AMD Ryzen 5 5600X

[Mussels]

@W1zzard can u bench again but now with 4 sticks of the same RAM?

that's... new

 

[80-watt Hamster]

the one big oddity about AM4, is that the mobos waste money and space on all these video out connectors that a large portion of users simply cant use. It causes confusion for the uneducated as to why the ports dont work, and then for the high end users we could at least use an APU's hardware encoding for streaming purposes. That said, i see it as more of a neutral fact, than a negative for the CPU review.

Do we know what proportion of AM4 boards get APUs installed (in aggregate, not just X*70)? In any event, as a manufacturer, I can understand hedging that bet, even on high-end models. It's likely that every review of a board without any video output headers would ding it for the lack thereof.

 

[TheUn4seen]

That's just not true. We can see some huge massive increases all over the place. Horizon and Death Stranding, Counter Strike, Shadow of the Tomb Raider, Hitman, Star Wars Squadrons, Serious Sam. Also the best power draw comparison ever. 67W beating Intel's 218W OC in gaming performance.

Check out the massive wins on Gamer's Nexus's reviews. Civ 6 for example. I'm not happy with the price of course, but AMD is in the lead, and the lead grows at 720p showing good future scaling.

Well, yes, but really no. In games it trades blows with 10700 non-K, with fairly similar power draw. It's better than some older CPUs, but not really worth the hassle and cost of changing the whole platform.
So yes, AMD is in the lead, I'm not disputing that, but the difference is not earth shattering and certainly not worth changing the platform for anyone who has a modern or even a semi-modern CPU.
The CPU matters mostly at 1080p and below. I don't remember when was the last time I played a game at such lousy resolution.
Yes, I know there are people still using Sandy Bridge and 768p screens. But even for them investing in a better GPU would bring a bigger performance increase.

 

[Metroid]

There is this video on gamernexus saying 4 stickies give almost 10% performance on zen 3, maybe wizard tested with 2 stickies?

Yes, his memory setup was,

Memory:	2x 8 GB G.SKILL Flare X DDR4 DDR4-3200 14-14-14-34​

@W1zzard, anyway to rerun with 4 stickies?

 

[ixi]

There is this video on gamernexus saying 4 stickies give almost 10% performance on zen 3, maybe wizard tested with 2 stickies?

Yes, his memory setup was,

Memory:	2x 8 GB G.SKILL Flare X DDR4 DDR4-3200 14-14-14-34​

@W1zzard, anyway to rerun with 4 stickies?

It would be great if tpu would re-check but now with quad channel if it was done with dual.

Anyway, that is almost nothing new when single channel performance is worse than dual channel, and dual is worse than quad.

 

[Metroid]

It would be great if tpu would re-check but now with quad channel if it was done with dual.

Anyway, that is almost nothing new when single channel performance is worse than dual channel, and dual is worse than quad.

It's not quad channel, b450 and x570 does not support quad channel, all dual channel, quad channel are for systems with threadripper.

 

[ixi]

It's not quad channel, b450 and x570 does not support quad channel, all dual channel, quad channel are for systems with threadripper.

Haven't really checked amd mobos if on dying costumers there are quad-channel mobos. If no, then yea 2 vs 4 sticks on dual channel.

 

[Mussels]

Haven't really checked amd mobos if on dying costumers there are quad-channel mobos. If no, then yea 2 vs 4 sticks on dual channel.

its ranks, that need to be tested. Most of the time we had single rank ram because it let us OC higher, but this time around the extra ranks might give more performance.

2/4 sticks can be 2/4/8 ranks... and this is gunna take up a lot of poor w1zzys time, but he probably will test it.

 

[Space Lynx]

@W1zzard can u bench again but now with 4 sticks of the same RAM?

well crap... now i have to buy two more sticks of ram lmao

its ranks, that need to be tested. Most of the time we had single rank ram because it let us OC higher, but this time around the extra ranks might give more performance.

2/4 sticks can be 2/4/8 ranks... and this is gunna take up a lot of poor w1zzys time, but he probably will test it.

so would my 32gb 2x16gb already be considered 4 ranks? so i dont need 2 extra sticks for this performance bump?

 

[InVasMani]

well crap... now i have to buy two more sticks of ram lmao




so would my 32gb 2x16gb already be considered 4 ranks? so i dont need 2 extra sticks for this performance bump?

2 to 4 ranks total from either 2x16GB (due to the higher capacity it's virtually all dual rank or was when DDR4 DIMM's that size launched with Skylake) or 4x8GB (depends on dual rank) the other aspect is bank groups has a impact on performance. Basically you reduce latency and increase bandwidth. DDR5 doubles the bank groups over DDR4 and DDR3 didn't have bank groups that was one of the new features implemented.

When DDR4 came out with Skylake frequency and latency of 16GB DIMM sticks were different and not as comparable to 8GB DIMM sticks. That one part of the matter perhaps that could explain some of the difference outside of the first two things. If the latency/frequency of higher capacity chips have caught up in performance parity despite the density that's going to obviously positively impact performance. The bigger issues of course is the ranks and bank groups of course.

 

[Space Lynx]

4 ranks total I think from 2x16GB (I believe they are all dual rank if I'm not mistaken in order to get 16GB on a single DIMM you need dual rank or did at least when those DIMM's came on the market for Skylake that could've changed though if the chip density has increase and thus half as many chips are required to fill that capacity) or 4x8GB (depends on double sided with dual rank), but outside of that you've also got 8 bank groups per DIMM which helps with prefetch performance. I think what people are missing is the prefetch performance from using 4-DIMM's because of the bank groups though the rank interleaving helps as well though it's the combination of both things. You can switch between 4 DIMM's with 8 banks per DIMM and get a bit more bandwidth by avoiding the refresh period which if you extend the refresh on a DDR DIMM you achieve more bandwidth, but at the expense of potential instability of course.

steve in that gn video says single rank 2x16gb performs best of all. video below i timestamped it for you
edit timestamp didnt work - its at 23:53

 

[InVasMani]

Wendell did actually if I'm not mistaken. Also I think he implies that it is or rather it's best compromise. I believe what Wendell was suggesting is that the bang for buck is with a 2x16 kit as opposed to 4x8GB while a 4x16GB won't net you much additional performance over a 2x16GB kit though didn't explicitly say it that way. I can flat out tell you when I tested 2x16GB and 4x16GB in the past 4x16GB performed better at the same settings and ran several tests of it. I don't think Wendell is the kind of guy that's going to blindly recommend 4x16GB as the "sweet spot" over 2x16GB. The keyword is "sweet spot" it's not that that 4x16GB might not have a tiny fraction of a advantage it's just that it isn't at all worth the extra price consideration unless you need the extra capacity itself instead focus more on a 2x16GB that's higher frequency with tighter timings hence once again the "sweet spot" he mentions. That's my take on his words, but you can interpret it any which way you like. I mean "sweet spot" or not I got better performance with 4x16GB at the same settings. Feel free to ask Wendell what he meant by it though doesn't phase me a bit. The OS itself will actually make use of more memory if you install more memory for the record which is another consideration. The other part of the equation is 2x16GB could have better stability than 4x16GB so that's a factor in the "sweet spot" consideration and perticularly true depending on the motherboard and I suppose CPU in question.

 

[Space Lynx]

Wendell did actually if I'm not mistaken. Also I think he implies that it is or rather it's best compromise. I believe what Wendell was suggesting is that the bang for buck is with a 2x16 kit as opposed to 4x8GB while a 4x16GB won't net you much additional performance over a 2x16GB kit though didn't explicitly say it that way. I can flat out tell you when I tested 2x16GB and 4x16GB in the past 4x16GB performed better at the same settings and ran several tests of it. I don't think Wendell is the kind of guy that's going to blindly recommend 4x16GB as the "sweet spot" over 2x16GB. The keyword is "sweet spot" it's not that that 4x16GB might not have a tiny fraction of a advantage it's just that it isn't at all worth the extra price consideration unless you need the extra capacity itself instead focus more on a 2x16GB that's higher frequency with tighter timings hence once again the "sweet spot" he mentions. That's my take on his words, but you can interpret it any which way you like. I mean "sweet spot" or not I got better performance with 4x16GB at the same settings. Feel free to ask Wendell what he meant by it though doesn't phase me a bit. The OS itself will actually make use of more memory if you install more memory for the record which is another consideration. The other part of the equation is 2x16GB could have better stability than 4x16GB so that's a factor in the "sweet spot" consideration and perticularly true depending on the motherboard and I suppose CPU in question.

yeah in the video steve says wendell will have more tests up next week. so we will see

 

[Mussels]

well crap... now i have to buy two more sticks of ram lmao

so would my 32gb 2x16gb already be considered 4 ranks? so i dont need 2 extra sticks for this performance bump?

Use CPU-Z to find out



This is why we need the 2/4/8 ranks comparison to know what gives best performance... this is new hardware, so we dont know yet

 

[InVasMani]

yeah in the video steve says wendell will have more tests up next week. so we will see

Can't speak for how it works with Ryzen, but I'd expect fairly similar if I had to wager a guess especially given Skylake is quite old relative to Zen 3.

 

[Megas]

The games used in the benchmarks, which are CPU vs. GPU dependent games?

 

[Searing]

Well, yes, but really no. In games it trades blows with 10700 non-K, with fairly similar power draw. It's better than some older CPUs, but not really worth the hassle and cost of changing the whole platform.
So yes, AMD is in the lead, I'm not disputing that, but the difference is not earth shattering and certainly not worth changing the platform for anyone who has a modern or even a semi-modern CPU.
The CPU matters mostly at 1080p and below. I don't remember when was the last time I played a game at such lousy resolution.
Yes, I know there are people still using Sandy Bridge and 768p screens. But even for them investing in a better GPU would bring a bigger performance increase.

No, Techspot has it beating the 10700k. The 5900X is beating the 10900k.

"For testing the AMD CPUs we're using the MSI X570 Godlike with four 8GB G.Skill TridentZ DDR4-3200 CL14 memory modules for a 32GB capacity and then cooling all test systems is the Corsair iCUE H150i Elite Capellix AIO." ................ so 4 sticks of ram might be the difference.

Also the power draw shown in Gamer's Nexus's review is the opposite of what you said. "with fairly similar power draw" is a load of nonsense. Confirmation bias here? Go watch Hardware Unboxed and Gamer's Nexus's reviews and then come back here. No one is asking you to change your platform. Do what you want. Intel buyers have been spending 100's of dollars and changing their platforms for 5 percent performance. Gamer's Nexus called the 5000 series the "largest generational improvement ever seen".

You pretended to respond to my comment and didn't respond to my comment, so I'll just repeat it: "We can see some huge massive increases all over the place. Horizon and Death Stranding, Counter Strike, Shadow of the Tomb Raider, Hitman, Star Wars Squadrons, Serious Sam. Also the best power draw comparison ever. 67W beating Intel's 218W OC in gaming performance."

 

[sepheronx]

just to use as reference, here is the techspot article.

AMD Ryzen 7 5800X Review

Having reviewed Ryzen 5000 12-core and 16-core models, today we're testing the Ryzen 7 5800X, AMD's latest 8-core CPU. So far we've been impressed by the Ryzen...

www.techspot.com

Now, if we average the gaming performance seen across all our game sample, we find that the 5800X is roughly on par with the 10900K and for the most part isn’t much slower than the 5900X and 5950X. It also comes out 6% ahead of the Core i7-10700K and 23% ahead of the older 3700X.

 

[80-watt Hamster]

It's a good CPU, but some people want to see it as a be-all-end-all of CPUs. For games it's fine, but whatever. Any CPU from 8600k upwards will suffice, and if you play at 1440/2160 it really doesn't matter, 6700k will get you within 3% of the performance, so spend your money on a GPU. For me personally. the lack of any iGPU is a show stopper - in a gaming PC I rarely need it, but when I do, i REALLY do, and in a work PC I don't want a dedicated GPU, I much prefer a low power, slim and silent machine.

Where this generation really shines seems to be the productivity oriented high end with the 5900X/5950X.

That's just not true. We can see some huge massive increases all over the place. Horizon and Death Stranding, Counter Strike, Shadow of the Tomb Raider, Hitman, Star Wars Squadrons, Serious Sam. Also the best power draw comparison ever. 67W beating Intel's 218W OC in gaming performance.

Check out the massive wins on Gamer's Nexus's reviews. Civ 6 for example. I'm not happy with the price of course, but AMD is in the lead, and the lead grows at 720p showing good future scaling.

Well, yes, but really no. In games it trades blows with 10700 non-K, with fairly similar power draw. It's better than some older CPUs, but not really worth the hassle and cost of changing the whole platform.
So yes, AMD is in the lead, I'm not disputing that, but the difference is not earth shattering and certainly not worth changing the platform for anyone who has a modern or even a semi-modern CPU.
The CPU matters mostly at 1080p and below. I don't remember when was the last time I played a game at such lousy resolution.
Yes, I know there are people still using Sandy Bridge and 768p screens. But even for them investing in a better GPU would bring a bigger performance increase.

No, Techspot has it beating the 10700k. The 5900X is beating the 10900k.

"For testing the AMD CPUs we're using the MSI X570 Godlike with four 8GB G.Skill TridentZ DDR4-3200 CL14 memory modules for a 32GB capacity and then cooling all test systems is the Corsair iCUE H150i Elite Capellix AIO." ................ so 4 sticks of ram might be the difference.

Also the power draw shown in Gamer's Nexus's review is the opposite of what you said. "with fairly similar power draw" is a load of nonsense. Confirmation bias here? Go watch Hardware Unboxed and Gamer's Nexus's reviews and then come back here. No one is asking you to change your platform. Do what you want. Intel buyers have been spending 100's of dollars and changing their platforms for 5 percent performance. Gamer's Nexus called the 5000 series the "largest generational improvement ever seen".

You pretended to respond to my comment and didn't respond to my comment, so I'll just repeat it: "We can see some huge massive increases all over the place. Horizon and Death Stranding, Counter Strike, Shadow of the Tomb Raider, Hitman, Star Wars Squadrons, Serious Sam. Also the best power draw comparison ever. 67W beating Intel's 218W OC in gaming performance."

If you're going to call someone out for not responding in kind, make sure you're not doing the same thing yourself. Here at TPU, the 10700 is ahead of the 5600X in games at 1080p and above, while the 5600X wins in the non-gaming benchmark aggregate. Techspot doesn't list the 10700 in it's 5600X review charts, though you're correct that the 5600X consistently comes out on top there. Just. All of that is at stock clocks. TPU's power consumption charts put the 5600X and 10700 within 11W of each other in all tests.

SO: There are massive increases in performance relative to AMD's previous generations of processors. The 5600X is an excellent value at its MSRP and 15-25% faster than its predecessors (an unheard of leap in probably a decade or more), and competes with more expensive Intel parts at lower power draw. TheUn4seen may have constructed a straw man, but they also tore him down successfully: A 5600X won't be a meaningful upgrade to a 6-core+ Coffee Lake or later chip in gaming terms. But yeah, if I were shopping for a ~$300 CPU, I'd absolutely go with the 5600X. Of course, none of this matters when the 5600X is out of stock or listing for $100-200 over MSRP (at time of post).

 

[Megas]

Do you guys play a lot of CPU dependent games cause in GPU dependent games the 3600+ really doesn't make a lot of difference. This fellow has some benchmarks that aren't just Death Stranding and Tomb Raider, etc.

Games like Apex, MW, etc.

 

[sepheronx]

Games are mostly GPU dependent. In 1440p and higher, the difference diminishes greatly. If one has a Ryzen 2000 series or Intel 8000 series and higher, I don't see the need to upgrade yet.

All that depends though on the whole PCIe 4.0 and the NVME. Future games may take advantage of NVME pcie 4.0 with how consoles plan to stream textures via the nvme.

But that is a wait and see.

 

[seth1911]

299$ USA

around 400€ here.

the funny Thing is , if i wanna buy a PC with an APU:
3000G, 3200G, 3400G, or Renoir, arent on Stock here for max. MSRP

Same with the new 5600X, 5800X, 5900X, 5950X

 

[RandallFlagg]

Here's why the 10700 vs 5600X comparison is a tough one - this shows a power unlocked and a power unlocked + BCLK 10700 on the right, and 5600X on the left, average of the CPU tests.

Look deeper into the CPU tests and you'll find the 5600X mostly wins on web browser and MS Office tasks. It's losing to the 10700 on other productivity benchmarks - due entirely to the additional cores on the 10700. I'm not sure what I think of that, those (web/office) used to be Intel strongholds but its undeniable Zen 3 has much higher single thread IPC.

Nevertheless, the 10700 can and does win at both games and productivity overall - *if* you power unlock it so that it doesn't turbo down :

 

[mahoney]

that's... new

Nope has been a thing for a long time
if you populate 4 ram slots for example 4x4gb vs 2x8gb -> 4x4gb will always be faster

Sadly Steve only just discovered hot water

 

[RandallFlagg]

Nope has been a thing for a long time
if you populate 4 ram slots for example 4x4gb vs 2x8gb -> 4x4gb will always be faster

Can't you do 2x double rank DIMMS as well, I thought this was why a lot of AMD folks favored Corsair LPX (If I recall correctly, haven't looked it up)?