New PC running benchmarks slower than expected (or hoped for!)

A

andymill

Member
Hi
Not sure whether this goes in technical support or desktop section. I received a my PC in July (see below). It's not had any software issues (no BSOD etc) but I've noticed it seems a tad slower than some benchmarks would suggest a 5900X with 3070 should be.

On graphics benchmarks like Firestrike, I'm getting about 5% worse than the average for my build. On Timespy, 4% worse. It's similar on some games. Frame rates a few percent worse than expected. Not a big deal. It's more more on CPU type benchmarks (Passmark, Geekbench, CPU-Z etx) where the difference is more substantial at over 10%. The worst is Geekbench 5 which is giving me 14% worse than for a typical 5900X.

As far as I can see PCS has set the machine up fine. They haven't enabled PBO or anything like that but the memory is running at DOCP 3600Mhz etc.

I suppose I'm wondering if the setup is somehow a sub-optimal in some way I cannot see or perhaps there is something a bit wrong with my hardware. Perhaps throttling for some reason. This is not my first custom built machine and prior ones haven't shown this much variation. Windows is a clean install.

Any obvious ideas?

Case
CORSAIR 4000D AIRFLOW TEMPERED GLASS GAMING CASE
Processor (CPU)
AMD Ryzen 9 5900X 12 Core CPU (3.7GHz-4.8GHz/70MB CACHE/AM4)
Motherboard
ASUS® ROG STRIX X570-F GAMING (USB 3.2 Gen 2, PCIe 4.0) - ARGB Ready!
Memory (RAM)
32GB Corsair VENGEANCE DDR4 3600MHz (2 x 16GB)
Graphics Card
8GB NVIDIA GEFORCE RTX 3070 - HDMI, DP
1st M.2 SSD Drive
500GB SAMSUNG 980 PRO M.2, PCIe NVMe (up to 6900MB/R, 5000MB/W)
1st M.2 SSD Drive
1TB SEAGATE FIRECUDA 520 GEN 4 PCIe NVMe (up to 5000MB/R, 4400MB/W)
Power Supply
CORSAIR 750W RMx SERIES™ MODULAR 80 PLUS® GOLD, ULTRA QUIET
Power Cable
1 x 1 Metre UK Power Cable (Kettle Lead)
Processor Cooling
Corsair H100i RGB PRO XT Hydro Series High Performance CPU Cooler
Thermal Paste
STANDARD THERMAL PASTE FOR SUFFICIENT COOLING
Sound Card
ONBOARD 6 CHANNEL (5.1) HIGH DEF AUDIO (AS STANDARD)
Network Card
10/100/1000 GIGABIT LAN PORT
Wireless Network Card
NOT REQUIRED
USB/Thunderbolt Options
MIN. 2 x USB 3.0 & 2 x USB 2.0 PORTS @ BACK PANEL + MIN. 2 FRONT PORTS
Operating System
NO OPERATING SYSTEM REQUIRED
Anti-Virus
NO ANTI-VIRUS SOFTWARE
Browser
Firefox™
Warranty
3 Year Silver Warranty (1 Year Collect & Return, 1 Year Parts, 3 Year Labour)
 
A

andymill

Member
Sorry slow to get back. Currently travelling. Don't have the links but I am recording them in Excel

TimeSpy: 13,009 (Graphics 13,348, CPU 11,665) vs. 13,595 average for 5900X/3070/32GB
Firestrike: 27,787 (Graphics, 31,056, Physics 34,635, Combined 13,268) vs. 29,309 average

So these are only 4-5% below their averages. I've got a Zotac 3070 so I assume others have better quality over overclocked cards, skewing the average up.

Geekbench 5: 12,040 vs 14,247 average
CPU-Z: 12,240 vs, 14,247 average for multi-core, 1,606 vs. 1,673 single core.
Passmark memory test: 3,195 vs. 3,618 for average of 3600MHz CL18

The Geekbench result is 15% below the average, CPU-Z 14% on multi-core but only 4% on single core. Passmark memory test is 12% below average. These are seem a bit more noticeable than the 3DMark tests. So seems it might be CPU or memory related.

@Martinr36: Yes, of course I ran the updates. I've clean installed twice so I really don't think it's OS related. It might be something in my BIOS but I've not touched that since receiving it.
 
Scott

Scott

Behold The Ford Mondeo
Moderator
We really need the links as there is a lot more information in the 3DMark report than just the scores. A simple check is to see whether it's the GPU score or the CPU score that's falling short.
 
D

DarTon

Well-known member
@andymill. I don't think it's your GPU or your CPU. I think it's your memory. Your Timespy CPU score is a bit of a tell but the Passmark memory score is really quite low for someone with 3600Mhz RAM, even if it's CL18.

Download Thaiphoon Burner to get a precise details on your memory DIMMs. Alternatively just grab the SKU and rank from CPU-Z.

Getting a sudden feeling of deja vu.
 
A

andymill

Member
Back home. According to CPUID/CPUZ, the DRAM frequency is 1800 (I assume that is somehow doubled to 3600 since dual channel). The CAS latency is 18, RAS to CAS 22, RAS to precharge 22 etc. Command rate is "1T"

How do I get the "SKU" and rank? I can't see it on the memory section. The memory is Corsair Vengeance LPX 3600MHz. Will Thaiphoon Burner do anything to my system? Never heard of it.
 
D

DarTon

Well-known member
It's on the tab called "SPD". You will have to choose a slot (1,2,3,4). PCS will have put the ram in slots 2 and 4 and it should bring up some details. Alternatively just take a look at the physical RAM in the case. It will have the SKU (Stock Keeping Unit code) and also a version number. Might be hard to see.

Thaiphoon is an app that can modify SPD firmware but all you want is for it to read what you have. It will give you a bunch of details.
 
Martinr36

Martinr36

MOST VALUED CONTRIBUTOR
andymill said:
Back home. According to CPUID/CPUZ, the DRAM frequency is 1800 (I assume that is somehow doubled to 3600 since dual channel). The CAS latency is 18, RAS to CAS 22, RAS to precharge 22 etc. Command rate is "1T"

How do I get the "SKU" and rank? I can't see it on the memory section. The memory is Corsair Vengeance LPX 3600MHz. Will Thaiphoon Burner do anything to my system? Never heard of it.
Click to expand...
If you're back home can you run firestrike and post the link
 
S

scarpa1

Enthusiast
Have you changed any Nvidia control panel settings?

If not a few of those add a good number of points to 3d mark scores which the scores you are comparing to will probably have set.

Also make sure G-Sync is off
 
SpyderTracks

SpyderTracks

We love you Ukraine
andymill said:
Back home. According to CPUID/CPUZ, the DRAM frequency is 1800 (I assume that is somehow doubled to 3600 since dual channel). The CAS latency is 18, RAS to CAS 22, RAS to precharge 22 etc. Command rate is "1T"

How do I get the "SKU" and rank? I can't see it on the memory section. The memory is Corsair Vengeance LPX 3600MHz. Will Thaiphoon Burner do anything to my system? Never heard of it.
Click to expand...
DDR RAM stands for double data rate, most tools give you the single data rate speed, so yes, yours is running at 3600MHz.

So mine you can see is at 800MHz which is a 1600MHz kit:

1629402143259.png


But yes, easiest way to get the SKU is through something like CPU-Z (the part number under the SPD tab)

But remember, these will most likely be OEM part numbers so will be harder to search online.
 
Last edited:
A

andymill

Member
@DarTon. Here is what Thaipohoon says

Thaiphoon - Copy.jpg


The SKU is CM4X16GD3600C18K4D. It's also written on the actual dimms. Not what I was expecting since the 3600Mhz memory from Corsair, according to their website, is CMK32GX4M2D3600C18. So what have I got and why is that an issue?
 
D

DarTon

Well-known member
@andymill. There is nothing odd about this RAM. The only thing that stands out is that it's single rank. These days you can get both single rank and dual rank under the same Corsair SKU though.

If we look at average scores on Passmark memory test for CM4X16GD3600C18K4D we see it's 3225.

PassMark - Corsair CM4X16GD3600C18K4D 16GB - Price performance comparison

www.memorybenchmark.net

If we look at the two retail SKUs for 3600Mhz 18-22-22-42 kits (Z is "Ryzen Optimized") then we get score of 3618 and 3748, so 12-15% better.

PassMark - Corsair CMK32GX4M2D3600C18 16GB - Price performance comparison

www.memorybenchmark.net

PassMark - Corsair CMK32GX4M2Z3600C18 16GB - Price performance comparison

www.memorybenchmark.net

It's the "Latency" number that is poor. Your SKU only has 5 samples, though, so significant small sample bias. It could be that the averages you see for the retail SKU are still dominated by dual rank kits from prior periods while this new SKU is only single rank kits.

This is the same RAM as I received, hence my deja vu comment. I was annoyed with PCS since I'd bought the RAM on condition it was dual rank. So I got a refund and bought my own - a standard Corsair retail SKU for £140, dual rank, same timings. Once fitted all my benchmark numbers jumped. 1500+ points on Firestrike, 500 on Timespy, 2000 on Geekbench etc. Everything back to "average" (tbh a totally meaningless term since you never compare apples with apples).

So all I know is that, empirically, I switched and got an instant jump. I'm blaming it on dual vs. single rank issue but that's just a guess really.
 
A

andymill

Member
@DarTon. Sorry not to respond but I wanted to check this by changing the RAM. Borrowed a friends 2x16gb ram (also Corsair) and instant improvement in all benchmark results and even in some games. Quite noticeable. Seems RAM is the cause.

Not sure what to do here. Paid PCS well over £200 for this ram and feeling like I got scammed. The ram I used from my friend is just £130 from another seller and only 3200Mhz. How can PCS justify charging this much for 3600Mhz ram that performs 10% worse than 3200Mhz?
 
SpyderTracks

SpyderTracks

We love you Ukraine
andymill said:
@DarTon. Sorry not to respond but I wanted to check this by changing the RAM. Borrowed a friends 2x16gb ram (also Corsair) and instant improvement in all benchmark results and even in some games. Quite noticeable. Seems RAM is the cause.

Not sure what to do here. Paid PCS well over £200 for this ram and feeling like I got scammed. The ram I used from my friend is just £130 from another seller and only 3200Mhz. How can PCS justify charging this much for 3600Mhz ram that performs 10% worse than 3200Mhz?
Click to expand...
If the RAM timings were important you'd just source your own, that's the beauty of buying a custom rig.
 
You must log in or register to reply here.
Top