The New AMD Building Thread

Shaff,

I'm not getting worked up, and I have no problem with constructive criticism. You say I am misinterpret what he says, and I muddle the response? Let's see! You and I are discussing "pears", two different reviews from Bit-Tech, and he come along and starts an argument about "Apples", Socket AM3 and socket AM3+, says we were discussing Bulldozer? Had he posted the page, rather than the Graph, we wouldn't be having this discussion, because the first thing I look at with any processor review is when it was written, especially with a brand new line of processors! I questioned the Bit-Tech article you posted, which was written in Jan. of 2011, and felt it was a very pro Intel, anti AMD review. I decided to take a look at Bit-Tech for some AMD reviews to try and get a clearer picture of how Bit-Tech does things. I found a review of the 1100T, Dated a month before yours, and found some discrepancies between the two reviews. I pointed these out from both sides, and feel that Bit-Tech seems to come across as very Pro Intel. We weren't discussing Bulldozer, we were discussing the article you posted. Sam comes out and slams me for "using such old reviews", and claims we were discussing Bulldozer. He had posted a graph showing power consumption from Bit-Tech that's totally insane and totally disagrees with Bit-Tech's articles from a few months before. You will note that I made sure that the material you and I were discussing was dated, and no accusation of using old articles was mentioned in regards to your first Bit-Tech link, nor was it raised with my post until I pointed out these discrepancies, with the argument being that it wasn't really a fair review because they tested all of the AMDs with 1 and 2 generation old motherboards, and that they should be tested on the newest AM3+ motherboards. Then there was the argument where Sam said that they were tested on AM3+ motherboards, and I provided links to show that they were tested on AM3 Crosshair IV Formula motherboards, which was what I showed in my response and the setup page from both articles. That's when he went on about the "using old reviews, and asking what the 1100T had to do with anything. I chose to make my comparisons using the 1090T because I own one and I also own a great AM3+ motherboard, and Bit-Tech's testing of the 1090T at 4.2GHz didn't come close to my own results at the same speed. Especially since it disagrees with just about every other review in the free world!

All of this could have been avoided had Sam dated the reference chart he posted. Then I would have known that we were not on the same page! We all know that with a major introduction like Bulldozer, that changes in bios and drivers can and do happen day to day, so reviews may differ considerably from day to day as well. It would have been apparent to me had I seen a date shown, after the introduction of the Socket AM3+ motherboards, and I would have pointed it out right away! Like I said I used the 1090T & 1100T power consumption numbers because I have never seen the 17 920 use less wattage than the 1090T before in any review. The whole point is we were making comparisons between your Bit-Tech Review and the one I posted. We were not discussing Bulldozer.

Where I take exception with Sam is with his wise a$$ comments, like "for people who know how to read", or "Bit-tech are also using an AM3+ board (Crosshair 5), not AM3, as you like to keep pretending they use", "Bulldozer has been a complete failure". All BS! I even jumped him (rightfully so) on the building thread for being disrespectful to me.

First, I read quite well, the flap with the AM3 and AM3+ boards was an integral part of the discussion you and I were having about Bit-Tech, and the information I provided was 100% on that conversation, so I wasn't pretending anything. As far as Bulldozer being a complete failure, that's utter nonsense. If it wasn't, I could have claimed the same thing when the 6600 Quads were released and the C2Ds were blowing the doors off of them, in particular the C2D E-8400, which decimated it. Intel worked it out and the 6600 eventually earned it's place, ahead of the Dual cores. What he fails to recognize is that when the criteria is met, where Bulldozer shines, it jumps right in there near the top in benchmarks, a fair number of games and apps, generally splitting the 2600K and the 2500K, although beating both at times.

What it's poor at is obvious, single threading. Multi threading is an entirely different story. Bulldozer can run with the big boys when you throw a lot of threads at it. in short, AMD has proven that their new architecture concept actually works. Not only that, they've demonstrated that their method of Core Boost, is superior to Intel's Turbo boost by a wide margin! Other little things as well, like having a very Intel-like Memory bandwidth, matching the Core i7 920!

I think that AMD has some work to do on the processor and needs to vastly improve single-threaded apps, and the Bios folks need to step up too. They also need to do some more work on power consumption. BTW! If you didn't notice, Intel is using the Core i5 2500K as a loss leader, $95 lower in price than the 2600K, and $80 less than the 2600. When Bulldozer is doing it's thing, it generally slots the FX-8150 right in between the two. Personally, I think the chips with the most potential are the FX-2120 and the FX-6100. They overclock well, and for the most part, consumers like them. Real consumers anyway (lot's of I Trolls)! I'm glad Newegg validates Ownership.

One more thing! It's been mentioned that a number of AM3 motherboards can be made ready for Bulldozer with a bios flash. That's true but, they will not run as full AM3+ chips, due to the different design architectures between AM3 and AM3+, particularly the Direct Connect Architecture, so it should be slower and not overclock as well. Something along the lines of the AM2/AM2+ boards supporting AM3 CPUs. They do support them and they do work fine, but it will take a genuine AM3+ motherboard to fully support some features of Bulldozer.

Here you go Shaff, I'll give you your chuckle for the day. I found this in a Newegg review from a verified owner.

"The Zambezi does 3 things extremely well, multitasking, overclocking, and attracting haters." He's right!

Best Regards,
Russ

 

I wouldn't feel comfortable running a newer CPU, on an older board. For exactly that reason. It's different if one wants to run an older cpu on a newer board though. AM2 on AM2+, AM3 on AM3+. In fact, it'd probably run better on the newer board. Revised technology

 

I've been doing some more testing with my new 990XA-UD3 motherboard. A while back I had said that I expected my 1090T to get to 90,000+ MIPS at a lower speed than the 4.0GHz it took to reach it on the 790X-UD4H. Here you go!

Processor Arithmetic



Memory Bandwidth



Multi-Media



CPUz



I had to cut the NorthBridge frequency to 2400MHz, which dropped the HT Link frequency to 2400 as well. This cuts the Memory Bandwidth a little bit, but doesn't hurt the performance at all. That's why I generally don't bother with overclocking memory. Don't get me wrong, it can be worth the effort to overclock the memory, but it can take months to do it successfully, as it can be tedious and time consuming work to do it right. It takes far more than just raising the memory speed to get meaningful results, as there's a bunch of trial and error involved.

Oman7,

How come you are running memory timings of 8-8-8-24, when they should be 7-8-7-24? You need to set the timings manually to 7-8-7-24, and you should see an improvement in memory bandwidth.

Best Regards,
Russ

 

I know russ Seems like I tried once upon a time, but it wouldn't stabilize. I'll be trying again tonight.

 

Oman7,

It seems to me that if the CPU Host frequency is left at 200, the memory should run at stock timings regardless of the overclock.



It runs the same at 3.6, 3.9, 4.0 and 4.2GHz, with timings of 7-7-7-21, and a memory bandwidth of just under 9000MB/s with both the NB and HT Link frequency set to 2600MHz. I was surprised that I had to lower the NB frequency with it at 3.9gHz, but since I was just fooling around, and with the obvious improvement in overall performance, it was just a minor trade-off that cost about 500MB/s in overall memory bandwidth, and left me with 4800MT/s instead of 5200MT/s. Still, the 1090T is only a 4000MT/s chip, so I'm still ahead of the game.

Weird that my R.O.G. version of CPUz was not working before. I removed it and dumped everything in the folder except the .rar file. Ran fine after I unzipped it.

Russ

 

I'm currently running Bd Rebuilder a second time(All cores maxed)! I'm loving the obvious speed improvement. But I still can't wait to see BD rebuilder doing the complete job in 30 minutes, Like Dvd rebuilder. Of course we're at least 2 - 3 yrs from that. Not sure how a 16 core CPU would handle it. I'd sure love a board that could take two quads or two 8 core Cpu's. But just for X264? Eh, time will tell LOL!

The job will be finished before too long, and then I'll try the Ram overclock again.

Russ, if it should run stable, then why do most(IF not all) AMD boards underclock Ram? Further, why does it increase the latency?

Well well well! A modest gain with the X264 benchmark. Not sure if it's stable at those settings yet, but I trust you

 

Oman7,

My guess is it's the bios not being able to tell what the memory is, in terms of settings, so the memory timings and speed are reduced to a safe level. It did the same thing when 1066 memory came around on the Intel's. It would see it as 800MHz memory. With my 1333 RipJaws Cas7 memory, it ran noticeably slow. I checked the setup, and it was set to cas9 settings. I put those in manually 7-7-7-21, and it's run beautiful ever since. If you do buy a series 9** motherboard, be sure you have a spare 1.5v DDR3 memory stick. The memory standard for AM3+ motherboards is 1.5v. It won't post with most 1.60v or 1.65v memory in it. That's why you need a 1.5v stick, to get it to post so you can raise the memory voltage. Save the adjustments and shut it down after the restart and install the higher voltage memory. We went through the same thing with the P35 Intel motherboards. That's why I didn't want to get 1600 memory. I could get cas7, but the voltage was 1.65v, and the timings were 7-8-7-34. From what I understand, 1600 memory is not too easy to get running at stock settings on a lot of motherboards. Even up, I don't lose but about 2% of memory performance, with the 1333 at 7-7-7-21, and it's just easy peazzy, set em and forget em! LOL!! The Motherboard is the most user friendly board I've ever seen, and I can google any setting on the board I don't understand, and actually get an answer! It does have one flaw, a delay in the Post where you wait for about 12 seconds as a cursor on the L top flashes, then it says Starting Operating System and the windows screen appears. I'm guessing that an SSD will cure that. I'll see what the Christmas sales offer, but I want the Patriot!

Russ

 

Yet another BD rebuild accomplished I would really like the 1090t. Russ, you should give me your processor LOL! Nah, I'll get one. I'm definitely liking the X264 benchmarks I'm seeing for the 6 core beast. I think it could be worth it. I must admit though, my quad doesn't seem to care, fully maxed, and running small background processes. E.g. browsing the web. Haven't tested much else. CPU's are multitasking freaks though! It's hard drives that don't like it I guess I'm just afraid I'll find an instability :S This is my first time running 4.0Ghz stable! I guess it feels a little too good to be true LOL!
I may just try to get a board, and the CPU. More so for the UEFI bios, and future upgrades.

 

http://www.youtube.com/watch?v=YjGmTKSfQhY&feature=player_embedded

 

What of it? Bulldozer is old news now. In the opinion of many, not old enough. A lot of AMD fans would wish Bulldozer had never happened...

 

While I'm disappointed with several factors of the Bulldozer, I wouldn't say I wish it hadn't happened. A step is a step, no matter how small. And no, I would not say a step backward. It has certain benefits over other processors, despite people ignoring them. I do acknowledge it's single thread performance(or rather lack thereof). I've read of people talking about windows being coded to handle them better. Performance increase? Don't know yet.
The only thing at this point that really disappoints me, is the power consumption. But it is a core beast. Perhaps we'll see what piledriver has in store. I'd really like to see a revision of the 8150/8170. E.g. B3 stepping?

 

Wonderful, New page bug...

 

npb


The thing about Bulldozer, is that in actual fact, it isn't a step forward. What it is, is a crossover from the server market to the desktop market. Everything about the way Bulldozer works makes it better for servers. The same is also true of Sandy Bridge-E (i7-3900 series). When you remember that Opterons exist, Bulldozer isn't a step forward at all. It's just a server chip that fits on desktop motherboards. They put server chips on server boards for a reason...

 

THanks sam. YOur message allowed me to view the page

 

That's what I'm here for

Edited my message into the other post.

 

Well, it certainly caters to multitasking freaks There are people that have compared it to 2500-2600k, and said the bulldozer felt smoother when running multiple processes/threads. Perhaps an Amd fanboy, giving points in Amd's favor though. It sure would be a step up in my configuration. I'd probably be concerned about load on the VRM's though :S

 

That would come from there being many cores, so there is more breathing room for other tasks if you really are maxing out the CPU 100%. If you want a system to be fluidly responsive but use 100% of its raw power, you're doing it wrong, but there's no reason you can't down-prioritise a task if you want to use the system for a bit during the middle of the work.

 

I wasn't expecting a miracle, but here it is. Not sure what I'm doing in february. I'd like to see my encode times cut in half, clearly I'm not gonna see that for at least a while longer. Perhaps when the 10 and 12 core beasts come out, I'll build a machine chiefly for encoding. Provided it isn't too expensive :S

Microsoft patches Bulldozer’s performance
http://www.extremetech.com/computing/113421-microsoft-patches-bulldozers-performance-we-investigate

 

4%, about what I expected. I was thinking along the lines of 5-10.

 

I've decided to add this post from the building thread, in hopes of discussing AMDs and overclocking. I meant to post this PDF the other day. I sent it to a couple of friends that are planning to OC their 1090ts. I got called for dinner and I forgot to send it. Sorry!

http://sites.amd.com/us/Documents/AMD_FX_Performance_Tuning_Guide.pdf

While this is for tuning with AMD OverDrive, I just used some of the same settings in my motherboard's setup. The results are pretty darn good, if I do say so myself!

120,504 MIPS!



MB 12,466MB/s



Cache and Memory performance




Processor Multi Media! 601,104/421,998it/s



This is at 4.03GHz, with great temperatures. That's where I'm going to leave it for now until I get another A/C 120mm 77 cfm, pwm push fan for the radiator.

http://www.newegg.com/Product/Product.aspx?Item=N82E16835186078

You can pair two of them to the CPU fan header on the motherboard, using the wiring that comes with the fan, and since they are the same fan, they both run the same speed, controlled by the CPU temperature. I think it will do at least 4.4gHz, maybe more. The definitions of all the new controls in the setup, is worth reading, if you want to OC an AMD.

Have Fun,
Russ