• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Poll: Ryzen 7950X3D, 7900X3D, 7800X3D

Will you be purchasing the 7800X3D on the 6th?

  • Total voters
    191
  • Poll closed .
Iron Man said:
I want to see some benchmarks on release where Intel memory is very poorly optimised so I can see a huge lead for the 7950x3D to justify my purchase.
Click to expand...

If it’s xmp and running out of the box as you see in techtuber videos, it’s poorly optimised. The faster the kit, the looser the subtimings so it’s got a higher chance of booting and not stressing the imc as much.

on ddr5, subtimings matter a lot, trrd/tfaw/trfc/trefi/tcwl more so than tightening down primaries.
 
Last edited:
humbug said:
I'm not sure that's unique to Ryzen, look at the post @LtMatt just made, there isn't much difference between memory speeds, no more than Intel.

If you use memory that isn't on the motherboard QVL list it defaults to extremely lose timings, i know this because my kit is not on my boards QVL, even at the XMP 32000MT/s the sub timings were 2X has high as i tuned them at 3800MT/s, TRFC for example was at 980, utterly ridiculous, i cut that down by more than half, this at 3800MT/s, i did the same for all the sub timings, i cut some by more than half.

I gained 20 to 30% in games by doing that, much more than the usual 10% you gain if the memory is on the QVL where they apply proper timings, i don't think Intel would be any different.

So its really easy, be it deliberately or by genuine ignorance to make any of these CPU's, be it Intel or AMD, look much worse in comparison than they actually are, all you have to do is use memory that isn't on one or the other's QVL, and with that you have plausible deniability if you get rumbled, which is very unlikely anyway.

People do do this, Ryan Shrout was notorious for manipulating his results in ways like this, he now works for Intel.
Click to expand...
That was my point, in Horizon Zero Dawn there isn't a difference in memory scaling between the two brands, however, that's an outlier for AMD as most games in that comparison show a far greater drop with memory speed reduction and timing.
 
LtMatt said:
I am curious to see how much faster Zen 4 can get in Horizon Zero Dawn. :eek:
RsKvBnb.png [/spoiler
Click to expand...
I posted some Horizon Zero Dawn 1080P numbers with the 7950X (SMT Off) and the 7900 XTX earlier in this thread using the favour quality preset with 6000C28 and achieved 273 CPU average FPS so in line with but also a little higher than what HUB got.
I3n6MZ7.jpg
When my X3D arrives, I will retest using the exact same memory speed and timings so we can measure the performance difference.
 
Last edited:
Curlyriff said:
Huh there is 8fps for Intel max and if you only compare the same RAM timing ranges as AMD max out at then only 4fps. There is a 20fps increase for AMD.

So that is 2% for Intel in the increase and 8.5% increase for AMD. So suddenly if you are saying Intel is 10% faster because you are using worse RAM and timings that are not giving performance expected you could actually only be more like 2-3% off Intel in those games if they gain that same 8-10% FPS increase by using decent RAM. All those RAM options with timing are supported via QVL listing so it is not just as long is it on list you are fine.
Click to expand...

4800MT/s > 7200MT/s Intel gain 4% (197 v 205)

4800MT/s > 6000MT/s AMD gain 9% (226 v 245)

Ok, granted.
 
LtMatt said:
I posted some Horizon Zero Dawn 1080P numbers with the 7950X (SMT Off) and the 7900 XTX earlier in this thread using the favour quality preset with 6000C28 and achieved 273 CPU average FPS so in line with but also a little higher than what HUB got.
I3n6MZ7.jpg
When my X3D arrives, I will retest using the exact same memory speed and timings so we can measure the performance difference.
Click to expand...

If you're running SMT off it probably explains the difference, weirdly after all these decades SMT can still have an effect on performance.

I trust Steve Walton, he is competent and smart, one of few i trust, the other two are Steve Burke and Dr Ian Cutress at Anand.
 
Last edited:
alec said:
formerly at Anand, sadly

and he was always using memory as per spec (so ddr5-5200 for zen4, ddr4-3200 for zen3), and anandtech are still doing that. I pretty much ignore their CPU reviews now.
Click to expand...
That is a shame, we are losing a lot of the "geek or nerd" aspect of reviews that people like him bring.

I'm ok with the use of XMP rated memory in reviews, as long as they do the same for Intel, which i think they do? Honestly i don't have a problem with memory being set at rated values, i think most people would actually run it like that.
 
Last edited:
Robert896r1 said:
If it’s xmp and running out of the box as you see in techtuber videos, it’s poorly optimised. The faster the kit, the looser the subtimings so it’s got a higher chance of booting and not stressing the imc as much.

on ddr5, subtimings matter a lot, trrd/tfaw/trfc/trefi/tcwl more so than tightening down primaries.
Click to expand...
The 7200 kits I’ve had have not been able to overclock any higher but the timings were very loose and the voltages were 1.45 at XMP. I managed to bring the latency down considerably from mid 60’s to mid 50’s and much improve performance as well as drop the voltage to 1.35. I’m thinking about getting an Apex and some 8000 memory, but first I am looking for a pre tested good example 13900k/ks. The 13900k’s I’ve had haven’t been great unfortunately.
 
Harshell123 said:
The 7200 kits I’ve had have not been able to overclock any higher but the timings were very loose and the voltages were 1.45 at XMP. I managed to bring the latency down considerably from mid 60’s to mid 50’s and much improve performance as well as drop the voltage to 1.35. I’m thinking about getting an Apex and some 8000 memory, but first I am looking for a pre tested good example 13900k/ks. The 13900k’s I’ve had haven’t been great unfortunately.
Click to expand...

You don't need a super mem bin like that for Apex. Just about any adie bin will stabilize 8000mhz daily tightened. RPL-Refresh will have CKD support so mem should see a big gain over what's possible today. If you saw Xaver's testing, you can see the gains you might get going to 8000. https://www.igorslab.de/en/basics-h...ics-and-ddr4-g1-in-synthetics-and-no-games/9/ The gain he's getting from 7000 to 7800, you'd get from 7200 to 8000. Upto really. But sometimes new toys are just fun to play with!

But yeah, the higher you go up with frequency on xmp kits, the looser the timings.
 
Harshell123 said:
The 7200 kits I’ve had have not been able to overclock any higher but the timings were very loose and the voltages were 1.45 at XMP. I managed to bring the latency down considerably from mid 60’s to mid 50’s and much improve performance as well as drop the voltage to 1.35. I’m thinking about getting an Apex and some 8000 memory, but first I am looking for a pre tested good example 13900k/ks. The 13900k’s I’ve had haven’t been great unfortunately.
Click to expand...
I'm running my 7600 kits at 7000 cause my 12900k doesn't play ball. 7000c30 compared to 7600c34 on my 13900k. Anything above 1.5v spits errors
 
When is the refresh due?

Yes it’s not so much the gains, I just enjoy the process. I’ve pretty much tweaked as much as I can with my current hardware :D. I’m selling one of my 13900k and will be getting a 7950x3d to test, although I doubt I will be able to play as much with that.

I was hoping a better bin would push beyond 8000 as I have read that some have pushed higher on air alone.
 
SilverCider said:
That was my point, in Horizon Zero Dawn there isn't a difference in memory scaling between the two brands, however, that's an outlier for AMD as most games in that comparison show a far greater drop with memory speed reduction and timing.
Click to expand...
There is more than double the scaling with AMD than Intel in that title as shown in the image posted.

As humbug pointed out;
humbug said:
4800MT/s > 7200MT/s Intel gain 4% (197 v 205)

4800MT/s > 6000MT/s AMD gain 9% (226 v 245)
Click to expand...

This to me shows Zen4 is held back by it's memory controller, so it bodes well for a decent uplift with the 3D cache versions. More cache, less going to main memory.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom