• 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.

Raptor Lake Leaks + Intel 4 developments

Puget's review/preview shows overall the Zen 3 Threadripper PRO line is still better than Intel's Sapphire Rapids Xeon. To be honest this was expected, considering the all core clock speeds of the Xeons being limited, and these are server parts for desktop, with servers being AMD's strength. At least there are more options though on the market.

Puget have reviewed the new Xeons here: https://www.pugetsystems.com/labs/articles/intel-xeon-w-3400-content-creation-preview/
And typical multi-threaded workloads reflect the performance vs Threadripper PRO:


b3iXd8n.png


CCcqZAg.png
 
Last edited:
That's kinda dissapointing, if they're not going to beat, or beat by much, the existing Zen3 TR then what's the impetus on AMD to get around to releasing the Zen4 TR goodness, or even non-pro TRs :(
 
SKILL said:
That's kinda dissapointing, if they're not going to beat, or beat by much, the existing Zen3 TR then what's the impetus on AMD to get around to releasing the Zen4 TR goodness, or even non-pro TRs :(
Click to expand...
There are things that Intel have that AMD don't that although Intel is behind in performance would be good to see AMD sort for their latest as per their expected latest lineup come September. I don't expect we will see non-pro though tbh still.

This is what AMD are missing or need to add to match the feature spec of Intel. Workflow that comes from larger memory bandwidth and more PCIe lanes seems to be the biggest shortfall we may see in the medium/longer term.
  • DDR5 RAM support
  • 4TB of support to RAM rather than 2TB (via 8 channels rarther than 4)
  • Double the PCIe Lanes
  • Higher boost frequency
  • Cheaper
  • Appart from the Pro 5995WX all Intel CPUs have greater L3 Cache
Stagnation isn't good and it seems like there could be some solid performance improvement with the newer cores also for Threadripper then. I do expect the power profile will significantly increase
Bencher said:
His numbers are from the wall, im pretty sure the cpu doesn't draw 500w stock. It's literally impossible since it's locked (at stock) to 350w. P cores are more efficient than zen 3 in cinebench R23.
Click to expand...
It isn't locked in that it will drawer more, for instance as per Puget system at stock it still pulled 374w. So well yes it is aiming for 350w, it clearly will have some tolerance.

In the video running R23 at stock settings it pulled 514.684w so unless Hardware info is miss-reporting that is a good deal higher than expected but they were burst power draw between 400-500w, still well above the claimed 350w. It is also really bad at ideal compared to TR CPUs. Nothing to do with numbers from the wall though.
 
Curlyriff said:
isn't locked in that it will drawer more, for instance as per Puget system at stock it still pulled 374w. So well yes it is aiming for 350w, it clearly will have some tolerance.

In the video running R23 at stock settings it pulled 514.684w so unless Hardware info is miss-reporting that is a good deal higher than expected but they were burst power draw between 400-500w, still well above the claimed 350w. It is also really bad at ideal compared to TR CPUs. Nothing to do with numbers from the wall though.
Click to expand...
In the video you are talking about you can see pretty clearly he manually removed the power limits. You see it right below the power draw. So he's not running stock.

Regarding puget, if they measure from the cables the difference is from the VRM efficiency. For Intel PL2 = Power limit, it can't draw more than that
 
Last edited:
Bencher said:
In the video you are talking about you can see pretty clearly he manually removed the power limits. You see it right below the power draw. So he's not running stock.

Regarding puget, if they measure from the cables the difference is from the VRM efficiency. For Intel PL2 = Power limit, it can't draw more than that
Click to expand...
No at the end when doing R23 he is on a completely different system to when he was doing the overclock and the footage with Intel prior.

The last one he shows it pulling 400-500w showing it running stock at 2.9Ghz turbo.
So either the power limits are removed and it does absolutely nothing or its at stock because it locked to 2.9Ghz and nothing more and still pulled 400watt +.
 
Last edited:
Curlyriff said:
No at the end when doing R23 he is on a completely different system to when he was doing the overclock and the footage with Intel prior.

The last one he shows it pulling 400-500w showing it running stock at 2.9Ghz turbo.
So either the power limits are removed and it does absolutely nothing or its at stock because it locked to 2.9Ghz and nothing more and still pulled 400watt +.
Click to expand...
Υοu are talking about 14:10 on the video right? You can tell right there that he has the limits removed
 
Bencher said:
Υοu are talking about 14:10 on the video right? You can tell right there that he has the limits removed
Click to expand...
He litereally says if you check the clock speed stock it is only 2.9Ghz and shows it, the power limits might be off but it doesn't seem to be doing anything else with it at 2.9Ghz still showing to be pulling 516.375w at the time of showing that so if it does overclock then it doesn't appear to be doing anything that is suggested to be doing at 350w other than forcing the power to be higher then?

Maybe he miss-spoke but saying at stock it is 2.9Ghz, showing it runing maxed out at 2.9Ghz I don't really see it being an OC of any sort. Maybe the power limit is moved up because on the PL1 350w the cores can't maintain the 2.9Ghz boost and would throttle for some reason? I can only go from what was stated in the video though.
 
Curlyriff said:
He litereally says if you check the clock speed stock it is only 2.9Ghz and shows it, the power limits might be off but it doesn't seem to be doing anything else with it at 2.9Ghz still showing to be pulling 516.375w at the time of showing that so if it does overclock then it doesn't appear to be doing anything that is suggested to be doing at 350w other than forcing the power to be higher then?

Maybe he miss-spoke but saying at stock it is 2.9Ghz, showing it runing maxed out at 2.9Ghz I don't really see it being an OC of any sort. Maybe the power limit is moved up because on the PL1 350w the cores can't maintain the 2.9Ghz boost and would throttle for some reason? I can only go from what was stated in the video though.
Click to expand...
You are correct, at stock 350w PL the ccores cant maintain 2.9ghz boost. What he meant by stock is that the cores are running at stock VF curve, and their maximum all core clockspeed is 2.9ghz.

The same way for example the 13900k has a stock all core of 5.5ghz but you can't get that unless you remove the power limits

TLDR : His score wouldn't be 70k and the cpu wouldnt be drawing 500w if he didn't remove the limits. I think it scores 55-58k @ 350watts or something
 
Last edited:
Bencher said:
You are correct, at stock 350w PL the ccores cant maintain 2.9ghz boost. What he meant by stock is that the cores are running at stock VF curve, and their maximum all core clockspeed is 2.9ghz.

The same way for example the 13900k has a stock all core of 5.5ghz but you can't get that unless you remove the power limits

TLDR : His score wouldn't be 70k and the cpu wouldnt be drawing 500w if he didn't remove the limits. I think it scores 55-58k @ 350watts or something
Click to expand...
Okay cool so the new Intel CPU at stock power limit is bogging down and thus why worse than TR-Pro parts whilst also being more power hungry for less cores than even last generation AMD cores? I assume these new Intel CPUs are same cores as the 13 series desktop parts generation wise?

So other than if needing PCIe lanes and double the memory throughput then they really aren't worth it. You would need to unlock the power limit to actually get to a similar performance to current gen TR-Pro so almost double the power drawer. They don't really seem all that appealing generally for most HEDT setups then I would have thought.
 
Curlyriff said:
Okay cool so the new Intel CPU at stock power limit is bogging down and thus why worse than TR-Pro parts whilst also being more power hungry for less cores than even last generation AMD cores? I assume these new Intel CPUs are same cores as the 13 series desktop parts generation wise?

So other than if needing PCIe lanes and double the memory throughput then they really aren't worth it. You would need to unlock the power limit to actually get to a similar performance to current gen TR-Pro so almost double the power drawer. They don't really seem all that appealing generally for most HEDT setups then I would have thought.
Click to expand...
Alder lake or golden cove cores for sapphire rapids. Raptor cove = raptor lake.

Also hedt uses mesh as ring doesn’t scale to high core counts. Mesh is notably slower that’s why desktop is much better for gaming.

Voltage wise hedt uses fivr while desktop uses svid.

The holy grail would be fivr on desktop eventually but keeping ring.
 
Last edited:
Curlyriff said:
Okay cool so the new Intel CPU at stock power limit is bogging down and thus why worse than TR-Pro parts whilst also being more power hungry for less cores than even last generation AMD cores? I assume these new Intel CPUs are same cores as the 13 series desktop parts generation wise?

So other than if needing PCIe lanes and double the memory throughput then they really aren't worth it. You would need to unlock the power limit to actually get to a similar performance to current gen TR-Pro so almost double the power drawer. They don't really seem all that appealing generally for most HEDT setups then I would have thought.
Click to expand...
If you are talking specifically about cbr23, since that's what der tested, I think these xeons are faster at same core counts than zen3. Have no idea about pricing though. The best part about these xeons is that finally, we will have cheap 64+ pcie lanes with just the entry level 349$ model, a market segment amd completely ditched, since now you need to play a couple of thousands to get lanes
 
Bencher said:
If you are talking specifically about cbr23, since that's what der tested, I think these xeons are faster at same core counts than zen3. Have no idea about pricing though. The best part about these xeons is that finally, we will have cheap 64+ pcie lanes with just the entry level 349$ model, a market segment amd completely ditched, since now you need to play a couple of thousands to get lanes
Click to expand...
I was on about the results from the test bench we've got linked generally from Puget and how it seems Intel still got a way to go for general workloads you'd expect a HEDT system to do.

In regard to R23 the 64core TR-Pro it scored 65-66k. If you took it as 56core that would be 57k then approx. So pretty similar all in then on that side of things.
 
Robert896r1 said:
Alder lake or golden cove cores for sapphire rapids. Raptor cove = raptor lake.

Also hedt uses mesh as ring doesn’t scale to high core counts. Mesh is notably slower that’s why desktop is much better for gaming.

Voltage wise hedt uses fivr while desktop uses svid.

The holy grail would be fivr on desktop eventually but keeping ring.
Click to expand...
Ta yeah I just wasn't sure which core being used. Useful info for future use generally though.
 
Last edited:
Been meaning to do this for a while so got some time today. This is a what I'd personally call a safe OC for those on RPL and DDR4 BDIE. You can load the CMO file onto a usb and import on Asus via the bios. The text file can help those not on Asus or those who just want to browse the settings.

The baseline is intended to work on any RPL+DR BDIE combo and I made sure it was rock stable. Another requirement was not actively cooling the ram with a fan or other methods and just use the natural airflow of the case. Most RPL systems will be able to reduce the vcore further and still be equally stable but I wanted a very safe number to start from.

Stability tested:
image.png


CMO file for import into asus bios: https://drive.google.com/file/d/11nK-_oDcwQdV_BnZtWm_GFcSgFUOugwt/view?usp=sharing

Txt file of the settings: https://docs.google.com/document/d/1KpuEQXhZBfUobm2rPbjfEdIlUyjJsmMNnTO0iKgw3fk/edit?usp=sharing
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom