AMD Ryzen 5 3600 Review Leaks, Shows Impressive Performance

[Hotobu]

Still not there in gaming. 6700k performance. 30fps less on far cry 5, ouch. I know 3800x is superior + oc but dont think it can beat Intel in games, so no upgrade to me. Change my mind AMD.

Isn't FC5 known to be a disproportionately poor performer on AMD platforms?

 

[dicktracy]

Isn't FC5 known to be a disproportionately poor performer on AMD platforms?

9900k looks to be GPU bound on the other games. One one hand, this just proves that you should spend more on the GPU instead. On the other hand, buying a brand new CPU today---you would want something that blows previous CPUs out of the water outside of fancy Cinebench numbers. I don't see that happening with Zen 2 or even Comet Lake.

 

[Bones]

IPC/ST only matters so much on AMD, because their latencies are higher. So in fact they would need like 20% better IPC than Intel to have same framerates on non GPU bound scenarios. This is why on Far Cry 5 you see the Ryzen chip doing the same as an old 6700k. It won´t still reach Intel numbers for high refresh gaming, but it´s getting better. By the time they catch Intel or even surprass, that´s when I will change to AMD. Until then, Intel it is to me.

The very fact AMD is catching up is good for both camps - No way I could have scored the two Apex IX boards for dirt-cheap ($89 per board - Shipped! ) as I did last week yet I did.

AMD stuff is performing better and since Intel is having so many issues, prices of their stuff is falling. All we can hope for is to not have a total flip of the situation and I don't believe that will happen, I'm expecting things to level out in time between the two.

Doesn't matter to me, I'm getting a Ryzen chip and possibly a board coming up on the 7th.

 

[Xzibit]

Isn't FC5 known to be a disproportionately poor performer on AMD platforms?

TPU had a 29fps difference at 1080p between 9900K to 2700X

9900K = 133.6
2700X = 104.3

Spoiler: TPU 9900k FC5 1080p chart

Going by the ELC charts 3600 is able to close the gap by half of the difference.

 

[Zareek]

If these are real numbers, that is a lot of performance for a $200 processor.

 

[Space Lynx]

impressive no way, i want better ST than intel, no point upgrading from haswell still

agreed, was hoping for actually beating intel at all levels with 7nm.

 

[Caqde]

I wonder what clockspeeds the 3600 was actually hitting in FC5 it seems to utilize 6 threads from what core utilization benchmarks show so I'm certain that the 3600 is running at < 4.2Ghz maybe 4Ghz or less and the 9900K should be running at ~4.8ghz to 5ghz depending on the Motherboard settings. Based on this the IPC of Ryzen 3x00 is equal to or slightly higher than the 9900k in FC5 if it is running at 5ghz which is likely.

 

[Zubasa]

impressive no way, i want better ST than intel, no point upgrading from haswell still

Skylake already has some IPC advantage over Haswell.
The fact that the 3600 edges out the 6700k is pretty impressive.
The 6700k has a base clock of 4Ghz, while the 3600 is a 65W part with a base clock of 3.6Ghz.
Then there is MCE, which runs all of the 6700K's cores at 4.2Ghz.
Chances are on average the 3600 is running at lower clocks than the 6700k while beating it.

IPC/ST only matters so much on AMD, because their latencies are higher. So in fact they would need like 20% better IPC than Intel to have same framerates on non GPU bound scenarios. This is why on Far Cry 5 you see the Ryzen chip doing the same as an old 6700k. It won´t still reach Intel numbers for high refresh gaming, but it´s getting better. By the time they catch Intel or even surprass, that´s when I will change to AMD. Until then, Intel it is to me.

The IPC numbers already have taken latency into account.
You cannot measure IPC without being affected by latency.
AMD claims 15% IPC over Zen+ and it does perform more than 15% faster than 2700X in Farcry 5,
The 3600 non-X should be running slightly lower clock speeds than the 2700X as well.
Given that the 2700X has 100Mhz higher base and boost clocks.

I wonder what clockspeeds the 3600 was actually hitting in FC5 it seems to utilize 6 threads from what core utilization benchmarks show so I'm certain that the 3600 is running at < 4.2Ghz maybe 4Ghz or less and the 9900K should be running at ~4.8ghz to 5ghz depending on the Motherboard settings. Based on this the IPC of Ryzen 3x00 is equal to or slightly higher than the 9900k in FC5 if it is running at 5ghz which is likely.

For the 9900k, high-end motherboards have MCE on by default, so in games it is pretty much running 5Ghz all core.
Only on things like encoding etc where you will see less boost.

 

[dyonoctis]

i'm confused. Isn't zen 2 supposed to have better lantency than zen ? And what's up with that write speed ? it's like they took a step back from even zen...

 

[biffzinker]

Isn't zen 2 supposed to have better lantency than zen ?

Might have something to do with the CPU cores on the chiplet die having to go through the revised Infinity Fabric bus out to the I/O die with the memory controllers. Adds distance between the cores and memory controller compared to everything before on one die.

 

[Metroid]

we suggest taking them with a grain of salt.
we suggest taking them with a grain of salt.
we suggest taking them with a grain of salt.

^ that

hehe, dont think people understood what you meant here, i did hehe,

 

[HwGeek]

I see many are complaining about high voltages for 7nm part, but as Lisa told us that going down to 7nm made moving electrons harder, so maybe they overcome this by increasing the voltage but the current decreased? [P=IV], after looking at 3600 leaks I saw that the max core voltage was 1.45V and max power was 10.5~11W, for contrary my 2700X at same voltage consumed 13~14W.
So who can test this?

Also They added the WR on LN2 for the 16C ES @5.27Ghz:

SampsonJackson`s Geekbench4 - Multi Core score: 64953 points with a Ryzen 9 3950X

The Ryzen 9 3950X @ 5273MHzscores getScoreFormatted in the Geekbench4 - Multi Core benchmark. SampsonJacksonranks #1 worldwide and #1 in the hardware class. Find out more at HWBOT.

hwbot.org

 

[Crackong]

If 4.2GHz is all they needed to reach 487 pts in R20 single threaded test, it is really impressive.

 

[Xuper]

Still not there in gaming. 6700k performance. 30fps less on far cry 5, ouch. I know 3800x is superior + oc but dont think it can beat Intel in games, so no upgrade to me. Change my mind AMD.

Why Should AMD change your mind when You're of Top AMD haters in entire Forum ?

 

[64K]

It's pretty impressive to me what AMD has managed to accomplish on the CPU side. Remember it was just a few years ago that some financial analysts were saying it was likely that they would have to file for bankruptcy and possibly sell off RTG. Lisa Su chose to put most of the company's efforts into Ryzen and let their graphics business flounder somewhat and for anyone that keeps up with the financials that was obviously the smart thing to do.

Last I saw AMD had a 20% market share in discrete GPUs and when Intel launches their lineup next year then AMD's market share will probably drop even further.

 

[Vya Domus]

IPC/ST only matters so much on AMD, because their latencies are higher.

If only you'd have a clue about how these things work you'd realize that if AMD was able to increase IPC even though memory bandwidth and latency remained constant that meant this wasn't the primary constraint.

But keep making stuff up for your template Intel fanboy comments and entertain us.

 

[HwGeek]

https://twitter.com/x/status/1140843448372797440

Looks like the CPPC2 ain't active yet.

 

[XiGMAKiD]

9900K is now within spitting distance

 

[bug]

View attachment 125734
9900K is now within spitting distance

Doesn't really matter. What matters is 3600 is the weakest of the new bunch
Besides vacation photo editing, I don't have much use for more than the 4 cores I currently have. But these new CPUs are really, really tempting.

 

[ic3r0ck]

I'm very curious to know why the memory write speeds are so low. They tested on x470 and also x570 and the same low speed was observed. I'm a bit worried because it's almost half the write speed of 2700x and also the higher latency does not explain the lower write speed. Let's hope it's a BIOS issue that will be fixed soon and/or that it's not present on all x570 motherboards.

 

[bug]

I'm very curious to know why the memory write speeds are so low. They tested on x470 and also x570 and the same low speed was observed. I'm a bit worried because it's almost half the write speed of 2700x and also the higher latency does not explain the lower write speed. Let's hope it's a BIOS issue that will be fixed soon and/or that it's not present on all x570 motherboards.

Well, the memory controller was moved onto a separate die. That never improves latency. But latency by itself doesn't mean much. If it does not impact the greater picture (i.e. performance), don't worry about it.

 

[Fabel]

3200 is below AMD recommendation for Zen2 RAM settings and latency does not match.

Either AMD was too optimistic or that review is pretty suspect.

 

[bug]

AMD's graph says 69, that review says 76. That's a 10% difference it could be explained by a number of factors. One of them being AMD knows better how to tweak memory timings.

 

[Fabel]

AMD's graph says 69, that review says 76. That's a 10% difference it could be explained by a number of factors. One of them being AMD knows better how to tweak memory timings.

10% in RAM latency and 3200 which affects IF frequency, not the best conditions. It looks it isn't even CL14.

I'd like info on the older AMD and Intel setups to compare apples to apples. 3200 is low even for Zen+ IMHO.

 

[bug]

10% in RAM latency and 3200 which affects IF frequency, not the best conditions. It looks it isn't even CL14.

I'd like info on the older AMD and Intel setups to compare apples to apples. 3200 is low even for Zen+ IMHO.

You'll have your answers in like 10 days. Don't fret