Are there Intel CPUs that perform better than the i9-7900X but not a lower cost?

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

modeonoff

Distinguished
Jul 16, 2017
1,415
18
19,295
Hello, last year some of you mentioned that one had to use water cooling system due to heat generated by the i9-7900X. I am afraid to use water cooling as water might damage the expensive system. Has the situation been improved? Can I just use a large heatsink and fans instead? Over the past few months, is there any new CPU that has similar or higher performance than the i9-7900X but does not require water cooling system to keep it cool?
 
Among the current Intel Core X-Series, I guess I prefer the 7900X because I don't need cores more that it can offer. Moreover , the number of lanes of more expensive ones are the same as the 7900X's. Besides these CPUs, is there any that performs better than the 7900X but at similar price (below $2000 each)? I think Xeon (especially dual versions) may have more lanes but they are more expensive and the single thread performance is not as good as the 7900X. Am I correct?

I read about the X299 VRM issue. Has it been fixed? If I am paying for that money, I hope the CPU is free from problems. I guess for current generation of CPUs (both INTEL and AMD), they still have those Meltdown and Spectre issues that we cannot avoid. Is that right?

 
Maybe the confusion is cased by my misunderstanding. What is the actual meaning of:

44- Lane CPU
3 x PCIe 3.0/2.0 x16 (x16, x16/x16, x16/x16/x8)
2 x PCI3 3.0x4 (max at x4 mode)
1 x PCIe 3.0/2.0x1

By "3 x PCIe 3.0/2.0 x16 (x16, x16/x16, x16/x16/x8)", what does it actually mean?
 
Not all the available PCIe lanes in an Intel PC are available from the CPU. Some come from the PCH (Platform Controller Hub) AKA the chipset. The PCH communicates to the CPU over DMI, which itself is usually connected to via 4x PCIe lanes. All devices on the PCH share that bandwidth, so typically things like GPUs, RAID cards, and other devices aren't made with this in mind. But all the ancillary devices in the computer hook up to this, SATA controllers, onboard lan, some NVMe drives.

Higher performance devices get hooked up directly to the CPUs available lanes.

Motherboard manufacturers have these lanes to allocate through switching or simple direct connection.

What they are telling you with those numbers is the configuration of the lanes and slots. 3x PCIe 3.0 is basically saying you have 3 PCIe slots that are hooked up to the CPU, available modes are 16x/16x/16x or anything less. (Though that is slightly simplifying it a little) All the others, including one x16 slot, is hooked up to PCIe 2.0 through the PCH (That should be an older motherboard, does anyone know if AMD uses PCIe 2.0 on threadripper?) Intel has in the past, but not on their recent stuff (I think)

Typical Intel consumer, Z class CPU, 16x PCIe 3.0, supported modes are a single 16x card (16x/0x). Two 8x cards (8x/8x) and occasionally (8x/4x/4x) for three cards. All the 1x slots and usually one x16 slot (at only 8x mode usually) run through the PCH.

28 Lane CPUs behave differently than 44 Lane CPUs in X99 and X299 boards. So the possible configurations get more complicated. Suffice to say, typical 28 Lane CPU can handle 16x/0x, 16x/8x/0x , 16x/8x/4x.

44 Lane CPUs 16x/16x/8x, 16x/8x/8x/8x, 8x/8x/8x/8x

If there is any room after that, then the smaller slots will still be running off the PCH.

It has been brought up before, but there are boards that support larger configurations using full PCIe bridge controllers (same ones found in the old dual GPUs), but in name only. They are still limited by the bandwidth available to the CPU.

I suggest posting your final build for a once over by others to make sure you get what you are after.
 
So, in this case, what they said is that there are three PCIe3.0 slots that can all run concurrently at full x16 speed? They are hard wired directly to the CPU?

The situation is that I chatted with tech support of ASUS yesterday. I asked him to check if there is any motherboard for i9-7900X, 7800K and Threadripper that allows running four GPUs all at top speed of x16. If not, if there is any motherboard for these CPUs that allows running three GPUS all at top speed of x16. I asked him the number of PCIe3.0 slots that have direct hard wired connections to the CPU. I also asked him the questions I posted in this thread (such as if I put 2 or 3 GPUs in the ASUS WS SAGE, will the dual PLX chips actually make the GPUs run slower than as if they were put in motherboards most people use). The guy had no idea and told me that he would forward all my questions to the support team. He told me to give them 1-2 days.

Today, somebody from the support team replied. He suggested me to use the PRIME X299-A because:

44- Lane CPU
3 x PCIe 3.0/2.0 x16 (x16, x16/x16, x16/x16/x8)
2 x PCI3 3.0x4 (max at x4 mode)
1 x PCIe 3.0/2.0x1

He did not address my questions in regard to the WS X299 SAGE motherboard. When I looked at what the guy I chatted with sent to him, I found that he only sent a small part of my questions to the tech team. He did not forward my questions about the WS X299 SAGE board to them.

As far as I know, none of the Threadripper boards allows 3 GPUs all running at top 16x speed.
 
And yet there isn't a gpu that'll soak up the entire bandwidth available to pcie 3.0 x16. Even a 1080ti is pushing it to use x8. So even if there was a performance loss by using pcie 2.0 (there isn't, it's just got less bandwidth available), even at x16 a 1080ti will still come up short. So it's a moot point.
 
Hello, I need some help in choosing a system. As you know, I planned to build a high-end workstation using 7900X. Then a colleague warned me that the Meltdown and Spectre patches could affect performance. The hit could be up to 20-40% depending on what I do. I decided not to spend too much money on such buggy system. Since Cascade Lake should have most of these issues removed at hardware level by this Fall, I decided to build an inexpensive system to get some work done during this transition period. After choosing between 8400 and 8700K, I chose the latter and searched for a motherboard that is compatible with Ubuntu Linux and Hackintosh. After some hard work, I settled for the ASUS ROG Strike-E Gaming board. Then, ASUS told me to follow their QVL list that does not list any 32GB-64GB RAM. Now, I figure out that if I build a 8700K system with 64GB RAM (don't know if I could re-use the RAM in a Cascade Lake system to get top performance as this up-coming system will use 6-channel memory. It means that I may need to buy a new 6-channel RAM kit for the new system), I may perhaps use it longer until Cascade Lake or even PCIe 4.0/5.0 components come out in 2019-2020. Then, rather than throwing out the 8700K and motherboard, I build a high-end system and get an extra computer by paying for an extra case and PSU. Then, I figured out that for this 8700K system with 64GB RAM, if I add about $300 more, I can get a Threadripper 1900X system. What suggestion do you have?

Is a TR 1900X system "much better" than a 8700K system? I heard that Intel system has better memory sub-system than AMD's.