*Official Benchmark Thread* - Post it here or it didn't happen :D

[Mr. Fox]

18 minutes ago, ssj92 said:

Guys need some RAM tuning help. So XMP tries to boot but it reverts to 5200mhz. It is 5600mhz RAM. I am able to change all timings using the RU.EFI bios mod method so I wanted to ask what timings should I try? here's my hwinfo ss. Right now it's on 5200mhz 38-37-37-73 

 

 

Try the same settings with 1.250V or 1.300V.

  

19 minutes ago, Papusan said:

I think you need test out with different drivers. This bench is quite glitchy and very picky. And you need run several benches to get to try get an ok score. Loads of people have many weird problems with this benchmark. Maybe a fun bench for bro @ryan 🙂
 

The fun part. When you posted I run Heaven on the 3070 🙂

Nah. At the end of the day it's not worth the effort. Not going to burn any calories on broken HWBOT garbage software released in 2011, LOL. I'll try one more time on a different OS and see if it works. If not, then I will move on to something else. Works fine on the Dark with the 4090.

[johnksss]

5 hours ago, Mr. Fox said:

I was going to upload these, but it appears the Catzilla web site is down. I don't know if that is temporary or permanent. So, I can't upload them to HWBOT without validation links.

 

 

@Papusan or @johnksssis there a patch or something for the HWBOT version of Heaven benchmark? When I click the clicks to "take a screenshot" I get the error show, it closes and there is no file created to upload.

I just checked and Catzilla is up and running, but I did find that one of my oses would not allow me to upload to them.

 

And the same with Heaven. I had some setting turned off and it was blocking heaven.

[Papusan]

6 hours ago, Mr. Fox said:

 

 Nah. At the end of the day it's not worth the effort. Not going to burn any calories on broken HWBOT garbage software released in 2011, LOL. I'll try one more time on a different OS and see if it works. If not, then I will move on to something else. Works fine on the Dark with the 4090.

Did it work? Yep, Hwbot Heaven bench is a mess. And I'm not so sure it would be so much better if it was a few years newer, LOOL

 

6 hours ago, Papusan said:

The fun part. When you posted I run Heaven on the 3070 🙂

https://hwbot.org/submission/5242103_papusan_unigine_heaven___xtreme_geforce_rtx_3070_10080.34_dx11_marks?recalculate=true

 

https://hwbot.org/submission/5242102_papusan_unigine_heaven___basic_geforce_rtx_3070_11376.12_dx9_marks

 

https://hwbot.org/submission/5242198_papusan_3dmark___fire_strike_geforce_rtx_3070_36776_marks

 

https://hwbot.org/submission/5242201_papusan_3dmark11___performance_geforce_rtx_3070_44546_marks

 

On 3/31/2023 at 7:56 PM, Raiderman said:

I found brother @PapusanaRGB Crocs

Nice find brother Raiderman. But I don't think the RGB style is quite disgusting😆 Maybe more in the line what bro @electrosoftwant and expect from nowadays modern style?😅 

 

[Etern4l]

9 hours ago, Mr. Fox said:

I've never owned memory that had two XMP profiles from the manufacturer. I have added my own second profile on DDR4 using Thaiphoon Burner, but none I have ever purchased or seen available for purchase have more than one.

 

FYI Kingston Fury Beast 5600 DDR5 modules (32GB at least, and others too I think) have 3 profiles: 5600CL40. 5200CL40 and 4800CL38. This is very handy due to all the issues with 4 2R DIMM support, and AFAIK nobody on the planet has managed to run 128GB at 5600MHz with 12/13 gen. AMD seems to struggle even more, another reason I made the right choice by going with Intel. 

 

I'm currently running mine at 5000CL30-38 super stable apparently. The performance improvment from this + the new mobo + the delid is insane. +15-20% roughly in practice, feels like 14th gen. 

 

Now I'm looking at these new 48GB modules from Corsair. They come in 5600CL40 and 5200CL38 variants (2x48GB kit, 4x48GB is only available as 5200). I know I will only be able to run them at 5200 or 5000, however, I'm assuming that those 5600 modules (just $10 or so more expensive) are better binned and might be able to run at lower latencies. Also MSI demoed 192GB running at 5200, actually using those 5600 modules. 

 

I guess I should do things the hard way as usual, and get the 5600 modules, then manually set the basic latencies and voltage, and hope the board does a good job with the rest. I'm not tweaking 30 RAM timings by hand lol Good return policy will be essential. 

[Mr. Fox]

6 hours ago, johnksss said:

I just checked and Catzilla is up and running, but I did find that one of my oses would not allow me to upload to them.

 

And the same with Heaven. I had some setting turned off and it was blocking heaven.

I cannot access catzilla.com or allbenchmark.com from any of my computer web browsers. I get the "This site can't be reached" error whenever I try.

  

5 hours ago, Papusan said:

Did it work? Yep, Hwbot Heaven bench is a mess. And I'm not so sure it would be so much better if it was a few years newer, LOOL

Haven't tried yet. Not a huge priority. I don't expect to get enough  points from it to matter too much.

[Etern4l]

10 hours ago, ssj92 said:

Guys need some RAM tuning help. So XMP tries to boot but it reverts to 5200mhz. It is 5600mhz RAM. I am able to change all timings using the RU.EFI bios mod method so I wanted to ask what timings should I try? here's my hwinfo ss. Right now it's on 5200mhz 38-37-37-73 

 

 

 

2 sticks of Kingston 32GB would run fine at 5600, or a bit higher, on the desktop, but the default latencies would likely be 40-40-40-40 and forgot the last one, 80-sh, you can look it up on Kingston website. Try that and maybe higher voltages, althoughmy experience with Kingston DDR5 suggests this won't help here. Does it revert straight away or after a long period of training? Quick reversion would indicate some lower-level issue with the BIOS. 

 

 BTWII recall watching a video where a claim was made that RGB variants of modules tend to perform not as well. Have you guys had similar experiences? I don't care about RGB, but sometimes it's the only variant available.

 

@Mr. Fox @Papusan @tps3443 @Rage Set

[Papusan]

Windows 10 was up 11.62% while Windows 11 crashed -9.65% (Steam Survey). Latest and greatest knocked down with a fantastic punch @Mr. Fox

 

https://www.techspot.com/news/98166-major-changes-steam-survey-new-top-gpu-language.html
 

[Rage Set]

2 hours ago, Etern4l said:

 

2 sticks of Kingston 32GB would run fine at 5600, or a bit higher, on the desktop, but the default latencies would likely be 40-40-40-40 and forgot the last one, 80-sh, you can look it up on Kingston website. Try that and maybe higher voltages, althoughmy experience with Kingston DDR5 suggests this won't help here. Does it revert straight away or after a long period of training? Quick reversion would indicate some lower-level issue with the BIOS. 

 

 BTWII recall watching a video where a claim was made that RGB variants of modules tend to perform not as well. Have you guys had similar experiences? I don't care about RGB, but sometimes it's the only variant available.

 

@Mr. Fox @Papusan @tps3443 @Rage Set

 

At least with DDR4, "RGB Powered" RAM sticks tend to need looser timings and that is generally due to the type of RAM requiring more voltage. 

 

So the thought process for me is simple.

 

If I want RGB bling and a certain speed, say 3600/4000 for DDR4 or 6400+ on DDR5, I make sure to get a RGB RAM kit that is specced higher and manually lower the speed. Timings won't be as tight. 

 

If I don't want nor care about RGB in a build, I can "usually" buy a cheaper set of nonRGB RAM and overclock higher, with tighter timings. 

[Etern4l]

1 hour ago, Rage Set said:

 

At least with DDR4, "RGB Powered" RAM sticks tend to need looser timings and that is generally due to the type of RAM requiring more voltage. 

 

So the thought process for me is simple.

 

If I want RGB bling and a certain speed, say 3600/4000 for DDR4 or 6400+ on DDR5, I make sure to get a RGB RAM kit that is specced higher and manually lower the speed. Timings won't be as tight. 

 

If I don't want nor care about RGB in a build, I can "usually" buy a cheaper set of nonRGB RAM and overclock higher, with tighter timings. 

 

Clear, no "bling-bling". 

[Mr. Fox]

12 hours ago, johnksss said:

I just checked and Catzilla is up and running, but I did find that one of my oses would not allow me to upload to them.

 

And the same with Heaven. I had some setting turned off and it was blocking heaven.

6 hours ago, Mr. Fox said:

I cannot access catzilla.com or allbenchmark.com from any of my computer web browsers. I get the "This site can't be reached" error whenever I try.

It must be only sporadic that it is functional. First bench I ran showed the leaderboard in the app, then on the second it went dark, so it is offline in my browser and the app on both desktops and my laptop.

12 hours ago, Papusan said:

Did it work? Yep, Hwbot Heaven bench is a mess. And I'm not so sure it would be so much better if it was a few years newer, LOOL

6 hours ago, Mr. Fox said:

I haven't tried yet. Not a huge priority. I don't expect to get enough  points from it to matter too much.

Yes, I got it to work on a different OS. It did not want to function at all on the OS I first installed it on. But, the GPU clock runs less than half normal boost. If I force it setting the minimum clock in the Radeon CP, that works but there is no change in score or FPS.

 

Now it mysteriously stopped working on my other desktop. I ran several without a problem with the 4090, now I just get a black background, Heaven music and a FPS counter that shows over 100K frames per second with no image rendered, LOL. I don't know what changed. Nothing I can identify. Just a buggy trash piece of software, I suppose. It is so old it looks like scores might be better stock or underclocked versus overclocked, but since it stopped working I can't test that theory. Given how it is "working" (quotes intended) I think it has no value as a benchmark now. The results are suspect. It is not designed for modern hardware or OSes and seems unreliable. Not surprising since it has not been updated since 2011.

 

[Rage Set]

 

On 3/17/2023 at 6:14 PM, Rage Set said:

 

Reliability and support are key in enterprise, as you likely have assumed. So far, in my experience using AMD in enterprise, they are surprisingly as reliable as Intel [once] you get through any bugs. That said, Intel is still overall, better in support. Although, some of my peers will argue that is only due to Intel's historic ironclad vertical integration.

 

 

Quote

In late March, Intel discussed its updated data center roadmap at its AI Investor Conference. The company shared its strategies and TAM to capture the opportunities afforded by AI over the next five years.

 

Notably, the company believes it retains the initiative to dominate the market with its hardware and software stack.

https://seekingalpha.com/article/4591798-intel-stock-buyers-strike-back

 

Unfortunately for Intel, AMD and Xilinx are eating their enterprise lunch. A lot of people are saying that the purchase of Xilinx is what saved AMD and I agree somewhat. 

 

The question is can Intel survive this onslaught from both Nvidia and AMD? While Intel is still the market leader, their lead is eroding and the ramifications are huge. For the first time, I see Intel as being only a Foundry. The mistakes they've made are costing them. 

[Etern4l]

17 minutes ago, Rage Set said:

 

 

 

https://seekingalpha.com/article/4591798-intel-stock-buyers-strike-back

 

Unfortunately for Intel, AMD and Xilinx are eating their enterprise lunch. A lot of people are saying that the purchase of Xilinx is what saved AMD and I agree somewhat. 

 

The question is can Intel survive this onslaught from both Nvidia and AMD? While Intel is still the market leader, their lead is eroding and the ramifications are huge. For the first time, I see Intel as being only a Foundry. The mistakes they've made are costing them. 

 

Are they eating it (the CPUs) or just approaching the table, hoping some stuff is left for them once Intel leaves the table full? Genuine question that can't be answered by chest-thumping marketing materials alone. 

 

As for the significance of Xilinx, how big is the FPGA market in relation to data centre CPUs for example? 

[Rage Set]

50 minutes ago, Etern4l said:

 

Are they eating it (the CPUs) or just approaching the table, hoping some stuff is left for them once Intel leaves the table full? Genuine question that can't be answered by chest-thumping marketing materials alone.  

 

Currently, AMD is the first to the table, followed by Intel. While in this scenario, Intel is eating a lot more than AMD, that is because they have had long-term contracts. Those contracts won't last forever, especially when Intel keeps delaying products. 

 

The article I shared is from an Intel Investor conference summary...aka they [Intel] have to tell the truth. They have admitted they are going to lose more marketshare over the next two years to AMD but hopefully their shift to AI products can give them another way to claw back market and mind share. 

 

 

50 minutes ago, Etern4l said:

As for the significance of Xilinx, how big is the FPGA market in relation to data centre CPUs for example? 

 

You do know that all of the major data centers run custom servers, with custom hardware? Very few run turn-key solutions. The emphasis on custom hardware designs are even greater now with AI. FPGA's offer the most customized options outside of a company spending billions of R&D on their custom silicon. 

 

Put it another way....GPU mining vs FPGA Mining. The GPU offers great general purpose data processing but a programmed FPGA can process specific datasets, sometimes upwards to 10X better at lower power. 

 

Another scenario, instead of using two 3090 TI in SLI and all the power those GPU's require to operate, I can use a 3090 plus a FPGA accelerator to transcode/process video much faster with much lower power requirements. 

[Mr. Fox]

Sometimes inexpensive things can be very useful. I got tired of pulling and plugging cords into outlets. These have a capacity of 15A and over 1800W, which is far beyond the demand of what I have connected to them. Very convenient, like a light switch.

 

https://www.amazon.com/dp/B07QKPBZX1

 

 

Brother @Ashtrix... Glad I already placed an order. Now they're going to get slammed and probably run out of stock, LOL.

 

And more bad news for Windows 7 hold-outs. Steam client is hostage to a dependency on Google Chrome and Steam will stop functioning. I'm glad that most Windows Steam games play fine on Linux using Proton.

 

 

[Etern4l]

3 hours ago, Rage Set said:

 

Currently, AMD is the first to the table, followed by Intel.

 

The article I shared is from an Intel Investor conference summary...

 

I quote: 

 

"Also, note that Intel is still expected to remain the data center market leader by a wide margin in 2023. Hence, if the company could stem and reverse the slide moving forward, it could lift buying sentiments further and put pressure on AMD's more expensive valuation."

 

3 hours ago, Rage Set said:

You do know that all of the major data centers run custom servers, with custom hardware? Very few run turn-key solutions. The emphasis on custom hardware designs are even greater now with AI. FPGA's offer the most customized options outside of a company spending billions of R&D on their custom silicon. 

 

I didn't know that, to me thst would be dependent on the use case for the servers. Do you need FPGAs to accelerate databases and web servers? My question was more concrete though: what's the size of the FPGA market relative to say the total global DCAI spend. My guess is it's tiny, a few %. 

 

3 hours ago, Rage Set said:

Put it another way....GPU mining vs FPGA Mining. The GPU offers great general purpose data processing but a programmed FPGA can process specific datasets, sometimes upwards to 10X better at lower power. 

 

Another scenario, instead of using two 3090 TI in SLI and all the power those GPU's require to operate, I can use a 3090 plus a FPGA accelerator to transcode/process video much faster with much lower power requirements. 

 

It thought bitcoin miners are ASICs rather than FPGAs? At least that's what they are called. 

 

2 hours ago, Mr. Fox said:

 

 

 

 

Respectfully, that 12VHPWR dongle looks like abother useless gimmick pushed by the influencers. Would take more vertical space than the native Corsair cable. 

[Papusan]

On 4/2/2023 at 9:14 AM, Papusan said:

Btw. The new modern locked down AMD Ryzen 9 7900X3D running 90.0 degrees °C under load at 5093.2MHz with below 130W😆 Not even an Corsair 360mm AIO help much on this 12 core’s disaster. 

https://hwbot.org/submission/5239144_neurotix_cinebench___r20_ryzen_9_7900x3d_10440_cb

 

 

@Mr. FoxDo you see the similarity between the two video's? 

 

 

 

 

[Rage Set]

33 minutes ago, Etern4l said:

 

I quote: 

 

"Also, note that Intel is still expected to remain the data center market leader by a wide margin in 2023. Hence, if the company could stem and reverse the slide moving forward, it could lift buying sentiments further and put pressure on AMD's more expensive valuation."

 

 

I didn't know that, to me thst would be dependent on the use case for the servers. Do you need FPGAs to accelerate databases and web servers? My question was more concrete though: what's the size of the FPGA market relative to say the total global DCAI spend. My guess is it's tiny, a few %. 

 

 

It thought bitcoin miners are ASICs rather than FPGAs? At least that's what they are called. 

 

 

Respectfully, that 12VHPWR dongle looks like abother useless gimmick pushed by the influencers. Would take more vertical space than the native Corsair cable. 

 

I'll say this and end it here. This thread for the most part, leans Intel. Nothing truly wrong with that, as Intel have historically offered the best overall overclocking products. Yet, if Intel continues down their current path, they can end up like AMD pre-Ryzen era. Basically, barely holding on from bankruptcy. It's not farfetched to conceive that in five years, Intel could simply be a fabless chip maker like AMD/Nvidia or worse.

 

I am not a true fan of Nvidia, AMD or Intel. I am a fan of computing and competition. Intel for the longest time has rested on their laurels and it caught up with them...in a big way. Who would have thought five/six years ago that AMD would be currently beating Intel in sales in enterprise, going toe to toe in mainstream (no matter how much people want to bash their mainstream parts) and completely eradicate Intel in HEDT/Workstation up until recently? I didn't.

[Mr. Fox]

3 hours ago, Etern4l said:

Respectfully, that 12VHPWR dongle looks like abother useless gimmick pushed by the influencers. Would take more vertical space than the native Corsair cable. 

Definitely not a gimmick, unless you have a 12VHPWR cable with a factory right angle made into it, assuming the right angle is pointing the desired direction. I am not sure there are any. I have only seen cables with straight connectors on the ends.

 

This type of adapter allows you to route the cable the way you want it, in a tighter space and without compromising the cable or connecor. It is no different than buying cables with 90° ends on them, which I already do for the reasons of convenience, ergonomics and preserving cable integrity. I have HDMI, DP and USB cables and adapters with both 90° and 180° ends on them.

 

Given the issues we have seen with putting sharp bends in the 12VHPWR cables near the GPU, this has value to add there as well when there is not sufficient space between the GPU and side panel. 

 

Let's use my current setup as the example. I cannot route the 12VHPWR cable behind the GPU return line due to the severe sharp bend it would require. While I have enough space to not kink the cable, it lightly touches the door. With the adapter, I will be able to have it positioned behind the tubing, rather than in front of it, pointing straight down into the case. The benefits are evident for both application and aesthetics.

  

2 hours ago, Papusan said:

@Mr. FoxDo you see the similarity between the two video's? 

That is very strange. Interesting video, but it seems there is no explanation at this point. If I had to guess, something is wrong with the Indium alloy. I doubt the CPU could get hot enough to melt it if the composition of the solder TIM were proper. The air pockets are also very weird. Perhaps something unusual happened in the manfacturing process that allow the CPU and/or IHS to get contaminated before soldering. I know the fab process has to be spotless and climate controlled.

[Etern4l]

1 hour ago, Rage Set said:

I'll say this and end it here. This thread for the most part, leans Intel.

 

While that may be true, I would like to think of myself as "minimally biased" (obviously, there tends to be some degree of ownership bias, unless the owner is deeply unhappy). I evaluated AMD 5000 series very closely and fairly at the time of the last personal purchase decision, and wasted no time to point out the flaws on the Intel side (below average likes on  those posts lol). Here, I've been mostly asking for more information :)

 

45 minutes ago, Mr. Fox said:

Definitely not a gimmick, unless you have a 12VHPWR cable with a factory right angle made into it, assuming the right angle is pointing the desired direction. This type of adapter allows you to route the cable the way you want it, in a tighter space and without compromising the cable or connecor. It is no different than buying cables with 90° ends on them, which I already do for the reasons of convenience, ergonomics and preserving cable integrity. I have HDMI, DP and USB cables and adapters with both 90° and 180° ends on them.

 

Given the issues we have seen with putting sharp ends in the 12VHPWR cables near the GPU, this has value to add there as well when there is not sufficient space between the GPU and side panel. 

 

Let's use my current setup as the example. I cannot route the 12VHPWR cable behind the GPU return line due to the severe sharp bend it would require. While I have enough space to not kink the cable, it lightly touches the door. With the adapter, I will be able to have it positiong behind the tubing, rather than in front of it, pointing straight down into the case. The benefits are evident for both application and aesthetics.

 

The picture clarifies things. That 12VHPWR cable looks much thicker than the Corsair one, which makes sense given the intended use.

[Mr. Fox]

22 minutes ago, Etern4l said:

The picture clarifies things. That 12VHPWR cable looks much thicker than the Corsair one, which makes sense given the intended use.

The highly flexible 12VHPWR "ribbon cable" that came with my Corsair RM1200x SHIFT PSU is awesome and they should all be made like that. I love that it is flat and all of the wires are attached rather than individual (like a ribbon cable).

[Etern4l]

14 minutes ago, Mr. Fox said:

The highly flexible 12VHPWR "ribbon cable" that came with my RM1200x SHIFT PSU is awesome and they should all be made like that. I love that it is flat and all of the wires are attached rather than individual (like a ribbon cable).

The ribbon one is pretty cool, but it's the slightly cheaper option:

https://www.corsair.com/uk/en/Categories/Products/Accessories-|-Parts/PC-Components/Power-Supplies/600W-PCIe-5-0-12VHPWR-Type-4-PSU-Power-Cable/p/CP-8920284

 

 

Mine has individually sleeved cables which gives it that extra flexibility:

 

https://www.corsair.com/uk/en/Categories/Products/Accessories-|-Parts/PC-Components/Power-Supplies/Premium-Individually-Sleeved-Type-4-12VHPWR-Cable/p/CP-8920331

 

Bends effortlessly at 90 degrees such that the cable requires about 2-3 cm clearance from the card. To be fair, that depends on the card, the socket is fairly deeply seated on the 3090 Ti FE.

 

In summary, yes - looks like I might gain maybe 5mm of clearance without any need for cable bend. Something to consider if I need to stick a 4090 in. The question though is about the reliability of the L-adapter itself, and I am a big fan of the KISS principle. 

 

[Mr. Fox]

30 minutes ago, Etern4l said:

The question though is about the reliability of the L-adapter itself, and I am a big fan of the KISS principle. 

Sketchy adapters are always something to be cautious about. Poorly made adapters can cause far more issues than they solve, and I have had a few that ended up going back for a refund or tossed into the trash. One of these being poorly made could cause the same issue as not plugging in the cable properly (i.e. fire or melted terminals).

 

Looking at how these are constructed, and given their solid brand reputation, I suspect they will be as reliable as necessary. It took them a good while to release the adapters, so I would hope that means the time was spent for R&D/testing. They even incorporate a thermal solution.

[Papusan]

11 minutes ago, Mr. Fox said:

Sketchy adapters are always something to be cautious about. Poorly made adapters can cause far more issues than they solve, and I have had a few that ended up going back for a refund or tossed into the trash. One of these being poorly made could cause the same issue as not plugging in the cable properly (i.e. fire or melted terminals).

 

Looking at how these are constructed, and given their solid brand reputation, I suspect they will be as reliable as necessary. It took them a good while to release the adapters, so I would hope that means the time was spent for R&D/testing. They even incorporate a thermal solution.

Yep, a short adapter connector should be more foolproof vs their cable adapters that was the major culprit for black screens. People had to use the awful adapters coming with the cards for troubleshooting all their problems with 4090. In the end a quality problem/QC problem from CableMod. 

[Reciever]

2 hours ago, Rage Set said:

 

I'll say this and end it here. This thread for the most part, leans Intel. Nothing truly wrong with that, as Intel have historically offered the best overall overclocking products. Yet, if Intel continues down their current path, they can end up like AMD pre-Ryzen era. Basically, barely holding on from bankruptcy. It's not farfetched to conceive that in five years, Intel could simply be a fabless chip maker like AMD/Nvidia or worse.

 

I am not a true fan of Nvidia, AMD or Intel. I am a fan of computing and competition. Intel for the longest time has rested on their laurels and it caught up with them...in a big way. Who would have thought five/six years ago that AMD would be currently beating Intel in sales in enterprise, going toe to toe in mainstream (no matter how much people want to bash their mainstream parts) and completely eradicate Intel in HEDT/Workstation up until recently? I didn't.

The moment AMD becomes Dell's offering in their enterprise rack solutions as the advertised pitch should be the moment that a cataclysmic shift just occurred much like an earthquake.

 

Might be a few awkward conversations behind closed doors and maybe a couple of heads on pikes as a peace compromise but still lol

[Etern4l]

14 minutes ago, Reciever said:

The moment AMD becomes Dell's offering in their enterprise rack solutions as the advertised pitch should be the moment that a cataclysmic shift just occurred much like an earthquake.

 

Might be a few awkward conversations behind closed doors and maybe a couple of heads on pikes as a peace compromise but still lol

 

Just took a quick look: 

Single socket: EPYC

Dual socket: half EPYC, half Xeon

Quad socket: Xeon

 

That's just the front page, didn't delve in to see if it's possible to configure a single-socket Xeon. 

 

https://www.dell.com/en-uk/dt/servers/poweredge-rack-servers.htm#tab0=0&accordion0