The Inquirer-Home

Athlon 64 3400+ model number criticism

Have some hardware review sites been overly critical?
Tue Jun 01 2004, 01:56
I WROTE an article last year about whether the AMD Athlon XP 3000+ processor had earned its model number rating, which highlighted the ongoing issues with benchmarks. This time round, however, the focus will be from the enthusiast perspective.

Since the launch of the AMD Athlon 64 last September, there has been what seems like a myriad of processor launches and accompanying reviews. AMD currently has five Athlon 64 processors that are available for the desktop - 2800+, 3000+, 3200+, 3400+, and the flagship FX model. New Athlon 64 additions should be launched tomorrow. Intel has responded to that onslaught by launching two Extreme Edition P4s, a 3.4 GHz Northwood P4, and Northwood's replacement - the 90 nm Prescott.

Let's not forget that Intel has now introduced model numbers for some of its product offerings. With that, when the next round of processor reviews are free of the NDA embargo, which is probably today, it will be interesting to see if there is any criticism of the model numbers used for AMD's newly launched Athlon 64 processors.

Praise for the 3400+
A large majority of hardware review sites have concluded that the Athlon 64 3400+ is a very hot part indeed, but naturally not its thermals - it's no fiery Prescott. Even though this 2.2 GHz device is just a 200 MHz speed bump improvement, when the pedal hit the metal and the rubber turned to smoke, reviewers heaped accolades on it.

Criticism of the 3400+
The Tech Report didn't seem totally happy when it said, "The benchmarks put the Athlon 64 3400+ just behind the Athlon 64 FX-51 in terms of overall performance, and I suppose AMD's "3400+" model number is warranted, at least for gamers, for whatever that's worth." (1).

HardcoreWare, which touts hardware reviews for hardcore gamers, felt somewhat let down when it said, "We put it against Intel's 3.2 GHz CPU which is now 6 months old, and while it performed well in many important applications - most notably gaming - it didn't dominate like I had hoped. It is still way behind in video and audio encoding, and rendering wasn't as good as I expected." (2).

Tom's Hardware (THG) came to a less than positive conclusion when it said, "A close look at all the benchmark results reveals that the new Athlon64 just barely earns the performance rating 3400+. Out of 32 benchmarks, only 13 were decided clearly in favor of AMD's new contender. If you were to evaluate each of the 41 individual disciplines, the result would be even poorer." (3).

Is this criticism fair?
So AMD's 3400+ processor wasn't totally approved across the board. But does that really resonate with the enthusiast?

In the car world, are those that buy exotic wheels really looking for a nicely rounded vehicle? If someone was looking to buy a Caterham Super 7, they're not going to be buying it for its creature comforts. (4). If another buys a Mercedes SL500, fuel economy is the very last thing on their mind. Those that buy a limousine clearly have no intention of being the driver. The point being made? Like those that fork out top dollar for their wheels, the PC enthusiast has one or two major requirements that will determine platform choice.

When enthusiasts have to make decisions about their next platform purchase, most of them won't have an unlimited budget to play with. So what really determines platform choice? If gaming performance and price are the two most important factors, then it should be possible to view the 3400+'s overall performance in a very favorable light. Also, but more importantly, we should also be able to corroborate that viewpoint as well.

The enthusiast weighted benchmark index
We'll use THG's Athlon 64 3400+ review as an example, since it was critical of the chip's performance and its benchmarks generally show the P4 in a good light; but this could be applied to any review. If we put THG's review benchmarks into their respective groups, and also add an appropriate weighting to reflect from an enthusiast standpoint the importance of each category, it becomes obvious that the 3400+ truly merits its rating.

As seen below, the gaming group gets a 4x weighting, so the 15 benchmark total becomes 60. Numbers in brackets denote more than one benchmark test. Encoding gets a 2x weighting, workstation and data compression a 1x weighting, and office productivity and synthetics a 0x weighting.

Why did the gaming group get a 4x weighting? Because from an enthusiast standpoint it is by far the biggest category. Don't agree? Just Google "gaming PCs" and compare it to any type of encoding PC search. Sysmark received a zero weighting because that benchmark has been derided so often. The synthetic tests from Sisoft Sandra and PCMark don't translate into real world performance, so they too received a zero weighting.

Quake 3 (3)
Serious Sam
Return to Castle Wolfenstein
Comanche 4
UT 2K3
Splinter Cell
Warcraft III
Gun Metal
Aquamark 3
3D Mark 2K1
3D Mark 2K3 (2)
Total of 15 x 4 = 60
Athlon 64 3400+ = 44
P4 3.2 GHz = 16

Main Concept
Pinnacle Studio 8.8
Xmpeg + DivX 5.1
Windows Media Encoder 9
MS Movie Make 2
Magix MP3 2K4
Lame MP3
Steinberg Nuendo 2
Total of 8 x 2 = 16
Athlon 64 3400+ = 0
P4 3.2 GHz = 16

SPECviewperf 7.1a (5)
Newtek lightwave 7.5
Cinema 4D XL R8.1
3D Studio Max 5.1
Total of 9 x 1 = 9
Athlon 64 3400+ = 4
P4 3.2 GHz = 5

Data compression:
Winrar 3.2
Total of 1 x 1 = 1
P4 3.2 GHz = 1

Office productivity:
Sysmark 2K2 (3)
Total of 3 x 0 = 0

PCMark 2K2 (2)
Sisoft Sandra Max3 (3)
Total of 5 x 0 = 0

The first thing to note in the gaming group is that the 3400+ won almost 75% of the disciplines. For many enthusiasts that would be enough to warrant the model number rating. When one also factors in that many would consider it the gaming equal to the 3.2 GHz P4EE as well, who is going to seriously tell the gaming enthusiast that the 3400+ barely earns its rating?

When everything is tallied together, the 3400+ comes out with 48 points and the P4 with 38, which is 56 and 44 percent of the weighted total. So when the weighted index is brought to bear, the 3400+ more than meets its requirement.

Does the weighted index
concur with the collective judgment?

For those that were critical of the 3400+'s performance, the weighted index certainly paints the 3400+ in a very different light. But can that be shown to be representative? One doesn't have to look very far to find out.
THG's own 2K3 Readers' Choice Awards voted the Athlon 64 FX-51 the best innovation in CPUs. (5). Its lead over the second placed P4 was 19.5%. From that it can be inferred that the 3400+, which offers comparable performance at over $300 less, would have won THG's award if it had been available at the time of the poll, and would probably have increased the winning margin as well.
X-bit labs had its own readership awards which corroborated THG's readership viewpoint as well, but to a far greater degree. AMD garnered two-thirds of the vote to overwhelmingly win the best CPU maker of 2003. Athlon 64 was no doubt the catalyst for that winning margin. Intel had to settle for the remaining third of the poll. (6).

The 64-bit desktop
Now all of this discussion has been based around the 32-bit world of today. The 64-bit desktop world of tomorrow hasn't even been raised. So let's open it with an interesting SuSE Linux quote about its 64-bit AMD64 OS.

"Moreover, 64-bit computers have larger caches and a more efficient memory access, which further increases the system speed. Example: A computer with an AMD Athlon(tm) 64 processor with 1.8 GHz is faster than a 32-bit computer with a Pentium(tm) 4 processor with 3.2 GHz." (7).

It seems evident that SuSE was talking in very general terms. So let's convert this 1.8 GHz into a representative model number rating, and then extrapolate the performance gains that a 64-bit OS should deliver. Conveniently, AMD has already launched a 1.8 GHz Athlon 64 processor, which has a 2800+ model number rating. So when referenced to the 3.2 GHz P4, the Athlon 64 2800+ is understated by at least two speed grades when using a 64-bit OS.

Reports suggest that Intel's x86-64 implementation isn't as good as AMD's. One report says it “sucks”. If that proves to be the case, the performance gap between Athlon 64 and Prescott P4 - when both are in 64-bit mode - could be greater than the two speed grade extrapolation mentioned above, which may explain why Intel hasn't yet turned on the 64-bit functionality in Prescott.

Paul Thurrot's SuperSite for Windows interviewed Microsoft's Bob Muglia, senior VP of the Windows Server Division, and Windows Server senior director Jeff Price. They were asked about the differences between Intel's and AMD's x86-64 implementation. (8).

Paul: Are you seeing any difference between AMD's [64-bit] stuff and Intel's stuff?
BM: Yes. [Smiles]
Paul: Would you care to clarify that? [Laughs]
BM: Well, AMD has done a good job ...

Enough said. The interview has other interesting tidbits.

So as things stand today, those looking to buy a x86-64 capable client platform will have to buy an AMD64 based solution. If it proves to be the case that AMD's x86-64 implementation is far superior than Intel's, then those that have already bought their Athlon 64 systems will be smiling like Cheshire cats.

Whatever one may say about the 64-bit desktop, it has arrived with Linux. And even though there are too few benchmarks to measure its actual performance, SuSE has clearly indicated the 64-bit performance advantage that Athlon 64 delivers over the 32-bit only P4. If Microsoft's 64-bit OS delivers anything close to that speed boost, Athlon 64 users are going to have more to look forward to.

The Wintel partnership is still very tight
While SuSE touts the benefits of its 64-bit OS on Athlon 64, Microsoft is still touting the Intel technologies of the past. Microsoft has a gaming PC roundup page dated November 3 last year, which was just a tad out of date about the PC gaming technology available at that time. Maybe someone from AMD can remind Microsoft that the gaming world changed after the September launch of the Athlon 64 processor. That page readily and dutifully plugs the P4 and HyperThreading, but there is no mention of the better performing Athlon 64. Not exactly a balanced and informative read, and this from a supposedly AMD64 partner. (9).

Athlon 64 is holding all
the gaming platform aces

For the enthusiast that is concerned about the environment, AMD's socket 754 platform and the upcoming socket 939 version can address that concern. Motherboards that support AMD's Cool 'n' Quiet technology have the ability to dynamically lower the processor's frequency and voltage when the workload doesn't require it, which will please those who run their PCs 24/7. So when the processor is in this low power state, it doesn't produce as much heat, so the processor fan speed can be reduced as well. That not only lowers the decibel level on the ears, it will also reduce the electricity bill as well.

When one looks at what the Athlon 64 platform brings to the enthusiast, it really does look like the chipmaker is holding all the gaming platform aces. Best price/performance, a scalable roadmap, and a 64-bit future. With the dual channel socket 939 platform about to debut, which won't require ECC registered memory, that platform will no doubt improve performance even more.

Can Intel make a credible response?
Oh to be a fly on the wall at Intel. How can the chip giant wrest the price/performance gaming crown back from AMD? It doesn't look like it's going to happen anytime soon.

Before Prescott's launch, Intel made ambitious claims about its performance - better HyperThreading, SSE3 software support, encoding performance improvements as well, but nothing about better gaming performance. Now that Prescott has been launched, gaming performance is arguably worse than Northwood, and it definitely hasn't delivered the performance gains that Northwood did over Willamette.

Looking at the new technologies that Intel is going to release for the desktop space, will it make a discernible impact? PCI express, new chipsets, higher FSB bandwidth, and higher frequency processors with more cache - will it be enough?

If AMD is able to stick to its roadmap then no. The AMD64 architecture has proven to be so very efficient and scalable. Athlon 64 just does more with less. When the socket 939 platform debuts, the non-FX devices will only need 512 KB of level two cache. And let's not forget that with the 90 nm shrink, AMD has the option of doubling the level two cache to 2 MB for its flagship FX model. Also, because Intel has to move to DDR2 before AMD needs to, AMD can fully exploit DDR1 and move to DDR2 at its leisure. So it should be clearly evident that when it comes to performance, AMD should have plenty in reserve.

What does all of this mean?
As Intel understands all too well, perception in this industry is everything. Since the launch of the Opteron processor over a year ago, AMD has already demonstrated continued two and four processor 32-bit server performance leadership. It seems apparent from a price/performance perspective that AMD will do the same in the enthusiast space as well.

At this year's CeBIT, an industry source said it will take Intel 18 months to catch AMD up. But as I've said already, AMD won't be taking a vacation while Intel is pulling out all the stops.

So from a perception standpoint it's not looking good for Intel. As I said last June using a naval metaphor, "AMD is blazing broadside with 16 inch guns. Intel can only respond with guns that are half the size."

I also said in the same article, but I've changed the context to reflect the P4:

The P4 is hemorrhaging performance because its northbridge based architecture is out of date. Athlon 64's ability to scale has made today's P4 yesterday's technology. Intel will have to pull the proverbial rabbit out of the hat if it's to rescue the P4 from its predicament, or its enthusiast business will suffer erosion as the market votes with its feet.

If Intel sees serious Athlon 64 penetration in the enthusiast space, it does have one very important card that it can play - price. The 90 nm process and 300 mm size wafers will contribute very significantly to Intel's economies of scale. If the chip giant is forced to play that card, all enthusiasts will benefit from lower platform costs. If that should happen, price competition will make 2004 an even more interesting year.

See Also
Did AMD's Athlon XP 3000+ earn its rating?
AMD's 64-bits better than Intel's 64-bits - Microsoft
Intel Nocona 64-bit mode "sucks”
Intel "will take 18 months" to catch AMD up



Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?