of the times! Only plebs keep their systems on 24/7 at idle

Yeah, I use sleep/suspend mode. The only time that’s a problem is when Windows randomly decides to wake up and never go back to sleep again.

 

[phanbuey]

Check out the other slides in HardwareLuxx's image gallery. They also include comparisons between Core Ultra 200K CPUs and Ryzen 9000s.

That said, AMD will basically have to slash prices for the lower non-X3D parts, unless they got the balls to release the 9800X3D for $550+.

edit: looks like they removed the slides now

They will release x3d at $500+.

They have no competition from ARL.

 

[Daven]

Intel will never win till they drop those E cores and make a pure 16 core P core CPU. 8 core vs 16 core do the math. I guess they do not care anymore they make enough on the business side with xeons.

Check the earnings report. Last quarter Intel made $7.8B on Core laptop and deskop and $3B on Xeons. Their Xeon (data center) revenue is also dropping quarter on quarter. If Core dies so does Intel. Xeons are a loss proposition for Intel as they are no longer competitive with AMD and ARM.

 

[jayjr1105]

People nit picking on AMD for a few extra seconds of boot time quickly forgot about the droves of folks that are having to wait 6+ weeks RMA time to get a raptor lake chip replaced. That is if they didn't dump that system for AM5 already.

 

[Dr. Dro]

Go look at Phoronix's review of the Zen5 consumer chips and then of the Zen5 Epyc chips....you'll see in professional tasks they had huge performance improvements. Perhaps gamers are unaware, but the enterprise x86 market dwarfs the gaming market in total addressable sales and revenue. Zen5 did have good performance gains where it is most lucrative for AMD.

Don't get me wrong, I agree and have stated the same before. In particular the gains in AVX-512 workloads are tremendous, which is why it works so well in the enterprise segment.

But in many aspects it has barely moved the needle - which is what puzzles me.

 

[AusWolf]

How disappointing... This means that the vast majority of this almost meaningless perf increase is coming from clock speed enhancements only. The Zen 5 architecture really is an utter disappointment. Something went very wrong inside AMD. The exact same monumental f-up as what caused the awful RDNA 3 chip which has all but destroyed the Radeon brand. It's almost as if nobody inside AMD has ever actually seen these chips in a real working system and measured their performance - It's like they are only looking at simulation results.

How this miniscule kind of performance increase is going to generate upgrade sales is beyond me. Hey look, you can have maybe 5-10% more frames than your old and awful 7800X3D for the low low price of $500+!!! Grab 'em while they're hot!

One thing is for sure, this Zen 5 is not a fixable situation, it's almost certainly either an errata in the design, or a deliberate lie by AMD marketing.

Somebody needs firing.

Why some people still think that every CPU/GPU generation is meant for single generational upgrades is beyond me.

 

[Beginner Macro Device]

But in many aspects it has barely moved the needle - which is what puzzles me.

There's nothing to be puzzled with.

AMD are milking business consumers who use their CPUs for professional workloads where popping as much of framerate as possible doesn't really matter to anyone. Nor does Google Chrome speed matter to them. They focused on the most profitable to fix weakspot and fixed it.

And gamers can be relaxed about CPUs since it's about 99.99% chance you'll end up being heavily GPU limited anyway (unless you really need these five million FPS). Excellent gaming CPUs go for 200 USD. 5800X and 12600K can drive whatever game to very playable experience, and these couple games that they don't... they don't run well on any CPU, really. Excellent gaming GPUs? DO THEY EXIST? Apart from 4090, of course, but it also makes for a bittersweet option even if the price ain't of any concern.

I don't see a reason for 9800X3D to exist in the first place. Just sell 7800X3D until it doesn't work and then release a new gaming CPU when anything AM5 goes obsolete. AMD are in zero rush with this.

 

[DemonicRyzen666]

I think the IF speeds are the problem need to push faster DDR5 at 1:1

but it doesn't look like that will be addressed until Zen 6.

not run DDR5 8000 at 2:1:1 and instead run it a 2:1: 2/3? There are dividers now for this.

 

[Beginner Macro Device]

Why some people still think that every CPU/GPU generation is meant for single generational upgrades is beyond me.

Some folks believe that their penis shrinks if they don't own the latest tech. It only costs you ♂three hundred bucks♂ annually to upgrade your PC to a performance tier monster every five or six years. Does a PC with i7-8700 and RTX 2080 look bad today? It doesn't, it's just not enough for 100+ FPS AAA gaming. You can bite this bullet.

 

[AusWolf]

Excellent gaming GPUs? DO THEY EXIST?

Of course they do. One just has to keep expectations in check. At 1080/1440 medium settings, you can game on a 6600 XT, too, if you want.

 

[Beginner Macro Device]

One just has to keep expectations in check.

No, no, I need Huang's expectations being in check. Not mine. Give me a GPU generation that actually provides uplift in FPS per $.

 

[AusWolf]

No, no, I need Huang's expectations being in check. Not mine. Give me a GPU generation that actually provides uplift in FPS per $.

His expectations are in check. Whatever crap he releases, people will gobble up like hotcakes. This is what he expects, and this is what he gets, unfortunately.

 

[Space Lynx]

Why some people still think that every CPU/GPU generation is meant for single generational upgrades is beyond me.

For me personally, it's just because it is a hobby and I like building and re-building. Also, there isn't much I spend money on beyond the basic living things everyone needs. Most things that people buy just for fun simply don't interest me, so I have no issues selling at a loss and building new every year or two. However, moving forward I am going to try to move to a three year or four year cycle, and I may keep my 7900 XT for life, because the world is just becoming more and more unstable, who knows when the next gpu shortage comes, probably sooner than we think. My GoG library is fully backed up offline, so even in worst case scenarios at least I can game.

 

[TheinsanegamerN]

How disappointing... This means that the vast majority of this almost meaningless perf increase is coming from clock speed enhancements only. The Zen 5 architecture really is an utter disappointment. Something went very wrong inside AMD. The exact same monumental f-up as what caused the awful RDNA 3 chip which has all but destroyed the Radeon brand. It's almost as if nobody inside AMD has ever actually seen these chips in a real working system and measured their performance - It's like they are only looking at simulation results.

How this miniscule kind of performance increase is going to generate upgrade sales is beyond me. Hey look, you can have maybe 5-10% more frames than your old and awful 7800X3D for the low low price of $500+!!! Grab 'em while they're hot!

One thing is for sure, this Zen 5 is not a fixable situation, it's almost certainly either an errata in the design, or a deliberate lie by AMD marketing.

Somebody needs firing.

It's simple, Zen 5 started as a server CPU, because EPYC makes way more money then desktop users. Where they made notable improvements are in energy efficiency, not outright performance. There's no reason for them to try anything radical given how poorly their competition is performing, my guess is zen 6 with the re designed interconnect controller will bring larger gains.

Or course, in desktop linux and anything AVX512 based, there are major improvements. Gone are the days of low hanging fruit from zen 1. Zens a mature arch now, gains will be smaller. Which is fine, means longer lifespans for hardware out there.

For me personally, it's just because it is a hobby and I like building and re-building. Also, there isn't much I spend money on beyond the basic living things everyone needs. Most things that people buy just for fun simply don't interest me, so I have no issues selling at a loss and building new every year or two. However, moving forward I am going to try to move to a three year or four year cycle, and I may keep my 7900 XT for life, because the world is just becoming more and more unstable, who knows when the next gpu shortage comes, probably sooner than we think. My GoG library is fully backed up offline, so even in worst case scenarios at least I can game.

I'm still doing fine with my 5800x3d/6800xt combo. The 7900xtx was tempting but given how much I had to pay before, I'm getting a long lifespan outta this card.

The hobbyist in me cries out, I've been satiating it with building small mini ITX media rigs and customizing some older cases. Also building a windows 7 SLI rig for funsies.

 

[Space Lynx]

I'm still doing fine with my 5800x3d/6800xt combo. The 7900xtx was tempting but given how much I had to pay before, I'm getting a long lifespan outta this card.

One of my best friends who lives in Scandanavia bought a 1080 ti on launch day, he still uses it daily and games on it. Truly a GOAT moment, similar to the intel i5-2500k folks who bought that on launch day. Talk about longevity, those are some rare wins.

 

[Makaveli]

not run DDR5 8000 at 2:1:1 and instead run it a 2:1: 2/3? There are dividers now for this.

i'm aware of the ratio's but 1:1 gives best performance.

if we could have even seen DDR5 7200 at 1:1 on Zen 5 I think we would see a fairly decent performance gain.

Some folks believe that their penis shrinks if they don't own the latest tech. It only costs you ♂three hundred bucks♂ annually to upgrade your PC to a performance tier monster every five or six years. Does a PC with i7-8700 and RTX 2080 look bad today? It doesn't, it's just not enough for 100+ FPS AAA gaming. You can bite this bullet.

Also some of us like to ride the hardware flip wave and if you wait to long to sell your gear resale value drops. That is the primary reason i'm doing Zen 3 to 5 if I wait until Zen 6 i'm going to get alot less for my Zen 3 gear on resale.

I however agree with the main point doing gen on gen upgrades are silly and mostly epeen. Skipping a gen is always worth it since you get much larger performance differences. I generally like to do gpu's flips when I see a 50% performance increase from my current gpu at a reasonable price. For CPU's that gain has to be 30%+.

 

[DemonicRyzen666]

i'm aware of the ratio's but 1:1 gives best performance.

if we could have even seen DDR5 7200 at 1:1 on Zen 5 I think we would see a fairly decent performance gain.

1:1: doesn't work if you're bandwidth limited you have to go beyond the sync their is no reason to use a sync at higher speeds, bandwidth usually wins over latency.
Which is what we keep coming back to with the infinity link.

What's the 2:1:1 speed sync for 8000mhz anyways?

 

[Makaveli]

1:1: doesn't work if you're bandwidth limited you have to go beyond the sync their is no reason to use a sync at higher speeds, bandwidth usually wins over latency.
Which is what we keep coming back to with the infinity link.

What's the 2:1:1 speed sync for 8000mhz anyways?

I don't want to choose Bandwidth vs Latency I want both.