AMD Phenom II X4: 45nm Benchmarked
Can AMD finally recapture its former market lead with the Phenom II? We tested their new 45-nm CPU, which is compatible with AM2 motherboards. How do things look from performance, energy consumption and overclocking perspectives? Read More
- Intel Plans to Ship Nehalem Chips (Core i7) Nov. 17
- Intel Core i7 (Nehalem): Architecture By AMD? : Introduction
- Intel reveals some More Nehalem information at recent presentation
- Intel to bring back Hyperthreading with Nehalem core
- Best Processor for Gaming ??
- OC'ing the DS3L vs. the DS3R
- Intel to drop overclocking for mainstream Nehalems
- Upgrade from e6600 to e8400 worth it? My OC'ing results...
- 8800 gts g92 bad 3dmark score
- Core i7 or Core 2?
- The Return of Intel's Pentium MMX
- Exclusive: 'Nehalem' at 2.9GHz and X58
- AMD Shows Ray-tracing Tech on YouTube
- Atom Shortage May Affect Low-cost Notebook Panel Shipments In July
- AMD Improves CPU Market Share - With Little Impact On Intel
- Gigabyte Goes Extreme Cold Overclocking
- AMD Introduces New Phenom Flagship Processor
- Asustek Initiates Price Cuts For P45 Motherboards
- Intel May Integrate DRAM into CPUs
- Playstation 3 Processor Powers The World's First Petaflop...
'Nehalem' 2.93 GHz Benches Revealed
Several days ago we took a look at Intel’s next generation X58 platform for Nehalem in which we explained that the new chipset will be bringing along several new features. The most significant of these features is QuickPath, which is a brand new link architecture very similar to AMD’s HyperTransport, and getting rid of the front-side bus (FSB) system altogether.
What this means for enthusiasts is that old methods of overclocking will no longer work with Nehalem. What will happen is that the connection between the CPU and the northbridge (X58 MCH), will be based on an external clock multiplier running at 133 MHz. At this time, with an early X58 sample motherboard on hand, we’re still unable to change this frequency, but our motherboard sources tell us that this is a feature that will be available via the system BIOS once things go into full production.

Today however, we were able to really play around with a 2.93 GHz quad-core ’Nehalem’ processor. Unfortunately, we were not able to test the system for long, and therefore very little comparison could be made to another system. However, we did manage to run the chip through some common synthetic benchmarks. We will definitely be providing a full performance review. For today though, we have a quick look at how Intel’s next generation platform performs.
We used the following system specifications:
| Processor | Intel Nehalem 2.93 GHz |
| Motherboard | Prototype X58 motherboard |
| Memory | Crucial Ballistix DDR3-1600 1GBx2 |
| Storage | Seagate SATAII 750GB (ST3750330AS/P) |
| Graphics | ATI Radeon 4850 HD |
| OS | Microsoft Windows Vista SP1 |
| Resolution | 1280 x 1024 |
| Patches | hotfix_vista32-64_dd_ccc_hd4800series_64906 |
The results so far are:
| 3DMark 06 1280x1024 noAA | Score |
| 3DMarks | 12786 |
| SM2.0 score | 4605 |
| HDR/SM3.0 score | 5600 |
| CPU score | 5183 |
| PCMark 05 | Score |
| PCMark | 9852 |
| CPU score | 9583 |
| Memory | 9010 |
Just for reference, a similarly configured Core 2 Extreme X6800 at 2.93 GHz delivers a CPU score of 7417 - roughly a 29 percent difference clock for clock. When compared to a similarly equipped AMD Athlon FX-62 system, the gap widened even more, to a staggering 36 percent difference, with ’Nehalem’ taking a long distance lead. Keep in mind t hat this is actually very impressive, because our Nehalem platform is running on very early hardware, specifically the motherboard. We are expecting even better margins in the coming months as the platform matures. Keep in mind though that the CPUs above are all dual-cores. We’ll update with quad-core comparisons shortly.
| 3DMark Vantage | Score |
| CPU score | 17966 |
| CPU Test1 (Plans/S) | 2515.1 |
| CPU Test2 (Steps/S) | 23.08 |
Although this is a short preview, it gives us an overall idea of the potential that’s in Intel’s next-generation CPU. Nehalem is a huge step forward for Intel in many ways, leaving behind some old concepts, and taking a big step forward in other areas.
At its expected announcement in Q4’08, Nehalem will be released initially with 4-cores, and later scaling up from there. One detail to note is that HyperThreading makes a nice comeback as well. Our CPU read as 8 processors in Windows Vista.
Frequency potential with Nehalem in its current state of silicon, is very healthy, Intel told us during Computex earlier in June. With our preliminary tests of the 2.93 GHz CPU, we’re confident in saying that enthusiasts can expect even faster processor frequencies down the line.
Keep in mind this article is not meant to be a review, but is a short expose on the potential of the Nehalem family. We hope it gives you an idea of what to expect.
Update: Quick tests between our 2.93 GHz Nehalem unit and a quad-core Core 2 QX6800 (2.93 GHz) as well as an AMD QuadFX Athlon FX-74 (3 GHz) reveal the following:
| PCMark 05 | Score |
| Nehalem 2.93 GHz | 9852 |
| Core 2 Extreme QX6800 2.93 GHz | 8925 |
| QuadFX Athlon FX-74 3 GHz | 7501 |
Performance between an early quad-core Nehalem at 2.93 GHz and a Core 2 Extreme QX6800 running at the same frequency reveals more than an 11 percent difference in performance. Meanwhile, Nehalem takes an even larger lead with a 24 percent between it and the higher clocked QuadFX Athlon FX-74. It’s interesting to note here that the quad core FX-74 is only roughly 7 percent faster than the dual-core Athlon FX-62m which is quite an old CPU now.
This gives a good view into the engineering behind the Nehalem architecture and the performance gains that Intel is able to achieve with even pre-production silicon. Very healthy? We definitely think so.
Update 2: As we mentioned earlier, this article was meant to give a brief introduction into some basic performance figures for Nehalem. Keep in mind that the CPU launch is still several months away. This is not meant to be an in depth review covering all CPU comparisons and charts. Fortunately, we were able to pull up PCMark 05 performance figures comparing Nehalem at 2.93 GHz and Phenom 9850 BE:
| PCMark 05 | Score |
| Nehalem 2.93 GHz | 9852 |
| AMD Phenom 9850 BE 2.5 GHz | 7400 |
The result is roughly a 33 percent difference. Right in line with what we predicted. For reference, the fastest Phenom available, the 9950 BE running at 2.6GHz was able to give a score of 7680, slightly above the Phenom 9850 BE by 3.7 percent. However, Nehalem is still faster than the fastest Penom 9950 BE by 28 percent.
We’ll release real-world gaming performance when we get more mature hardware through the labs.
Thanks to jumpingjack6 for correcting us on our percentage math. Nehalem is actually 33 percent faster than Phenom 9850 BE with a 17 percent clock difference
-
Previous News Article
LCD Panel Makers Cutting Small- To... -
Next News Article
Hitachi Fires Up Second-generation...
React! Return to news index
- Seagate Boosts Hard Drive Storage Capacity To 1.5 TB
- Hitachi Fires Up Second-generation Terabyte Hard Drives
- 'Nehalem' 2.93 GHz Benches Revealed
- LCD Panel Makers Cutting Small- To Medium-size Output Amid Weakening Demand
- Philips To Transfer PC Monitor Brand To TPV
- Samsung Puts 128 SSDs Into Mass-production

I'm not sure how relevant it is, but you don't mention if you're running Vista SP1 32 or 64 bit. Anyway really looking forward to a full review.
Can't wait for the full review
isn't the x6800 a dual core? you are comparing a dual and a quad?
Nice. So I assume the NDA is being lifted?
Uhhh, so the QUAD CORE nehalem is 23% faster than a DUAL CORE? Errr.
Someone made a type-o, or that new cpu is useless
isn't the x6800 a dual core? you are comparing a dual and a quad?
Obviously you haven't read the entire article, here is the answer to your question quoted directly from the article:
"Keep in mind though that the CPUs above are all dual-cores."
I didn't understand the 23% comparison. Please elaborate.
Uh, isn't the FX-74 going on two years old now? Odd setup to benchmark against.
WHY would you compare to Athlon and not (Quad) Phenom????
Obviously you haven't read the entire article, here is the answer to your question quoted directly from the article:"Keep in mind though that the CPUs above are all dual-cores."
What's your point? There are quad core cpus from both intel and amd to compare with, and has been for a while now. There's no point what so ever to use a dual core for reference.
There's only one scenario where that comparison would make sense, and that's if 3dmark is exclusively single threaded, and as far as I know, it isn't.
Scoring 23% better with twice as many cores is a rubbish score.
What's your point? There are quad core cpus from both intel and amd to compare with, and has been for a while now. There's no point what so ever to use a dual core for reference.There's only one scenario where that comparison would make sense, and that's if 3dmark is exclusively single threaded, and as far as I know, it isn't.Scoring 23% better with twice as many cores is a rubbish score.
How is it rubbish? i have a dual core cpu and i'd like to know how much performance i'd get from going to Nehalem when it comes out? Seems reasonable for me and any other dual core owner. Get a better grasp on the scope of things and not such a narrow mind. In fact, i and many others didn't plan on upgrading to quad cores for the very reason that the performance difference were small, as indicated in the article. now that we see even bigger differences with nehalem, I feel now it's reasonable to upgrade for the price. Sheesh.
And on top of it all, they have even stated that the AMD is a quad, but the FX74 is a ancient dual core with 90nm technology?!
Even the intel dual core is rather old, with 65nm. An e8500 would have been a better choice, being 1333fsb/45nm and so on...!?
cool
And now they add an amd quad fx test...... Is this a joke Toms?
Have you heard of Phenom recently?
I think people should wait for the full review. This is just a teaser really.
Unfortunately 3DMark is by no means a good way to show a CPUs power. A real test will come in the encodig/decoding and also how well it can run the games physics and such since it is supposed to have a onboard physics engine, courtesy of Havok.
Wake me up when we have non synthetic, real world benchmarks.
I have but one question: Is that core voltage for real?
Good Point, That is a VERY low voltage at that speed.
asdasd123123 dude don't vote me down because what I said makes sense.
maybe the 'Quad' Athlon FX74 (isn't this actually a twin dual-core processor MB?) was used as this is the only AMD 'Quad-ish' offering to run at 3GHz native, I'm probably wrong but no Phenoms run that fast without O/C do they? Just a thought.
LOL.
Here's an idea: Take that unlocked multiplier on that Phenom BE, and set it to 3GHZ, AND THEN COMPARE THE BENCHMARK SCORES... Surely the experienced overclockers at Toms can get it 3ghz stable. The fact that it was OCed from 2.5 to 2.6 makes me think that Nehalem is not that much better clock-for-clock, your choice of platforms for comparison are just silly.
How is it rubbish? i have a dual core cpu and i'd like to know how much performance i'd get from going to Nehalem when it comes out? Seems reasonable for me and any other dual core owner.
Because the entire point is to show how much difference there is between Nehalem and Core 2 processors, which is why they used a processor with the same clock speed, and yet, it literally tells you nothing.
You say that it tells you how much of an improvement a Nehalem will be for you and others switching from a dual core Core 2, but what if a quad core Core 2 would have made just as much or extremely close to the same difference? It does not tell you what a Nehalem gives you that you cannot get today. There is absolutely nothing you can really take away from this, because even in the event that you have one of these exact processors compared to, no one is going to have this particular Nehalem setup.
This site absolutely baffles me lately. Not that I'm ungrateful that they do tests and take the time to write about it, but on a very consistent basis, there have been glaringly obvious things wrong with the articles that totally defeat the purpose of bothering.
And now they add an amd quad fx test...... Is this a joke Toms?Have you heard of Phenom recently?
This is not a full review, and YES, read the end of the post, there are the Phenom 9850, and 9950.
isn't the x6800 a dual core? you are comparing a dual and a quad?
Read more before making a comment. QX6800 is an QUAD-CORE CPU.
WHY would you compare to Athlon and not (Quad) Phenom????
FX-74 is not an Althon, the AMD setup combined of two FX-74, and it made a quad-core setup (dual-sockets mother board).
Please only make a comment, when you are fully understand the post.
onearmedscissorb, i don't think that's the case here. I think its more on the lines of showing some figures in there to see some numbers for nehalem. That was the bottom line from the get go, at least to what the author stated. All those updates with comparisons only came up when people started whining. Bottom line is that the platform is still way early anwyay so any figures would not be final anyway. Just wanna see some numbers.
I have but one question: Is that core voltage for real?
You saw that too? I asked the same question when they first posted the CPU-Z chart cuz I was like "Wow thats freakin low for that speed".
LOL.Here's an idea: Take that unlocked multiplier on that Phenom BE, and set it to 3GHZ, AND THEN COMPARE THE BENCHMARK SCORES... Surely the experienced overclockers at Toms can get it 3ghz stable. The fact that it was OCed from 2.5 to 2.6 makes me think that Nehalem is not that much better clock-for-clock, your choice of platforms for comparison are just silly.
Or here is a better idea. Lets not use synthetic benchmarks that show nothing of true real world performance.
You also forget that this is not even in the final stages and is really early hardware. There will be performance enhancements to come more than likely and will change a lot of things really.
Also in 3DMark/PCMark a OC normally only gives a few hundred points so what would the point be? Right now on a clock per clock Q6600 vs a Phenom 9750BE, the Q6600 gets a higher 3DMark CPU score. So whats that tell ya?
Right now on a clock per clock Q6600 vs a Phenom 9750BE, the Q6600 gets a higher 3DMark CPU score. So whats that tell ya?
This is a plain fact, and it can not be denied, AMD fans please accecpt it.
Quad 6600 > Phenom 9750 clock vs clock
And Nehalem > QX6800 clock vs clock
=> Nehalem > Phenom clock vs clock.
Dude, this is so easy to understand.
I don't know if these benchmarks are good or not. I looked over at my 3DMark 06 and PCMark 05 benchmarks for my new PC (E8400, P35 motherboard, 8800GT 512MB, 2GB DDR2-1066) and the Nehalem system beats mine by like 10%-15%. Sure, I'm not using the same OS, but unless Vista cripples the 4850 to 8800GT levels, then wow.
Based on the phenom score of 3.5% increase for 4% more clock you could predict a score of 8695 at 3ghz for phenom. Still a good choice for the money. Intel has a very good product now but I feel that if AMD could rework the core to get rid of L3 and double the L2 cache they would be competitive again due to lower latencies and a wider memory bus than they have now.