AMD Ryzen 7 9700X

[AMDK11]

Mike Clark admitted that Zen5 is the starting point for a new base and will allow for further, constant IPC increases in subsequent generations.

Contrary to some claims that certain solutions in Zen5 have been postponed to the next generations, AMD has already confirmed that nothing of the sort happened. A new core microarchitecture that is new from scratch takes a lot of time.

I understand that if Zen5 had an average IPC gain of +30-40%, everyone would be shouting that Zen5 is a revolutionary microarchitecture.

The microarchitecture revolution rarely brings greater IPC benefits. Especially since Zen4 is not a bulldozer.

Zen5 achieves an average IPC increase of +13% in SpecINT and +26% in SpecFP. Is it wrong?

 

[Zach_01]

BTW why do people expect that prices should be compared only with current retail prices of previous gens?
I can understand that the users want the best performance/price ratio but consider this...

Lets say they priced 9700X close to current 7700X price... What will happen in 1-2 years when Zen6 is out?
9700X will be down in price further more and...?
Should we expect Zen6 replacement to be priced again close to future 9700X price?
And by 2030 CPUs will be handed for 100$ (why not for free) while better and better fab nodes cost more and almost everything else going up too.

There is a regression in price from prev gen and this is a good sign. Not the best but at least is something.
Better even with higher SKU if the -100$ is true.

 

[R0H1T]

BTW why do people expect that prices should be compared only with current retail prices of previous gens?

Probably to make room for a $3k 5090

They would skip a CPU entirely if they could & just buy the GPU to go 8k 240fps on their latest AI powered Nvidia superchip

It's not like AMD/Intel have to pay their bill right, right? Why are they charging us for this inconvenience

 

[Mr_Engineer]

Zen 5 numbers are all over the place. Tom's hardware review shows better numbers especially for games, the numbers are below for 9700x (PBO) vs 7700x
@FHD: 21% faster, 1% low: 19% faster
@ 2K: 16% faster, 1% low: 22% faster
For productivity, 9700x (PBO) was 10% faster in single core, and 22% faster in multi core.

It's possible as some have mentioned that it could be Motherboard, BIOS, Memory, or AGESA differences/issues that is causing some numbers to be lower than expected.

I think AMD messed up as they should have removed the "x" from the 9700x and called it 9700, and had it priced at $300. And then release a better binned 9700x running at 105w default setting.

Also, AMD did a lot of work on Zen 5 but used the same old weak IOD from 2 years ago, which didn't have a great memory controller to begin with. I think this was a mistake, as it seems that in certain situations Zen 5 is starved for higher memory bandwidth. Note the great single core performance but weak multicore in many situations.

 

[DemonicRyzen666]

Following the words of Mike Clark, who admitted that only the approach with two decoders in Zen5 allowed for the expansion of the Front-end. Is history repeating itself?

At least Zen5 has no IPC regression compared to Bulldozer.

The Zen5 Core is no less "revolutionary" than the Bulldozer "Module".

That was my point. Not ever cpu had gains in ipc to become good. There is a give & take at some point.

The original phenoms the 9000/9050 series were 1% to 2% faster per clock than the phenom II 900 in IPC. The difference was lower power cosumption and higher overall clocks by 20%. The later could just reach a much higher clock then the first out doing the earlier cpus. That was thanks to a new node & other changes.

Like I said its not easy ro increase the front end without some losses here and there. Since its such a complex part in of the main design of the cpu. Its like your completely igonring the fact it was even achived this time.

 

[nguyen]

Can't believe that AMD could only managed to squeeze out a 5% progress to their uarch after 2 years.

For gamers, just grab a 7800X3D now, 9800X3D wouldn't be much faster anyway. For gamers+productivity, there are way better options out there already

 

[R0H1T]

That's BS x3d chips will have the same IPC increase & maybe slightly lower/higher gaming uplift

 

[Pumper]

Another thing that never happened with Intel is seeing a gen-to-gen power usage drops like that, so I think this is pretty good, there is lots of progress here.

Are they power usage drops, really? Zen4 increased the wattage by 50% compared to Zen3, then they toned down the TDP back to just below Zen3 levels with Zen5, so in reality, the power usage dropped by only ~10% in two generations.

 

[R0H1T]

So you're ignoring the work done per watt completely?

 

[THU31]

So you're ignoring the work done per watt completely?

I see a 5% win over the 7700 (non-X, obviously). Surely you're not comparing against the 7700X.
And the gaming result is actually worse on the new CPU.

This level of efficiency improvement is basically irrelevant on sub-100 W parts. 10-20 W makes no difference. Maybe the 9950X will benefit from this a bit more.

GPUs is where efficiency is really important. Going from a 320 W 3080 to a 200 W 4070 is an insane difference. Or getting 50% more performance with a 4080 over the 3080 at the same power.

 

[R0H1T]

Why not? Both can be tuned for even better (efficiency) results but I bet 9700x would still easily win against 7700 & 7700x by probably double digits. At stock most chips are less efficent, 7700x more so than vanilla 7700 & any number of zen 1/2/3 before them.

 

[LittleBro]

Zen 5 numbers are all over the place. Tom's hardware review shows better numbers especially for games, the numbers are below for 9700x (PBO) vs 7700x
@FHD: 21% faster, 1% low: 19% faster
@ 2K: 16% faster, 1% low: 22% faster
For productivity, 9700x (PBO) was 10% faster in single core, and 22% faster in multi core.

It's possible as some have mentioned that it could be Motherboard, BIOS, Memory, or AGESA differences/issues that is causing some numbers to be lower than expected.

I think AMD messed up as they should have removed the "x" from the 9700x and called it 9700, and had it priced at $300. And then release a better binned 9700x running at 105w default setting.

Also, AMD did a lot of work on Zen 5 but used the same old weak IOD from 2 years ago, which didn't have a great memory controller to begin with. I think this was a mistake, as it seems that in certain situations Zen 5 is starved for higher memory bandwidth. Note the great single core performance but weak multicore in many situations.

This is quite my opinion as well.

Zen 5 overall is not a fail, it's just underwhelming but it depends on how you take a look at it.
Zen 4 efficiency was terrible (except for 7800X3D) so I approve the approach that AMD took with Zen 5.
Yes, those chips are more like non-X. When paired with 105W TDP, the intergenerational perf. uplift would raise from 5-6% to 9-10%. Still not much, though.

I totally agree with the weak IOD opinion. This must be definitely holding the new chips back and a lot. They had time to do revamp, so this in particular sounds like a fail to me.
AMD is spending majority of R&D money on EPYCs and AI stuff, so does nVidia. That's the path that actually makes some money. Not good for us consumers.

I'm wondering what will 9800X3D bring.

 

[phanbuey]

I think they're trying for the laptop market so the power drops make sense. Also probably much lower failure rate and overall cost with low power parts (as intel can attest )

This is quite my opinion as well.

Zen 5 overall is not a fail, it's just underwhelming but it depends on how you take a look at it.
Zen 4 efficiency was terrible (except for 7800X3D) so I approve the approach that AMD took with Zen 5.
Yes, those chips are more like non-X. When paired with 105W TDP, the intergenerational perf. uplift would raise from 5-6% to 9-10%. Still not much, though.

I totally agree with the weak IOD opinion. This must be definitely holding the new chips back and a lot. They had time to do revamp, so this in particular sounds like a fail to me.
AMD is spending majority of R&D money on EPYCs and AI stuff, so does nVidia. That's the path that actually makes some money. Not good for us consumers.

I'm wondering what will 9800X3D bring.

As a desktop product it's a flop - they should have just released the x3d only variants - no reason to "upgrade" to use 4% less total system power overall. the 9800X3D can probably boost higher and have better MT than 7800X3D and it also is going to age better with the newer instructions, makes much more sense.

As a laptop product it's much better, and I think they're really pivoting well to compete with snapdragon x elite and the likes.

 

[b1k3rdude]

One thing thats common among ALL the reviews is that AMD has artificlally gimped the perf of the 9600/9700 by setting a myopicly low power limit

• 7600x, 4.7-5.3Ghz, 105W
• 9600x, 3.9-5.4Ghz, 65W
• 7700x, 4.5-5.4Ghz, 105W
• 9700x, 3.8-5.5Ghz, 65W
  

@W1zzard, wouldnt a more accurate comparison be to benchmark the 9600/9700 set to 105W or even 95w?

 

[phanbuey]

One thing thats common among ALL the reviews is that AMD has artificlally gimped the perf of the 9600/9700 by setting a myopicly low power limit

• 7600x, 4.7-5.3Ghz, 105W
• 9600x, 3.9-5.4Ghz, 65W
• 7700x, 4.5-5.4Ghz, 105W
• 9700x, 3.8-5.5Ghz, 65W

@W1zzard, wouldnt a more accurate comparison be to benchmark the 9600/9700 set to 105W or even 95w?

They don't gain anything at 105W - even with PBO enabled the gains are minimal. Better to get the same performance with 35% less power than to get 8% better performance at the same power.

 

[oZ65]

I see a 5% win over the 7700 (non-X, obviously). Surely you're not comparing against the 7700X.
And the gaming result is actually worse on the new CPU.

This level of efficiency improvement is basically irrelevant on sub-100 W parts. 10-20 W makes no difference. Maybe the 9950X will benefit from this a bit more.

GPUs is where efficiency is really important. Going from a 320 W 3080 to a 200 W 4070 is an insane difference. Or getting 50% more performance with a 4080 over the 3080 at the same power.

In my book, this is a performance increase per watt of 13%. Really curious to see the new Ryzen 9s.

 

[Quicks]

They should have increased the cache on base models to 64 and the 3D chips to 128. I think that would have increased performance enough for everyone to be happy.

 

[LittleBro]

They should have increased the cache on base models to 64 and the 3D chips to 128. I think that would have increased performance enough for everyone to be happy.

Indeed, but that would take a lot of space. This is probably planned for AM6 socket or waiting for suitable node.

AMD Zen 4c Not an E-core, 35% Smaller than Zen 4, but with Identical IPC

AMD on Tuesday (June 13) launched the EPYC 9004 "Bergamo" 128-core/256-thread high density compute server processor, and with it, debuted the new "Zen 4c" CPU microarchitecture. A lot had been made out about Zen 4c in the run up to yesterday's launch, such as rumors that it is a Zen 4 "lite"...

www.techpowerup.com

Ryzens 9000 lacks NPU, that's another thing that will eat some part of the chip area in the future.

 

[AnotherReader]

I get that it's ok in some specialized use cases but in general both gaming ant MT uplifts aren't very impressive.

Again if this was after maybe 12 months sure but it's been 2 years and sometimes loses to the 7700X which should never happen....

I feel like this is the 4060ti of CPU's sure it isn't bad at everything but price still sucks and generationally overall it isn't very good.

I agree with that sentiment too. For some of my use cases, it's better than anything else, but for general consumer use, it isn't significantly better than Zen 4. However, I'm also interested in why that's the case as IPC increased significantly in most workloads. I think it's a combination of two things:

• lower clocks than Zen 4 in demanding all core workloads, e.g. in the multithreaded test of Cinbench 2024, the 9700X ranges between 4.3 and 4.5 GHz compared to 4.8 to 4.9 GHz for the 7700
• identical cache capacity to Zen 4 coupled with the reuse of the IO die penalizes applications with high L3 miss rates which is the case for many games

 

[b1k3rdude]

They don't gain anything at 105W - even with PBO enabled the gains are minimal. Better to get the same performance with 35% less power than to get 8% better performance at the same power.

Fair comment, but with PBO unlocked/unlimited, the 9700X is upto 35% faster than the 7700x -

Imagine what we could do with 105W and 'curve optimizer' and the new 'curve shaper', I imagine even then the 9700X would still significantly best the 7700x. I would really like some of the tech channels to try this... My 5900x runs a -0.05v drop and CO of negative 20, and will boost to 4.95Ghz sc and 4.35 ac.

 

[THU31]

Fair comment, but with PBO unlocked/unlimited, the 9700X is upto 35% faster than the 7700x -

What kind of math is this? With PBO you get 22% more performance over the 7700X at 14% more power. You can see it in the graph. That's a 7% efficiency improvement in this scenario.

I'm genuinely baffled that so many people are happy with this kind of progress after 2 years. Then again, it just means that we don't have to upgrade. Clearly they don't want our money anymore.

 

[Mr_Engineer]

Using the Phoronix review as a reference (I know it is Linux based and mainly server benchmarks). The 9700x (65W) was 15% faster than the 7700x (105W), and 19% faster than the 7700 (65W). This is actually pretty good, higher efficiency and higher performance. On Windows things may improve, especially with newer drivers, firmware, BIOS etc.

They also mentioned that they will be comparing Zen 5 performance using DDR5-6000 and DDR5-8000 which can show if the CPU is memory bandwidth constrained in certain situations.

For now I do agree that Zen 5 is underwhelming

 

[Ayhamb99]

While the efficiency improvements are nice, this kind of performance uplift over Zen 4 reminds me of Intel's stagnation days where each gen there was only like a 5% increase. This is essentially AMD's Kaby Lake moment. Very underwhelming. Like I was hoping for a meaningful performance jump but for this to be barely outperforming the 14600K by 1% for Applications and only 2% for 1080p gaming, like come on.

The only advantage is that due to Intel being a complete clown show rn with how they're handling the CPU failures, AMD is more appealing than ever.

 

[Lesha]

This video shows the true potential of 9700X

 

[R0H1T]

This is essentially AMD's Kaby Lake moment. Very underwhelming.

KBL was an OCed 6700k, did you know that before KBL-R on laptops Intel had the same dual core ULV chips for i3/i5/i7

You can thank AMD for quad core ULV Intel chips as well

You're talking stagnation? Sitting on dual core ULV for 6-8 years or quad core desktop chips for 8-10 years that is called stagnation or underwhelming! Lots of exaggeration here with very little facts to back them up