AMD Ryzen 7 7700X

[W1zzard]

First things that come to mind are memory speed, game selection, GeForce 3080, Windows 11, VBS enabled

 

[Denver]

Thanks @W1zzard for keeping the 2700X in this review.
On the downside it's making me consider upgrading to a 5700X or 5900X if a good deal shows up...

Meanwhile can someone help em understand how a ~5% performance gap between the 2700X and the 5900X in 2020 grew to ~25% in 2022

View attachment 263108	View attachment 263109
AMD Ryzen 9 5900X Review The Ryzen 9 5900X dominates Intel's Core i9-10900K in our testing because of AMD's massive IPC improvements. At $550, this processor is certainly not cheap, but it offers so much more performance, especially single-threaded, that AMD has a clear winner on their hands. www.techpowerup.com	AMD Ryzen 7 7700X Review - The Best Zen 4 for Gaming With the Ryzen 7 7700X, AMD is introducing their fastest Zen 4 processor for gamers. In our review we found out that gaming on the 7700X runs better than the 7900X and 7950X, thanks to the single CCD design of the 7700X. Just like on other high-end AM5 CPUs, temperatures are a problem though. www.techpowerup.com

Probably because the GPU and games used have changed...

 

[Gica]

Meanwhile can someone help me understand how a ~5% performance gap between the 2700X and the 5900X in 2020 grew to ~25% in 2022

1. RTX 3080 is way above RTX 2080Ti
2. Many AAA games in 2022 (Cyberpunk, Forza Horizon 5 and even FarCry 6 vs. 5, etc.)
3. W10 2019 edition vs W11 VBS ON.
The impact of the memory was small on the Alder Lake in games, but on the Ryzen 7000 it must be tested.
Edit: sorry, can't be tested because Ryzen doesn't support DDR4.

 

[tvshacker]

First things that come to mind are memory speed, game selection, GeForce 3080, Windows 11, VBS enabled

In that particular order?
What is VBS?
In the games that are in common between the 2 reviews (Battlefield V and Civ VI), the 5900X increased FPS on both but the 2700X didn't. I've heard Civ is very CPU intensive, did a patch or game update cause this?
Thanks for the quick replies!

 

[W1zzard]

In that particular order?

In that order I'd that

What is VBS?

Virtualization based security

 

[JAB Creations]

I'd like to see dedicated articles to running these CPUs at reasonable temperatures.

On an unrelated note: this is just one of the reasons why it's painful to read comments. ︀

 

[jsven008]

Much appreciation to W1zzard for another great review! Love the reviews, the information, the logical format, the analysis

 

[mechtech]

Was hoping to, but didn't have the time.

"Which benchmark would you like to run for IPC?" was my biggest obstacle

That's a good question. Not sure, perhaps
Cinebench........since it seems to be the go to?
one or 2 games
something else which highlights IPC and/or per core decently?

 

[Wirko]

@W1zzard , wouldn't it be more informative if you combined the three "Average Clock Frequency vs Thread Count" graphs into one, to make differences between curves more obvious? With all the vertical space saved, you could also make the graph taller than each of the three are now.

Also, it's just great that you are able to measure the CPU-only power consumption, this information is hard to come by. (Actually it's CPU+VRM, right?)

 

[xSneak]

I would like to see a game using UE4 on the benchmarks, and a gaming while streaming test.
Thanks.

 

[StellarShots]

Ryzen 7000 Delidding - Unreal Temperature improvement with Direct-Die Cooling​

Almost 20 degrees cooler.

 

[AlwaysHope]

I know I'm asking a lot, but before I raise this point, like to thank W1zzard for this comprehensive review that helps a lot with perspective on it all.

Be interesting to see the effects of OC the ring/LLC bus inside not only rocket lake but also alder lake cpus with this testing regime the author has put in place. I never hear mentions of this in reviews of these processors. Also to see if high bandwidth RAM in gear 2 for DDR4 (4400MHz+) effects on overall performance I believe will alter results compared to gear 1.
High & mid end motherboards do have options to OC these 2 parts of this 2 architectures in the bios for these platforms, & I'm sure enthusiasts would be keen to see this in testing results.

I do understand time is an issue, but in some applications, the speed of the ring/LLC bus does have an effect in performance outcomes & even more when combined with high bandwidth gear 2 RAM.

However, like some have commented before, the total platform costs for AM5 are a hinderance atm for some & I'm in that club atm.

 

[techguru177]

Bro you jump the gun I got Asus latest Agesa BIOS and the board post quicker and performance way better. You should retest everything this is irrelevant

 

[Jism]

The bios quality of Asus is way better then any other vendor in my experience. I regret buying a Gigabyte X570 Auros Elite. I had to because the Nic on the asus board died before that.

 

[Iain Saturn]

* The bios quality of Asus is way better then any other vendor in my experience. I regret buying a Gigabyte X570 Auros Elite. I had to because the Nic on the asus board died before that.​

Respectfully…why not buy a nic - why buy a whole new board?

 

[Jism]

Well, the nic "died" sporadicly... (Intel) Not before throwing the most weirdest erratic behavour in regards of my internet experience. The NIC was gone (not visible in bios even) and it took less then 2 weeks to take the whole board along with it. RMA would take weeks and i was in the need of a new board.

 

[monseven7]

95 deg is hot, so if everything is ok with it, then its actually not bad.
Originally, I thought yeh, more fan to cool, noisy, just harder to cool. But it should actually be the opposite.
In theory, higher operating temp means higher Td to ambient air temp, therefore less air to move to keep it cool.
NFI if higher Td makes existing coolers more efficient, e.g. wicking withing heat pipes works better. For liquid, should work better, transfer should be more efficent.

To spell it out a bit, 105W @ 40 deg Td vs 105W @ 70 deg Td. Power the same but Td changes.

Which in theory means, 2l/s air vs 1.25l/s so less has to be shifted.

So just boosting fan curves up to ramp at 100 deg perhaps.

Useful heat though, can boil water and heat rooms better.

 

[W1zzard]

@W1zzard , wouldn't it be more informative if you combined the three "Average Clock Frequency vs Thread Count" graphs into one, to make differences between curves more obvious? With all the vertical space saved, you could also make the graph taller than each of the three are now.

Also, it's just great that you are able to measure the CPU-only power consumption, this information is hard to come by. (Actually it's CPU+VRM, right?)

I tried, it will look like a colorful mess, and you can't see anything when the 3 curves are identical like on 7600X

Also, it's just great that you are able to measure the CPU-only power consumption, this information is hard to come by. (Actually it's CPU+VRM, right?)

Yes, I measure on the 12V side, so there's still the VRM, but it's completely impractical to reliably measure power on the CPU's ~1.2V side. The losses in the VRM are tiny. You're pumping hundreds of amps through it and can still passively cool it

and a gaming while streaming test.

streaming is free on modern graphics cards?

I would like to see a game using UE4 on the benchmarks

Borderlands 3, that's why it's included. Or did you mean UE5? No games with that yet. I did look into making my own with UE5, but felt it couldn't reach the AAA quality required

Bro you jump the gun I got Asus latest Agesa BIOS and the board post quicker and performance way better. You should retest everything this is irrelevant

Where I can I get that BIOS?

https://imgur.com/9Le1aTA

0605 is a typo, it's supposed to be 0604
"1) The reviewer's guide mentions ASUS Hero BIOS 0605 as optimum, the press site only lists 0604. I guess this is a typo in the RG?"
"1. Typo, 0604 is OK"

The bios quality of Asus is way better then any other vendor in my experience. I regret buying a Gigabyte X570 Auros Elite. I had to because the Nic on the asus board died before that.

Had no noteworthy issues on ASUS or I would have mentioned it. The only bug I encountered is that loading a profile does not restore the "Show ASUS logo on POST screen: off" setting. Happens on Z690, too

 

[AusWolf]

Very nice!

So... for gaming, this is a 7950X with a much lower power consumption.

It's also interesting to see that these CPUs essentially work like GPUs with their boost behaviour, that is they boost until they hit thermal limits.

I was tempted to buy a 7950X, but the more I think about it, the more I'm tempted to buy the 7700X instead.

 

[Wirko]

95 deg is hot, so if everything is ok with it, then its actually not bad.
Originally, I thought yeh, more fan to cool, noisy, just harder to cool. But it should actually be the opposite.
In theory, higher operating temp means higher Td to ambient air temp, therefore less air to move to keep it cool.
NFI if higher Td makes existing coolers more efficient, e.g. wicking withing heat pipes works better. For liquid, should work better, transfer should be more efficent.

To spell it out a bit, 105W @ 40 deg Td vs 105W @ 70 deg Td. Power the same but Td changes.

Which in theory means, 2l/s air vs 1.25l/s so less has to be shifted.

So just boosting fan curves up to ramp at 100 deg perhaps.

Useful heat though, can boil water and heat rooms better.

We'll all learn from experience but this doesn't look good. It may be safe to run the new chips at 95°C in the long term, I can believe that. But I'm sure that they have less headrom than before. Even without overclocking, there are chip-to-chip variations. Some live in hot areas and may not have an air conditioner. A processor pushed that far up the temperature scale is less forgiving to every detail like having less than ideal airflow in the PC case.

Are heat pipes more effective at higher temps? That's highly questionable, and more testing will be necessary. Everything with an evaporation - condensation cycle is designed to function best in a certain range of temperatures (and pressures). Go cooler than that, and evaporation becomes slower; go hotter, and condensation becomes slower.

 

[Tyl3n0L]

Yesterday I said ill be holding off on AM5 because of price of entry however this morning I ordered the 7700x, as well as Artic Freezer II 360mm & 32Gb DDR5

Should be a nice upgrade from my i7 3770, 8GB DDR3

Just need a 500$ motherboard now... Lol.

 

[AusWolf]

We'll all learn from experience but this doesn't look good. It may be safe to run the new chips at 95°C in the long term, I can believe that. But I'm sure that they have less headrom than before. Even without overclocking, there are chip-to-chip variations. Some live in hot areas and may not have an air conditioner. A processor pushed that far up the temperature scale is less forgiving to every detail like having less than ideal airflow in the PC case.

Are heat pipes more effective at higher temps? That's highly questionable, and more testing will be necessary. Everything with an evaporation - condensation cycle is designed to function best in a certain range of temperatures (and pressures). Go cooler than that, and evaporation becomes slower; go hotter, and condensation becomes slower.

The problem with that question is that heatpipes only get hot if they get soaked in heat. You need a high power consumption CPU with a relatively large die that sits right below the heatpipes. AMD's chiplets are small and offset to one side (or corner) of the CPU package. That's why my be quiet! Shadow Rock LP didn't work well with a R5 3600, but works brilliantly with the Core i7 11700.

As much as I love the design of direct-touch heatpipes (the ones that touch the CPU IHS directly, without a coldplate in between), I would never use such a cooler with a chiplet-based CPU.

 

[trparky]

I knew that prices for this new platform were going to be high but damn, I never imagined that it would be this damn high.

Thanks @W1zzard for keeping the 2700X in this review.
On the downside it's making me consider upgrading to a 5700X or 5900X if a good deal shows up...

Me too. It'd cost me nearly $1200 to $1300 to do a full platform upgrade; that includes a new power supply, chip, motherboard, cooler, and memory. And for what? Maybe a ten percent improvement over that of current-gen AMD?

Meanwhile, I could go the cheaper route and do an upgrade at less than $600 and just replace the motherboard, chip, and cooler.

 

[HD64G]

Disappointed in the RPCS3 figures...was the 12900K used in the test an older example, still running with AVX 512 enabled?

Nearly 40% slower seems strange given that the new Ryzens have AVX 512 as well

RPCS3 didn't use the AVX512 of Zen4. It got updated yesterday though and I think @W1zzard should know about that by now.

 

[AusWolf]

I knew that prices for this new platform were going to be high but damn, I never imagined that it would be this damn high.

Me too. It'd cost me nearly $1200 to $1300 to do a full platform upgrade; that includes a new power supply, chip, motherboard, cooler, and memory. And for what? Maybe a ten percent improvement over that of current-gen AMD?

Meanwhile, I could go the cheaper route and do an upgrade at less than $600 and just replace the motherboard, chip, and cooler.

My conflict is either a new graphics card for my Rocket Lake system, or a platform upgrade and then graphics card later. I technically don't need either, it's just for the fun of building.