AMD Ryzen 4000 Renoir Desktop Benchmarks Show Up To 90 Percent Uplift Over Last-Gen Flagship

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
10,785
90% on the HEDT Parts though or just low and midrange budgets?>

"The preliminary results for desktop Renoir look very encouraging on the processing side. This much is to be expected since Zen 2 and the 7nm process node opened the doors to octa-core APUs. It's a shame that the iGPU performance didn't exhibit the same level of gains. That said, overclocking is still on the table, so perhaps it's still possible to squeeze some more performance out of the iGPU in that aspect."

https://www.tomshardware.com/news/amd-ryzen-4000-renoir-desktop-benchmarks
 
The article itself states it (in a buried line). 90% CPU uplift (benchmark, anyways) over 3400G. IGP is roughly the same.
 
The GPU was far better than anything Intel had in the first two gens of APU. They’re just now catching up, so we get parity for a little while. It’s still fine for even some casual gaming. I think it might benefit from higher memory speeds available now than previously, but probably anyone running DDR4-4000 or whatever isn’t going to be using one of these anyways.

It would have been nice to have a graphics improvement, but I expect RDNA 2 just wasn’t ready and RDNA wasn’t worth the effort. I still think they should do onboard memory for the gpu - HBM or something.

At any rate, this means my 2400g TV PC / emulator box will just keep on keeping on, and I’ll see what the zen3-based APUs bring before we move on to AM5, DDR5, etc
 
Yawn, they they doubled the cores and gained 90% 3dmark physics performance....get a better headline
 
  • Like
Reactions: erek
like this
Nice, Ill be getting one for a low power Emulator build when they come out. Probably overkill, but maybe not. Some of those shader's can hog the system, especially with multiple passes.
 
My Pi 4 isn't cutting it for the MAME titles I want to be running atm, Gauntlet Legends and Dark Legacy. I may use this as an upgrade for that box.
 
My Pi 4 isn't cutting it for the MAME titles I want to be running atm, Gauntlet Legends and Dark Legacy. I may use this as an upgrade for that box.

Same. Bought the Pi4 for MAME, but quickly found out it doesn't have the horsepower to run Killer Instinct 1 or 2 or many others like even PS1 with Shaders, was kind of disappointed.

Currently doing MAME & other emulation on an old i5 2500K with a GTX 980, but I'd rather have lower power draw. Hopefully these 4000 series APU's will fit the bill.
 
Same. Bought the Pi4 for MAME, but quickly found out it doesn't have the horsepower to run Killer Instinct 1 or 2 or many others like even PS1 with Shaders, was kind of disappointed.

Currently doing MAME & other emulation on an old i5 2500K with a GTX 980, but I'd rather have lower power draw. Hopefully these 4000 series APU's will fit the bill.
Here's hoping, I was recently excited by that AMD based SBC that has been circulating around but I finally managed to track down pricing and it was more than double what I was expecting so it was a big let down. The Cabinet I built doesn't have a lot of room inside it for anything too large so I was thinking of trying to track down one of the supermicro embedded boards or something. MAME is 90% CPU rendered so even an Intel onboard can do a good part.
 
A CPU with 8/16 cores is about twice as fast as one with 4/8 cores. 🙄

That's the sort of clickbait headline I'd expect to see on WCCF not Toms. :facepalm:

I suppose that the increased corecount doesn't badly bottleneck on RAM bandwidth is mildly newsworthy; but it's also old news since mainstream 8 core 2 channel processors have been out for a while.
 
IGPs can't get much faster until RAM gets faster or they use something else faster than RAM like a l4 cache or HBM
 
if you start having dedicated video ram, then you technically no longer have an iGPU, but a discrete gpu integrated into a single package, basically a SoC.

I'm all for that, but adding vram may start pushing the cost of the part out of the price range of the audience that buys apu's. I'm sure the market for high performance apu but not willing or able to do a conventional cpu + gpu setup is very tiny. At least for desktops.

That kind of setup for laptops though would probably be nice. I hate how every good amd laptop has to be married to an nvidia gpu for any decent graphics performance. It's the only reason why I haven't purchased a new laptop in basically a decade. I only want an amd laptop and it has to be amd graphics so i dont have to muck around with proprietary nvidia drivers and so far, the only all-amd options are low end.
 
IGPs can't get much faster until RAM gets faster or they use something else faster than RAM like a l4 cache or HBM

DDR5 should be available on consumer systems sometime in the next few years (servers next year, then trickle down) which should give a nice 2x bump to IGP performance; and double the number of cores a 2 channel memory system can feed when not using an IGP for gaming.
 
DDR5 should be available on consumer systems sometime in the next few years (servers next year, then trickle down) which should give a nice 2x bump to IGP performance; and double the number of cores a 2 channel memory system can feed when not using an IGP for gaming.
There is a pretty big difference in the performance characteristics of DDR and GDDR memory types, I am not so sure that the increase in ram speeds from DDR4 to DDR5 is going to scale as expected. It will help but I don't expect it to be overly stellar.
 
There is a pretty big difference in the performance characteristics of DDR and GDDR memory types, I am not so sure that the increase in ram speeds from DDR4 to DDR5 is going to scale as expected. It will help but I don't expect it to be overly stellar.

For a given amount of bandwidth, the more data streaming optimized setup of GDDR will be better; but when your GPU cores spend most of their time twiddling their thumbs waiting for data doubling the bandwidth will still be a major performance bump; which is why bottom tier discete GPUs always upgrade to the next tier of DDR as soon as their hardware is refreshed. DDR5 splitting each channel in half to allow more parallel accesses might even allow more scaling that the raw bandwidth increase offers. The primary goal is better scaling for many core CPUs; but presumably it should also give GPUs a bit of a bump even though the absolute performance gap between DDR and GDDR/HBM will leave IGPs far behind indefinitely.
 
Back
Top