• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Discussion Ryzen 3000 series benchmark thread ** Open **

Page 30 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Hans Gruber

Senior member
Dec 23, 2006
942
262
136
@Hans Gruber

I think Samsung will forgive you.
I wrongly figured that all the SATA 3 ports were the same. I ran benchmarks and they were identical. Meaning the SATA 3 ports were throttled. I looked up ASmedia SATA 3 and complaints about their ports being slow were everywhere. I switched to the other set of SATA 3 ports and the speeds are back where they should be.

I've never experienced such bad speeds from SATA 3 ports before.
 

VirtualLarry

Lifer
Aug 25, 2001
48,829
5,326
126
I wrongly figured that all the SATA 3 ports were the same. I ran benchmarks and they were identical. Meaning the SATA 3 ports were throttled. I looked up ASmedia SATA 3 and complaints about their ports being slow were everywhere. I switched to the other set of SATA 3 ports and the speeds are back where they should be.

I've never experienced such bad speeds from SATA 3 ports before.
As a general rule-of-thumb, boot devices like SATA SSDs go on SATA #0, which should be a "native" SATA port, part of the main chipset. Add-on (connected via PCI-E) SATA port controllers (ASMedia, JMicron, Marvell, Via), have traditionally never quite performed as well as the native SATA ports, Intel or AMD platform. Also, historically, Intel native SATA6G ports have been slightly faster than AMD native SATA6G ports, but at that point, you're mostly splitting hairs, especially if you have a less-than-top-of-the-line SATA SSDs. (Basically any SATA SSD other than a Samsung, and the most recent Intel SATA SSDs.)

Part of that is how that they are wired up via PCI-E (the add-on controllers), it has been largely the case, at least with add-on cards with those controller chips, using a PCI-E x1 lane connection (I forget if PCI-E 2.1 or 3.0/3.1), but it's basically limited to 500MB/sec each way, and each SATA port can theoretically handle 600MB/sec one way, so there's kind of a "bandwidth haircut" going on there too, and it wouldn't surprise me much if the add-on controller chips were wired and limited similarly.
 

Hans Gruber

Senior member
Dec 23, 2006
942
262
136
The max read speed I have seen with a SATA 3 SSD is 560MB. Those are my Samsung 850evo numbers. The userbenchmark numbers are low for all the SSD's. But it is apples to apples.
 

psolord

Golden Member
Sep 16, 2009
1,290
326
136
I wrongly figured that all the SATA 3 ports were the same. I ran benchmarks and they were identical. Meaning the SATA 3 ports were throttled. I looked up ASmedia SATA 3 and complaints about their ports being slow were everywhere. I switched to the other set of SATA 3 ports and the speeds are back where they should be.

I've never experienced such bad speeds from SATA 3 ports before.
There's a reason they are called asmedia. However there's one s missing!
 

Asterox

Senior member
May 15, 2012
376
422
136

Techpowerup about "1usmus custom power plan for Ryzen R9 3900X."




Posting links or images without comments is not allowed.


esquared
Anandtech Forum Director
 
Last edited:

Abwx

Diamond Member
Apr 2, 2011
9,100
859
126
3950X review at Computerbase, resuilts are more or less what was expected, for the ock crowd it s worth pointing that they got 10% frequency overclock on all cores.

 
  • Like
Reactions: lightmanek

TheGiant

Senior member
Jun 12, 2017
695
295
106
I did, but that freq means povray testing and I cant find in the article of the power numbers are for that testing
I dont see any extra binning except it can reach single core of 4,7GHz, for a hefty increase in single core power from 18 to 24W for 100 MHz
my guess on the clocks is
the cinebench allmighty gives 9228 vs 7220 score, so 7720/12*16=9626 and 9626/9928=1.043 so the freq is about 4.3% lower
so with that freq/voltage curve this means nothing special, just standard board regulation for power limit

however this is new level of performance one can get on standard desktop, but for who, for me my 3900X is a big overkill

I still think AMD should place the price higher, they are not using their chance

as for TDP,


most of thise CPUs in the chart (except HEDT) are 91,95 or 105 W TDP
TDP=new internet joke of this age
 

Abwx

Diamond Member
Apr 2, 2011
9,100
859
126
the cinebench allmighty gives 9228 vs 7220 score, so 7720/12*16=9626 and 9626/9928=1.043 so the freq is about 4.3% lower
so with that freq/voltage curve this means nothing special, just standard board regulation for power limit
4.3% lower frequency should cut power by 15% with the frequency/voltage at play, so that s not enough to cope with 33% more cores, not counting that 16C will have higher peak currents than a 12C even at same power, wich mandate more voltage to compensate for bigger voltage drops.

So there s some binning that has been made surely during several months, a hint is that Computerbase.de sample hit 4.725MHz on single core tests, not sure that previous SKUs could hit such a value even when overclocked with higher voltages.
 
  • Like
Reactions: lightmanek

Hitman928

Platinum Member
Apr 15, 2012
2,773
2,124
136
most of thise CPUs in the chart (except HEDT) are 91,95 or 105 W TDP
TDP=new internet joke of this age
That's system power, not CPU power.

With that said, both AMD and Intel allow for sustained boost levels at stock as long as your cooling solution and power delivery can handle it. That means that both 95W intel chips and 105W AMD chips will use ~140W all core if your setup can handle it. Additionally, both AMD and Intel allow for essentially unlocked power limits which will obviously use more power for higher sustained boosts. Some motherboards (especially on the intel side as of late) will unlock these power limits by default so CPU power use will depend largely on the motherboard you choose. This behavior has evolved over time and I think both Intel and AMD need to be more transparent about it (like listing both a base TDP and boost TDP) but this is where the unfortunate lack of traditional technology scaling has brought us.
 
  • Like
Reactions: lightmanek

DrMrLordX

Lifer
Apr 27, 2000
16,371
5,281
136
not sure that previous SKUs could hit such a value even when overclocked with higher voltages.
My 3900x definitely can't. $750 is too rich for my blood, but I will say that anyone who waited this long for the 3950x won't be disappointed.
 
  • Like
Reactions: lightmanek

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
20,603
8,456
136
Where are the 3960 and 3970x reviews ? Wasn't that today also ?
 
  • Like
Reactions: Drazick

amrnuke

Senior member
Apr 24, 2019
844
1,027
96
The reviews show FPS in gaming gaining a lot on 9900KS. That's remarkable IMO for a 16C/32T chip whose focus was ostensibly to bridge the gap productivity-wise between mainstream and HEDT. These are well-binned. TechSpot's review of the productivity stuff - the 3950X just destroys everything.

If they have been able to scale this well so far, I think this bodes extremely well for the impending 3960, 3970 TR chips (and hopefully that most elusive of moarcoarbois, the 3990WX). That being said, I agree now more than ever with many on here that a 16C/32T TR chip, with that massive surface area for heat dissipation, would be a potential gaming king, maybe even hit that elusive 5 GHz mark on water.

Lastly, I wonder, as the process matures, if we are going to see some new 3x50 lines for budget gaming. I now wonder about a 3850X with 8C/16T at 4.7-4.8 GHz? Sounds pretty sweet. A 3650X with 6C/12T at 4.6-4.7 GHz would be pretty cool too.
 
  • Like
Reactions: lightmanek

lobz

Golden Member
Feb 10, 2017
1,345
1,505
106
I did, but that freq means povray testing and I cant find in the article of the power numbers are for that testing
I dont see any extra binning except it can reach single core of 4,7GHz, for a hefty increase in single core power from 18 to 24W for 100 MHz
my guess on the clocks is
the cinebench allmighty gives 9228 vs 7220 score, so 7720/12*16=9626 and 9626/9928=1.043 so the freq is about 4.3% lower
so with that freq/voltage curve this means nothing special, just standard board regulation for power limit

however this is new level of performance one can get on standard desktop, but for who, for me my 3900X is a big overkill

I still think AMD should place the price higher, they are not using their chance

as for TDP,


most of thise CPUs in the chart (except HEDT) are 91,95 or 105 W TDP
TDP=new internet joke of this age
You got so caught up in trying to prove your point that you forgot about the 16 cores vs the 12 in the 3900X. Nothing special, eh? OK.
 

Kocicak

Senior member
Jan 17, 2019
279
149
76
I cannot imagine any reason why the power consumption of the 3950X as a function of the cores used should have this weird characteristic, as measured by Anandtech. Why does it not go like I indicated with the green line?

3950X PowerLoading2.png
 

jpiniero

Diamond Member
Oct 1, 2010
8,288
1,355
126

AMD apparently is now ripping off Intel's low end core marketing with the Athlon Gold 3150U. Not sure if legit. Despite what it says it is likely Pinnacle Ridge based.
 

VirtualLarry

Lifer
Aug 25, 2001
48,829
5,326
126
AMD apparently is now ripping off Intel's low end core marketing with the Athlon Gold 3150U. Not sure if legit. Despite what it says it is likely Pinnacle Ridge based.
It could just be HP doing it, with AMD being complicit in setting the CPUID name string for a batch of OEM CPUs for them. Remember that thread about the Future Shop ad, with AMD A8 APUs in laptops, being sold as "10 Core". Maybe someone will do the favor of looking that one up. I believe that was also HP's doing.
 
  • Like
Reactions: lightmanek

ASK THE COMMUNITY